EE 840 Mathematical Methods in Engineering

- 1 -
Eigenvalues and Eigenvectors

1. State-Space Models
Dynamic behaviour of systems can be modeled by differential equations. In general,
nonlinear differential equations are required to model actual dynamic systems. In many
cases, however, linear approximations can be obtained to describe the dynamic behaviour
of such systems in the vicinity of some nominal operating point. The mathematical
models of some simple linear dynamic systems are described below.

1.1 A Series RLC Electrical Circuit
Figure 1 shows a series circuit connecting a resistance R, an inductance L and a
capacitance C. A voltage ( (( ( ) )) ) t v
i
is applied to the circuit which results a loop current ( (( ( ) )) ) t i .
The dynamic response of this circuit, ( (( ( ) )) ) t i can be described by a differential equation
using Kirchhoffs voltage law. According to Kirchhoffs law, voltage applied in a
closed loop at any instant must be equal to the sum of all voltage drops around that loop.
( (( ( ) )) ) t v
i
( (( ( ) )) ) t i

Figure E 1.1 A series RLC circuit
Therefore,
( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) t v t v t v t v
C L R i
+ ++ + + ++ + = == = (E 1.1)
Where
( (( ( ) )) ) t v
R
= voltage drop in the resistance, R at time t
( (( ( ) )) ) t v
L
= voltage drop in the inductance, L at time t and
( (( ( ) )) ) t v
C
= voltage drop in the capacitance, C at time t.
( (( ( ) )) ) ( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )

+ ++ + + ++ + = == = dt t i
C dt
t di
L t Ri t v
i
1
(E 1.2)
Differentiate Eqn. (E.2) with respect to time. The mathematical model of the system is:
( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) )
C
t i
dt
t i d
L
dt
t di
R
dt
t dv
i
+ ++ + + ++ + = == =
2
2
(E 1.3)

1.2 An Automobile Suspension System
The spring-mass-damper system shown in Figure E 1.2 can be used to represent a
simplified suspension system for each wheel of an automobile.
EE 840 Mathematical Methods in Engineering
- 2 -

Figure E 1.2 An automobile suspension system
A force applied to the mass, M in the direction of x will be balanced by three other
forces in the opposite direction of x. The balancing forces are the gravitational force on
the mass, the damping force due to the shock absorber and the spring force.
The damping force is: ( (( ( ) )) )
( (( ( ) )) )
dt
t dx
B t F
B
= == = (E 1.4)
The spring force is: ( (( ( ) )) ) ( (( ( ) )) ) t Kx t F
K
= == = (E 1.5)
The gravitational force is: Mg
Applying Newtons Second Law we can write
( (( ( ) )) ) ( (( ( ) )) )
( (( ( ) )) )
2
2
dt
t x d
M t F t F Mg
B K
= == =
or,
( (( ( ) )) ) ( (( ( ) )) )
( (( ( ) )) ) Mg t Kx
dt
t dx
B
dt
t x d
M = == = + ++ + + ++ +
2
2
(E 1.6)

1.3 A Two-tank Liquid Level System
Consider the two-tank liquid level system shown in Figure E 1.3.

Figure E 1.3 A two-tank liquid level system
EE 840 Mathematical Methods in Engineering
- 3 -
Assume linearized flow rates. Therefore, the flow rates can be considered
proportional to the corresponding liquid heights. If
1
R and
2
R are constant resistance
factors of the control valves
1
C and
2
C then
( (( ( ) )) )
( (( ( ) )) ) ( (( ( ) )) )
1
2 1
1
R
t h t h
t q

= == = (E 1.7)
( (( ( ) )) )
( (( ( ) )) )
2
2
2
R
t h
t q = == = (E 1.8)
If ( (( ( ) )) ) t V
1
and ( (( ( ) )) ) t V
2
are volumes of liquid in tanks 1 and 2 respectively and ( (( ( ) )) ) t q
i
is the
inflow in the first tank then
( (( ( ) )) ) ( (( ( ) )) ) { {{ { } }} }
( (( ( ) )) ) ( (( ( ) )) ) t q t q
dt
t h A d
dt
t dV
i 1
1 1 1
= == = = == = (E 1.9)
( (( ( ) )) ) ( (( ( ) )) ) { {{ { } }} }
( (( ( ) )) ) ( (( ( ) )) ) t q t q
dt
t h A d
dt
t dV
2 1
2 2 2
= == = = == = (E 1.10)
where
1
A and
2
A are respective surface areas of tank 1 and 2.
Substituting Eqn. (E 1.7) in Eqn. (E 1.9) we get
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) ) ( (( ( ) )) )
) )) )
` `` `




= == =
1
2 1 1
1
R
t h t h
t q
dt
t dh
A
i
(E 1.11)
Substituting Eqns. (E 1.7) and (E 1.8) in Eqn. (E 1.10) we get
( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) )
2
2
1
2 1 2
2
R
t h
R
t h t h
dt
t dh
A
) )) )
` `` `




= == = (E 1.12)
Eqns. (E 1.11) and (E 1.12) mathematically model the dynamic behaviour this two-tank
liquid level system.

2. State-Space Representations
The dynamic behaviour of many physical systems can be modeled with the help of a
set of n first-order differential equations involving n state variables. These n state
variables completely define the internal behaviour of the system. The state variables are
usually denoted as ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) t x , t x , t x
n
L
2 1
. The state model of a system is not unique, but
depends on the choice of a set of state variables.
The general state-space model of a linear time-invariant system can be expressed as:
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






+ ++ +
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






= == =
( (( (
( (( (
( (( (
( (( (

( (( (






t u
t u
t u
B B
B B
B B
t x
t x
t x
A A A
A A A
A A A
t x
t x
t x
m nm n
m
m
n nn n n
n
n
n
M
L
M O M
L
L
M
L
M O M M
L
L
&
M
&
&
2
1
1
2 21
1 11
2
1
2 1
2 22 21
1 12 11
2
1
(E 2.1)
EE 840 Mathematical Methods in Engineering
- 4 -
Eqn. (E 2.1) can be written in a concise form as
( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] t u B t x A t x + ++ + = == = & (E 2.2)
where [ [[ [ ] ]] ] A is the n n state matrix and [ [[ [ ] ]] ] B is the m n input matrix.
It is important to note at this point that although the entire n-dimensional state vector
( (( ( ) )) ) [ [[ [ ] ]] ] t x characterizes the complete internal behaviour of a state-space system, it is usually
impossible to directly measure or observe all n components of ( (( ( ) )) ) [ [[ [ ] ]] ] t x . The external
output vector can be expressed as
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (
( (( (

( (( (







= == =
( (( (
( (( (
( (( (
( (( (
( (( (

( (( (







t x
t x
t x
C C
C C
C C
t y
t y
t y
n pn p
n
n
p
M
L
M O M
L
L
M
2
1
1
2 21
1 11
2
1
(E 2.3)
Eqn. (E 2.3) written in a concise form as
( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] t x C t y = == = (E 2.4)
where [ [[ [ ] ]] ] C is the n p output matrix.
Eqns. (E 2.2) and (E 2.4) are the state-space representation of a linear, time-invariant
system.
3. Transfer Function Matrices and Stability
Transfer function is utilized to determine the stability of a dynamic system. For an n-
dimensional dynamic system the transfer function matrix can be found by applying
Laplace transform to Eqns. (E 2.2) and (E 2.4). The Laplace transform of a vector is the
vector of the Laplace transform of its elements. The Laplace transform of ( (( ( ) )) ) [ [[ [ ] ]] ] t x and
( (( ( ) )) ) [ [[ [ ] ]] ] t x& are as follows:
( (( ( ) )) ) [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ]
( (( ( ) )) ) { {{ { } }} }
( (( ( ) )) ) { {{ { } }} }
( (( ( ) )) ) { {{ { } }} }
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( (
( (( (
( (( (
( (( (

( (( (






= == =
( (( (
( (( (
( (( (
( (( (

( (( (






= == = = == =
s X
s X
s X
t x L
t x L
t x L
t x L s X
n n
M M
2
1
2
1
( (( ( ) )) ) [ [[ [ ] ]] ]
( (( ( ) )) ) { {{ { } }} }
( (( ( ) )) ) { {{ { } }} }
( (( ( ) )) ) { {{ { } }} }
( (( (
( (( (
( (( (
( (( (

( (( (






= == =
t x L
t x L
t x L
t x L
n
&
M
&
&
&
2
1

where
( (( ( ) )) ) { {{ { } }} } ( (( ( ) )) ) ( (( ( ) )) ) 0
i i i
x s sX t x L = == = &
Therefore, the Laplace transform of ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] t u B t x A t x + ++ + = == = & is
( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] s U B s X A x s X s + ++ + = == =
0

or, [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ]
0
x s U B s X A I s + ++ + = == = (E 3.1)
where [ [[ [ ] ]] ] I is the unit matrix.
The Laplace transform of ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] t x C t y = == = is ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] s X C s Y = == =
EE 840 Mathematical Methods in Engineering
- 5 -
Premultiplying both side of Eqn. (E 3.1) by [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )
1
A I s we get
( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ]
0
1 1
x A I s s U B A I s s X

+ ++ + = == =
The output in Laplace domain is
( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ]
0
1 1
x A I s C s U B A I s C s Y

+ ++ + = == =
( (( ( ) )) ) [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ]
0
1
x A I s C s U s G s Y

+ ++ + = == =
where ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] B A I s C s G
1
= == = is called the transfer function matrix.
A transfer function matrix relates the transforms of the input and output vectors for
zero initial conditions.
The inverse of matrix [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) A I s can be expressed as
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )
[ [[ [ ] ]] ] [ [[ [ ] ]] ] A I s
A I s
A I s


= == =

adj
1
(E 3.2)
Using Eqn. (E 3.2), the transfer function matrix can be written as
( (( ( ) )) ) [ [[ [ ] ]] ]
[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )[ [[ [ ] ]] ]
[ [[ [ ] ]] ] [ [[ [ ] ]] ] A I s
B A I s
s G


= == =
adj C
(E 3.3)
The dynamic system modeled by Eqns. (E 2.2) and (E 2.4) is stable if all the roots of
the characteristic equation [ [[ [ ] ]] ] [ [[ [ ] ]] ] 0 = == = A I s lie on the left half of the s-plane. The roots of
the characteristic equation, [ [[ [ ] ]] ] [ [[ [ ] ]] ] 0 = == = A I s are called the eigenvalues of the system
coefficient matrix (system matrix) [ [[ [ ] ]] ] A . That means the eigenvalues of a stable system
are all negative.
Example: Consider a 2-dimensional system with the following system and input
matrices.
[ [[ [ ] ]] ]
( (( (

( (( (




= == =
3 2
1 0
A [ [[ [ ] ]] ]
( (( (

( (( (



= == =
0
1
B and [ [[ [ ] ]] ] [ [[ [ ] ]] ] 0 1 = == = C
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )
( (( (

( (( (



+ ++ +

= == =
3 2
1
s
s
A I s
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )
[ [[ [ ] ]] ] [ [[ [ ] ]] ]
( (( (

( (( (




+ ++ +
+ ++ + + ++ +
= == =


= == =

s
s
s s A I s
A I s
A I s
2
1 3
2 3
1 adj
2
1

The transfer function of the system is
( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] B A I s C s G
1
= == =
( (( ( ) )) ) [ [[ [ ] ]] ]
[ [[ [ ] ]] ]
( (( ( ) )) )( (( ( ) )) ) 2 1
3
0
1
2
1 3
2 3
0 1
2
+ ++ + + ++ +
+ ++ +
= == =
( (( (

( (( (



( (( (

( (( (




+ ++ +
+ ++ + + ++ +
= == =
s s
s
s
s
s s
s G
EE 840 Mathematical Methods in Engineering
- 6 -
The denominator of the transfer function is
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )( (( ( ) )) ) 2 1 2 3
2
+ ++ + + ++ + = == = + ++ + + ++ + = == = s s s s A I s
The roots of the characteristic equation, i.e., the eigenvalues are -1 and -2 and therefore,
the system is stable.
4. Coordinate Transformation and Eigenvalues
It has been mentioned earlier that the state model of a dynamic system is not unique.
The model depends on the choice of a set of state variables. It is often beneficial to
define a new state vector [ [[ [ ] ]] ] z with the help of a coordinate transformation [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] z T x = == = .
The state model of the system described by [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] u B x A x + ++ + = == = & and [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] x C y = == =
will be transformed by this coordinate transformation in the following manner.
[ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] u B z T A z T + ++ + = == = & (E 4.1)
[ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ][ [[ [ ] ]] ] z T C y = == = (E 4.2)
Premultiply Eqn. (E 4.1) by [ [[ [ ] ]] ]
1
T .
[ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] u B T z T A T z T T
1 1 1
+ ++ + = == = &
[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] u B T z T A T z
1 1
+ ++ + = == = &
The transformed system (with new state vector [ [[ [ ] ]] ] z ) is
[ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] u B z A z

+ ++ + = == = & [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] z C

y = == =
where [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] T A T A
1

= == = and [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] T C C

= == =
The use of the new state vector (internal state variables) should not affect the stability
(i.e., eigenvalues) and the input-output relation of the system. However, this can be
verified by evaluating the characteristic equation and the transfer function matrix of the
transformed system in the following manner.
Characteristic equation:
[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] T A T I T T s T A T I s A I s
1 1 1

= == = = == =
Because, [ [[ [ ] ]] ] I is the unit matrix we can write
[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )[ [[ [ ] ]] ] T A I s T T A T T I T s A

I s = == = = == =
1 1 1

Remember the properties of determinants.
[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] T A I s T T A I s T A

I s = == = = == =
1 1

[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) A I s A

I s = == =
So the characteristic equation remains unchanged.
EE 840 Mathematical Methods in Engineering
- 7 -
Transfer function matrix:
The transfer function matrix of the transformed system is
( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ][ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] B T T A T I s T C B

I s C

s G

1
1
1
1




= == = = == =
[ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )[ [[ [ ] ]] ] { {{ { } }} } [ [[ [ ] ]] ] [ [[ [ ] ]] ] B T T A I s T T C
1
1
1


= == =
[ [[ [ ] ]] ][ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] B T T A I s T T C
1 1 1
= == =
[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] G B A I s C = == = = == =
1

The transfer function remains unaffected by the transformation.
The question is, what kind of transformation would be helpful? The particular
transformation that we are interested is the one for which [ [[ [ ] ]] ] A

would be a diagonal
matrix. Why we want a coefficient matrix that is diagonal? To answer this question, let
us have another look at the system state equations, particularly Eqn. (E 2.1).
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






+ ++ +
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






= == =
( (( (
( (( (
( (( (
( (( (

( (( (






t u
t u
t u
B B
B B
B B
t x
t x
t x
A A A
A A A
A A A
t x
t x
t x
m nm n
m
m
n nn n n
n
n
n
M
L
M O M
L
L
M
L
M O M M
L
L
&
M
&
&
2
1
1
2 21
1 11
2
1
2 1
2 22 21
1 12 11
2
1

In order to solve for each state, we have to integrate each equation. Integrating the
second term on the right-hand side of the equation should be done with little difficulty.
However, the integration of the first term on the right-hand side cannot be done that
easily. Because, we will have to wait for the determination of other state variables. Now
consider a coefficient matrix, [ [[ [ ] ]] ] A with only diagonal elements, all other elements being
zero.
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( ( ) )) )
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






+ ++ +
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






= == =
( (( (
( (( (
( (( (
( (( (

( (( (






t u
t u
t u
B B
B B
B B
t x
t x
t x
A
A
A
t x
t x
t x
m nm n
m
m
n nn n
M
L
M O M
L
L
M
L
M O M M
L
L
&
M
&
&
2
1
1
2 21
1 11
2
1
22
11
2
1
0 0
0 0
0 0

Each individual equation as shown in the matrix form can be expressed in terms of its
self term only. Therefore, each equation can be integrated with little difficulty.

( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) )
( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) )
( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) )
( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) t u B t u B t u B t x A t x
t u B t u B t u B t x A t x
t u B t u B t u B t x A t x
t u B t u B t u B t x A t x
m nm n n n nn n
m m
m m
m m
+ ++ + + ++ + + ++ + + ++ + = == =
+ ++ + + ++ + + ++ + + ++ + = == =
+ ++ + + ++ + + ++ + + ++ + = == =
+ ++ + + ++ + + ++ + + ++ + = == =
L &
M M M
L &
L &
L &
2 2 1 1
3 2 32 1 31 3 33 3
2 2 22 1 21 2 22 2
1 2 12 1 11 1 11 1
(E 4.3)
EE 840 Mathematical Methods in Engineering
- 8 -
The set of Equations (E 4.3) represents a system model that contains only the self
terms. All mutual terms are zero. This type of system is often called a decoupled system.
It is relatively very straightforward to carry out mathematical analysis of a decoupled
system.
We are, therefore, interested in a particular type of coordinate transformation that
would transform our coupled system into a decoupled system. We have already
discussed that the transfer function matrix and the characteristic equation would remain
unchanged even after this transformation. Assume that this transformation can be
achieved with the help of a transformation matrix [ [[ [ ] ]] ] M .
Therefore, [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] z M x = == = and [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] { {{ { } }} }
n
, , , M A M A

L
2 1
1
diag = == = = == = = == =


The characteristic equation of the system is

[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) ( (( ( ) )) ) ( (( ( ) )) ) [ [[ [ ] ]] ] { {{ { } }} }
( (( ( ) )) )( (( ( ) )) ) ( (( ( ) )) )
n
n
s s s
s , s , s I s A I s


= == =
= == = = == =
L
L
2 1
2 1
diag det

The elements of the diagonal matrix [ [[ [ ] ]] ] are the roots of the characteristic equation
[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] 0 = == = = == = I s A I s and are therefore the eigenvalues of the system.
Now the question is what would be the elements of such a transformation matrix [ [[ [ ] ]] ] M ?
[ [[ [ ] ]] ]
( (( (
( (( (
( (( (
( (( (

( (( (






= == =
nn n n
n
n
m m m
m m m
m m m
M
L
M O M M
L
L
2 1
2 22 21
1 12 11

[ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] = == =

M A M
1

[ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] M M A M M = == =
1

[ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] M M A = == =
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






= == =
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






n nn n n
n
n
nn n n
n
n
nn n n
n
n
m m m
m m m
m m m
m m m
m m m
m m m
A A A
A A A
A A A

0 0 0
0 0 0
0 0 0
0 0 0
2
1
2 1
2 22 21
1 12 11
2 1
2 22 21
1 12 11
2 1
2 22 21
1 12 11
O
L
M O M M
L
L
L
M O M M
L
L
L
M O M M
L
L

Equate columns on both sides. Equating the 2nd column we can write
( (( (
( (( (
( (( (
( (( (

( (( (






= == =
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






2
22
12
2
2
22
12
2 1
2 22 21
1 12 11
n n nn n n
n
n
m
m
m
m
m
m
A A A
A A A
A A A
M M
L
M O M M
L
L

EE 840 Mathematical Methods in Engineering
- 9 -
In the same manner equating the ith column we can write
( (( (
( (( (
( (( (
( (( (

( (( (






= == =
( (( (
( (( (
( (( (
( (( (

( (( (






( (( (
( (( (
( (( (
( (( (

( (( (






ni
i
i
i
ni
i
i
nn n n
n
n
m
m
m
m
m
m
A A A
A A A
A A A
M M
L
M O M M
L
L
2
1
2
1
2 1
2 22 21
1 12 11

(E 4.4)

[ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ]
i
i
i
m m A = == = (E 4.5)
[ [[ [ ] ]] ] [ [[ [ ] ]] ][ [[ [ ] ]] ] 0 = == =
i i
i
m A m
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )[ [[ [ ] ]] ] 0 = == =
i
i
m A I (E 4.6)
Eqn. (E 4.6) can be solved to find the ith column of the transformation matrix [ [[ [ ] ]] ] M .
Observe that associated with each eigenvalue there is a corresponding column or vector
for the transformation matrix [ [[ [ ] ]] ] M . These vectors are called the eigenvectors of the
system. A physical (geometric) interpretation of an eigenvector can be derived from Eqn.
(E 4.5). The operation by the system matrix on an eigenvector is equivalent of changing
the magnitude and direction of the vector by a factor of its corresponding eigenvalue. If
the eigenvalue is positive, the pre-multiplication operation simply changes the magnitude
of the vector. If the eigenvalue is negative, the operation in addition to changing the
magnitude reverses the direction of the resulting vector. The magnitude is amplified by a
factor equal to the absolute value of the corresponding eigenvalue. From Eqn. (E 4.6) it
becomes obvious that any multiple of the eigenvector [ [[ [ ] ]] ]
i
m will satisfy the equation.
Therefore, any multiple of [ [[ [ ] ]] ]
i
m is also an eigenvector.
The transformation matrix [ [[ [ ] ]] ] M that diagonalizes the system matrix [ [[ [ ] ]] ] A is called the
modal matrix.
Example: Find the eigenvalues and eigenvectors
[ [[ [ ] ]] ]
( (( (

( (( (





= == =
0 1
2 3
A
Characteristic equation is
[ [[ [ ] ]] ] [ [[ [ ] ]] ] 0 = == = A I
0
0 1
2 3
0
0
= == =
( (( (

( (( (






( (( (

( (( (

0
1
2 3
= == =
+ ++ +

0 2 3
2
= == = + ++ + + ++ +
EE 840 Mathematical Methods in Engineering
- 10 -
( (( ( ) )) )( (( ( ) )) ) 0 1 2 = == = + ++ + + ++ +
The eigenvalues are 2
1
= == = and 1
2
= == =
The eigenvector [ [[ [ ] ]] ]
1
m corresponding to the eigenvalue 2
1
= == = is related in the following
manner.
[ [[ [ ] ]] ] [ [[ [ ] ]] ]
1
1
1
0 1
2 3
m m = == =
( (( (

( (( (






( (( (

( (( (



= == =
( (( (

( (( (



( (( (

( (( (





21
11
21
11
2
0 1
2 3
m
m
m
m

21 11
2m m = == =
[ [[ [ ] ]] ]
( (( (

( (( (



= == =
( (( (

( (( (



= == =
1
2
21
11 1
m
m
m
The eigenvector [ [[ [ ] ]] ]
2
m corresponding to the eigenvalue 1
2
= == = is related in the following
manner.
[ [[ [ ] ]] ] [ [[ [ ] ]] ]
2
2
2
0 1
2 3
m m = == =
( (( (

( (( (






( (( (

( (( (



= == =
( (( (

( (( (



( (( (

( (( (





22
12
22
12
1
0 1
2 3
m
m
m
m

22 12
m m = == =
[ [[ [ ] ]] ]
( (( (

( (( (



= == =
( (( (

( (( (



= == =
1
1
22
12 2
m
m
m
The modal matrix is
[ [[ [ ] ]] ]
( (( (

( (( (



= == =
1 1
1 2
M
Now we will discuss another method of finding eigenvectors. For this method we will
use Eqn. (E 4.6) again.
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )[ [[ [ ] ]] ] 0 = == =
i
i
m A I
We know that [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ] I K K = == =
1
and also [ [[ [ ] ]] ]
[ [[ [ ] ]] ]
K
K
K
adj
1
= == =


Therefore, [ [[ [ ] ]] ][ [[ [ ] ]] ] [ [[ [ ] ]] ]
[ [[ [ ] ]] ]
[ [[ [ ] ]] ] I
K
K
K K K = == = = == =

adj
1

EE 840 Mathematical Methods in Engineering
- 11 -
[ [[ [ ] ]] ] [ [[ [ ] ]] ] [ [[ [ ] ]] ] I K K adj K = == =
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) )[ [[ [ ] ]] ] I A I A I A I
i i i
= == = adj
However, [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) A I
i
is equal to zero since
i
is an eigenvalue.
Therefore, we can write
[ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) 0 adj = == = A I A I
i i
(E 4.7)
A comparison of Eqns. (E 4.6) and (E 4.7) show that the eigenvector [ [[ [ ] ]] ]
i
m is proportional
to any nonzero column of [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) A I
i
adj .
We can, therefore, use [ [[ [ ] ]] ] [ [[ [ ] ]] ] ( (( ( ) )) ) A I
i
adj to obtain the modal matrix.
Example:
Find the eigenvalues and eigenvectors for the system with a system matrix A.
A
5
0
13
1
4
1
0
1
0
|

\
|
|
|

:=


eigenvals A ( )
6.1
2.339
0.561
|

\
|
|
|

= eigenvals A ( ) :=
6.1
2.339
0.561
|

\
|
|
|

=


D
1

1
identity 3 ( ) A :=

D
2

2
identity 3 ( ) A :=

D
3

3
identity 3 ( ) A :=

1
identity 3 ( ) A
1.1
0
13
1
2.1
1
0
1
6.1
|

\
|
|
|

=

adj D
1
( ) D
1
D
1
1
:=


adj D
1
( )
11.554
12.714
26.705
5.966
6.566
13.791
0.978
1.076
2.261
|

\
|
|
|

=

adj D
2
( )
4.825
12.84
21.328
2.31
6.147
10.211
0.988
2.628
4.366
|

\
|
|
|

=


adj D
3
( )
3.332
14.79
50.867
0.638
2.832
9.739
1.138
5.051
17.37
|

\
|
|
|

=


EE 840 Mathematical Methods in Engineering
- 12 -
The modal matrix [ ] M is
[ ]
(
(
(

=
37 17 366 4 261 2
051 5 628 2 076 1
138 1 988 0 978 0
. . .
. . .
. . .
M

M
1
A M
6.1
0
0
0
2.339
0
0
0
0.561
|

\
|
|
|

=


References:
Feedback Control Systems, John Van de Vegte, Prentice Hall
Automatic Control Systems: Basic Analysis and Design, William A. Wolovich, Saunders
College Publishing