Sie sind auf Seite 1von 17

CEIC3000 Process Modelling and Analysis

Week 1
Session 1, 2012
A/Prof. Jie Bao

1 About the subject


1.1 What are process models
This subject is about process modelling, analysis and optimisation. The ever increasing global demand
for better utilisation of raw material and energy resources in an environmentally-sustainable manner
means that greater ingenuity must be employed in the development of new processes (or improvement of
existing ones). Thus, the traditional but time-consuming and expensive practice of rst building a small-
scale replica of the process for the sole purpose of investigating its dynamic and steady-state behaviour
to acquire optimal operating conditions is a luxury that can no longer be aorded. Consequently, a more
quantitative approach to process design and operation is now more attractive to the industry. Process
modelling, analysis and optimisation deals with the principles and methods for eectively carrying out
this exercise and as a result, this course is a central component of the chemical and process engineering
curriculum. Specically, the objectives of the course are to:
Provide you with key mathematical methods useful for the analysis of wide-ranging chemical
engineering problems.
Acquaint you with the dierent techniques to mathematical model building, for example, phe-
nomenological and empirical approaches.
Integrate results from modelling exercise with known physicochemical features of the system to
evolve better process design ad operation.
Indeed, this course allows you to rene and couch your knowledge of other core subjects in a
more analytical fashion therefore making it possible to consolidate your understanding and even draw
extensions to other disciplines.
In the rst part of this subject, we will study process modelling. That is to describe a dynamic
process using mathematical equations. We will also learn some basic analysis methods to analyse process
models in the context of systems theory. Before proceeding, let us look at the denition of models.
The Webster dictionary denes the term model as: A small representation of a planned or existing
object. From McGraw Hill Dictionary of Scientic and Technical Terms, we can nd the following
denition: A mathematical or physical system, obeying certain specied conditions, whose behaviour is
used to understand a physical, biological or social system to which it is analogous in some way.
In this class, we refer to a mathematical model:

Last modied Tuesday, February 28, 2012 14:59


1
A process model is a set of equations (including the necessary input data to solve the equations) that
allows us to predict the behaviour of chemical process system.
While you have been developing mathematical equations to describe chemical processes since Year
1, in this subject, we focus on the dynamics of processes. That is, the process behaviour as a function
of time.
Example 1 (Gravity tank) Process variables are function of time during non-steady state. For ex-
ample, a gravity tank, its liquid level, outlet ow rate may change with time. To represent the dynamics
of the process, an dierential equation is often used:
A
dh(t)
dt
= F
i
(t) k

h(t) (1)
F
o
(t) = k

h(t) (2)
where k is the valve coecient.
1.2 Why modelling
Many reasons for developing process models:
Improving or understanding chemical process operation is a major overall objective for develop-
ing a dynamic process model. These models are often used for the following purposes:
Operator Training. The people responsible for the operation of a chemical manufacturing pro-
cess are known as process operators. A dynamic process model can be used to perform simulations
to train process operators, in the same fashion that ight simulators are used to train airplane
pilots. Process operators can learn the proper response to upset conditions, before having to
experience them on the actual process.
Process Design. A dynamic process model can be used to properly design chemical process
equipment for a desired production rate. For example, a model of a batch chemical reactor can
be used to determine the appropriate size of the reactor to produce a certain product at a desired
rate.
Safety. Dynamic process models can also be used to design safety systems. For example, they
can be used to determine how long it will take after a valve fails for a system to reach a certain
pressure.
Control System Design. This is something really exciting! Youll learn this next session in
my process control subject. Feedback control systems are used to maintain process variables at
desirable values. For example, a control system may measure a product temperature (an output)
and adjust the steam owrate (an input) to maintain that desired temperature. For complex
systems, particularly those with many inputs and outputs, it is necessary to base the control
system design on a process model. Also, before a complex control system is implemented on a
process, it is normally tested by simulating the expected performance using computer simulation.
1.3 Basic elements of a model
A process model is a set of mathematical equations which predicts certain unknown variables we are
interested in from the initial conditions and some independent variables. The unknown variables we are
interested in are called model output, the independent variables are called model input. Therefore
a model is just a relation described in the language of mathematics, which maps input space to output
space. The output is the information you want to get from the model and the input is the information
2
you must specied before the problem can be solved. Then we can have the following block diagram
to represent the process model:
Process
Inputs: u
Outputs: y
States: x
Figure 1: Block diagram
Please note the input variables are not necessary the inlet mass ow/energy ow to
the system, the output variables are not necessary the outlet mass ow/energy ow out
of the system.
1.4 General form of dynamic models
Dynamic models have the following typical form:
_
_
_
dx
dt
= f (x, u)
y = g (x, u)
(3)
where x, u and y are the state, input and output vectors.
If the functions f (x, u) and g (x, u) are linear (i.e., they are the linear combinations of x and u),
then we have the following linear models
_
_
_
dx
dt
= Ax+Bu
y = Cx +Du
(4)
If the process has n states, p inputs and q outputs, then A R
nn
, B R
np
, C R
qn
, D R
qp
.
As most chemical processes are multivariable and dynamical, the main features of this
subjects are: (1) process dynamics; (2) multivariable system analysis.
2 Mathematical tools in multivariable system modelling
Most processes are multivariable in nature. That is, they have multiple input and multiple output.
Example 2 (Distillation column) The system under consideration is a distillation column as shown
in Figure 2. At the steady-state, the tray temperatures at trays 7 and 21 (T
7
and T
21
) are related to
liquid reux (L) and vapour boilup rates (V ).
_
T
7
T
21
_
=
_
33.89 32.63
18.85 34.84
_ _
L
V
_
. (5)
Vectors and matrices are compact notations for systems of linear equations. Therefore, vector and
matrix algebra are essential tool in engineering system analysis. We can use vectors to represent
process variables and matrices to represent linear relationships between input and output
variables (either steady-state or dynamical).
3
7
21
TC
1
TC
2
Figure 2: The Distillation Column
x
z
y
k a j a i a a
v r r
v
3 2 1
+ + =
Figure 3: A vector in Cartesian Coordinate
2.1 Vectors & vector operations - a review
Scalar quantities are measurements which are completely specied when the individual magnitude is
given. For example, mass, temperature, area, voltage, etc. However, when both direction and magnitude
are required for full specication, we say the quantity is a vector. For example, if we need to describe
the movement of an object and measure its velocity, if we need both the magnitude of the speed, but
also its moving direction, we need to specify the speed as a vector.
In the Cartesian coordinate system, we can express the speed with its direction as:
a = a
1
i +a
2
j +a
3
k (6)
where a
1
, a
2
, a
3
are real numbers.
a = a
1
i +a
2
j +a
3
k (7)
b = b
1
i +b
2
j +b
3
k (8)
c = c
1
i +c
2
j +c
3
k (9)
4
A vector can also be written as a row vector
a =
_
a
1
a
2
... a
m

(10)
or a column vector
a =
_
a
1
a
2
... a
m

T
=
_

_
a
1
a
2
.
.
.
a
3
_

_
(11)
2.1.1 Addition
Commutativity
a +b = b +a (12)
Associativity
(a +b) +c = a + (b +c) (13)
Distributivity
(a +b) = a +b (14)
( +) a = a +a (15)

a + (a) = 0 (16)
If all these algebraic operations are do-able and both a and b belong to a set of V (non-empty) then
V is said to be a real vector space. Thus,if a
1
, a
2
...a
m
are vectors in V then an expression of the form

1
a
1
+
2
a
2
+... +
m
a
m
(17)
is called a linear combination of these vectors. In particular, we note that the summation in Eq.
(17) a vector.
Linear Independence and Dependence
Now consider the equation

1
a
1
+
2
a
2
+... +
m
a
m
= 0 (18)
Clearly, this holds if we choose all

i
s zero, because then it becomes 0 = 0. If this is only m-tuple
of scalars for which (18) holds, then our vectors a
1
, a
2
...a
m
are said form a linearly independent
set or, more briey, we call them linearly independent. Otherwise, if (18) also holds with
scalars not all zero, we call these vectors linearly dependent, because then we can express (at
least) one of them as a linear combination of the others. For instance, if (18) holds with, say,

1
= 0, we can solve (18) for a
1
:
a
1
= (
2
a
2
+... +
m
a
m
) /
1
(19)
e.g.
a
1
=
_
1 0 2 2

T
(20)
a
2
=
_
2 1 0 1

T
(21)
a
3
=
_
3 2 2 4

T
(22)
5
are linearly dependent because
a
1
+ 2a
2
a
3
= 0 (23)
What is the point of linear independence and dependence?
Well, from a linearly dependent set we may often omit vectors that are linear combinations of
others until we are nally left with a linearly independent subset of the really essential vectors,
which can no longer be expressed linearly in terms of each other.
m independent vectors can be used as a base to form a vector space with the dimension of m. In
the case of real space, the space is denoted as R
m
.
2.1.2 Product
Inner product (dot product):
a =
_
a
1
a
2
... a
m

T
, b =
_
b
1
b
2
... b
m

T
then,
a b =
_
a
1
a
2
... a
m

_
b
1
b
2
.
.
.
b
m
_

_
(24)
=
m

k=1
a
k
b
k
= a
T
b (25)
a b = a
2
b
2
cos (26)
when a = 0,b = 0, where is the angle between a and b.
= cos
1
_
a b
a
2
b
2
_
(27)
a and b are orthogonal if
a b = 0 (28)
The angle between a and b ? (/2)
The inner product can be also represented as (a, b) or a, b .
From the denition of inner product, we have
1. Symmetry
a b = b a (29)
2. Linearity
[
1
a +
2
b] c =
1
a c +
2
b c (30)
3. Positive-deniteness
a a 0 (31)
a a = 0 only if a = 0 (32)
If two vectors a and b, its inner product is zero, then a and b are called orthogonal.
6
Figure 4: Cross product
Figure 5: Right hand triple of a, b,v
Quick question:
For the i, j, k dened before, what are the dot products of i j, i k, j k?
i j = 0 (33)
i i = 1 (34)
2-Norm (Euclidian norm) representing the length of a vector
a
2
=

a a 0
Norm is the largest absolute value of the elements in a vector: for a =
_
a
1
a
2
a
n

T
,
||a||

= max
n
i=1
{|a
i
|} .
Normalization: multivariable inputs and outputs are often normalized so that their norms
are more meaningful:
x

=
x x
min
x
max
x
min
(35)
In this case, x

is called the normalized variable for x. x

is also dimensionless.
Cross product
Denition
v = a b
|v| = |a| |b| sin
Properties of Cross Product
7
Anti-commutative:
a b = (b a) (36)
Non-associative:
a (b c) = (a b) c (37)
The parenthesis cannot be omitted.
Vector product in terms of components
Let
a = a
1
i +a
2
j +a
3
k (38)
b = b
1
i +b
2
j +b
3
k (39)
then
a b =

i j k
a
1
a
2
a
3
b
1
b
2
b
3

(40)
= a
2
b
3
i +a
1
b
2
k +a
3
b
1
j a
2
b
1
k a
1
b
3
j a
3
b
2
i (41)
Scalar Triple Product
(abc) = a (b c)
=

a
1
a
2
a
3
b
1
b
2
b
3
c
1
c
2
c
3

Properties
Scalar: (ka bc) = k (abc)
Commutative: a (b c) = (a b) c
Geometric Interpretation of Scalar Triple Products
|a (b c)| = |a| |b c| cos
where is the angle between a and (b c) , |a| cos is the height h. |b c| is the area of
the parallelogram.
Example 3 (Velocity of a rotating body) A rotating body B in space can be simply and uniquely
described by a vector w: The direction of w is that of the axis of rotation and such that the rotation
appears clockwise if one looks from the initial point of w to its terminal point. The length of w is equal
to the angular speed of the rotation, that is, the linear speed of a point of B divided by its distance
from the axis of rotation.
Let P be any point of B and d its distance from the axis. Then P has the speed d. Let r be
the position vector of P referred to a coordinate system with origin 0 on the axis of rotation. Then
d = |r| sin , where is the angle between w and r. Therefore,
wd = |w| |r| sin = |w r| . (42)
From this and the denition of vector product we see that the velocity vector v of P can be represented
in the form
v = w r (43)
This simple formula is useful for determining v at any point of B.
8
Figure 6: Velocity of a rotating body
2.2 Matrices
A matrix is a rectangular array of numbers written with the notation:
A =
_

_
a
11
a
12
a
13
a
21
a
22
a
23
a
31
a
32
a
33
a
41
a
42
a
43
_

_
A = {a
ij
, i = 1, 2...n, j = 1, 2, ...m}
2.2.1 Type of matrices:
Real matrices
Square matrix: Has the same number of rows and columns, i.e.,
A = {a
ij
, i, j = 1, 2, ...n}
The elements a
ii
are called the diagonal elements.
Diagonal matrix: A square matrix having zero elements everywhere except on the diagonal
line. The identity matrix, I is a diagonal matrix having 1 s on the unity diagonal elements.
For the mapping y = Au , if A is diagonal, then there are no interations between the inputs.
That is, u
1
will only aect y
1
, u
2
will only aect y
2
, ... so on so forth.
Triangular matrix: A square matrix is which all the elements below the diagonal are zeros
is called an upper triangular matrix U. Similarly, a lower triangular matrix, L is one for
which all elements above the diagonal are zero.
U =
_
_
b
11
b
12
b
13
0 b
22
b
23
0 0 b
33
_
_
, L =
_
_
b
11
0 0
b
21
b
22
0
b
31
b
32
b
33
_
_
Tridiagonal matrix: A tridiagonal matrix is one in which all elements that are not on the
diagonal line and the two diagonals immediately above and below it are zero. e.g.,
T =
_

_
a
11
a
12
0 0
a
21
a
22
a
23
0
0 a
32
a
33
a
34
0 0 a
43
a
44
_

_
(44)
It is encountered often in nite dierence solution method for dierential equations.
9
Sparse matrix: A matrix in which most elements are zero. These are encountered frequently
in engineering systems.
Diagonally-dominant matrix: For a given matrix A, if the absolute value of the diagonal
term is no less than the sum of all other elements on the same row, then it is said to be
diagonally-dominant matrix, mathematically,
|a
ii
|
n

j=1
j=i
|[a
ij
]| for i = 1, 2, ...n. (45)
Matrix A is said to be strictly diagonally dominant if:
|a
ii
| >
n

j=1
j=i
|[a
ij
]| for i = 1, 2, ...n. (46)
The requirement for diagonal dominance is encountered in iterative methods for solution of
set of linear equations. For the mapping y = Au , if A is diagonally dominant, then the
mapping can be approximated by a diagonal one..
Complex matrices: some or all of the elements are complex numbers. Complex matrices are
encountered when you apply Laplace transform to an ODE.
Complex conjugate transpose
A

=

A
T
(47)
e.g.
A =
_
1 + 2i 4i
6 5 2i
_
, A

=
_
1 2i 6
4i 5 + 2i
_
(48)
Hermitian: A complex matrix A is said to be Hermitian if
A

= A (49)
a
ij
= a
ji
(50)
Skew-Hermitian: A complex matrix A is said to be skew-Hermitian if
A

= A (51)
a
ij
= a
ji
(52)
Unitary:
A

= A
1
(53)
It can be seen from the denitions that, the diagonal elements of an Hermitian matrix must real
(why? a
ii
= a
ii
) and for the same reason, the diagonal elements of a Skew-Hermitian matrix must
be imaginary numbers.
10
2.2.2 Matrix Algebra
Matrix addition and subtraction
A = (a
ij
) is mn and B = (b
jt
) is mn. Then
C = {c
ij
} = A+B = {a
ij
+b
ij
} = B +A (54)
Matrix Product
If A = (a
ij
) is m n and B = (b
jt
) is n k, t = 1, 2, ..., k, then the matrix product, AB, is the
mk matrix C = (c
ij
), where
c
ij
=
n

r=1
a
ir
b
rj
Example 4
_
a b
c d
_ _
x y
z w
_
=
_
ax +bz ay +bw
cx +dz cy +dw
_
Properties of Matrix Multiplication
Suppose A = (a
ij
) is mn and B = (b
jt
) is n k, t = 1, 2, ..., k. Then
AB = BA Non-commutative
(AB) C = A(BC) Associative Law of Multiplication
A(B +C) = AB +AC Left Distributive Law of Matrix Multiplication over Addition
(B +C) A = BA+CA Right Distributive Law of Matrix Multiplication over Addition
For any scalar r, r (AB) = (rA) B = A(rB) Associative Law of Scalar and Matrix Multipli-
cation
Inverse
The inverse of a matrix A is a matrix B such that AB = I = BA. The inverse of A is denoted by
A
1
. For the mapping y = Au , an inverse mapping from y to u will be A
1
, if it exists.
Computation of matrix inverse:
Adjoint Matrix: For each entry a
ij
of A, let A
ij
be the cofactor of a
ij
. The adjoint matrix
of A is dened to be the n n matrix whose entry in the i
th
row and j
th
column is A
ji
.
A
1
=
1
det (A)
[A
jk
]
T
=
1
det (A)
_

_
A
11
A
21
A
n1
A
12
A
22
A
n2
.
.
.
.
.
.
.
.
.
.
.
.
A
1n
A
2n
A
nn
_

_
=
1
det A
adj A (55)
where A
jk
is the cofactor of a
jk
, i.e.
A
jk
= (1)
j+k
M
jk
(56)
Where M
jk
is the reduced determinant of a determinant expansion that is formed by omitting
the jth row and kth column of a matrix A.
For two-by-two matrices:
A =
_
a
11
a
12
a
21
a
22
_
, A
1
=

a
22
a
12
a
21
a
11

det (A)
=

a
22
a
12
a
21
a
11

a
11
a
22
a
12
a
21
(57)
11
Only square matrices can have inverses.
If a square matrix has an inverse, the matrix is said to be invertible or nonsingular.
Singular matrix: Square matrices that do not have inverses are called singular matrices.
Singular matrices determinants are zero:
|A| = 0 (58)

(ABCD)
1
= D
1
C
1
B
1
A
1
(59)
Example 5
_
1 2
3 2
_ _
1
4
1
4

3
8
1
8
_
=
_
1 0
0 1
_
=
_
1
4
1
4

3
8
1
8
_ _
1 2
3 2
_
Transpose
If A is an m n matrix, the transpose of A, denoted A
T
, is the n m matrix whose entry a
st
is
the same as the entry a
ts
in the matrix A.

_
A
T
_
T
= A.
(A+B)
T
= A
T
+B
T
.
(AB)
T
= B
T
A
T
.
For any scalar r, (rA)
T
= rA
T
.
If A is a diagonal matrix, then A = A
T
.
Symmetric Matrix:When every pair elements that are symmetrically placed in a matrix with
respect to the main diagonal are equal, the matrix is said to be a symmetric matrix.
a
ij
= a
ji
(60)
A
T
= A (61)
e.g.,
A =
_
2 1
1 4
_
, B =
_
_
2 1 4
1 3 1
4 1 5
_
_
(62)
Skew-symmetric Matrix: A matrix A is said to be skew-symmetric if A
T
= A.
a
ij
= a
ji
(63)
A
T
= A (64)
e.g.
A =
_
_
0 5 1
5 0 7
1 7 0
_
_
, A
T
=
_
_
0 5 1
5 0 7
1 7 0
_
_
(65)
Note: every square matrix A can be written as the sum of a symmetric matrix and a skew-
symmetric matrix (why?)
A =
Symmetric
..
1
2
_
A+A
T
_
+
Skew-symmetric
..
1
2
_
AA
T
_
(66)
12
Orthogonal matrix: Given a matrix A, if the product of the transpose of a matrix and itself
yields the identity matrix, then the matrix A is said to be orthogonal. (or orthogonal or unitary).
i.e.,
A
T
A = I (67)
e.g.
A =
_
3
5
4
5

4
5
3
5
_
(68)
For the mapping y = Au , if A is orthorgonal, then, the 2-norm (Euclidian norm) of y equals
that of u:
||y||
2
= ||Au||
2
=

(Au)
T
(Au) (69)
=

u
T
A
T
Au =

u
T
u = ||u||
2
(70)
Idempotent matrix: A matrix A is an idempotent matrix if
AA = A (71)
e.g.,
_
3 2
3 2
_
is an idempotent matrix. Clearly, the identity matrix I is a special form of the
idempotent matrix and it is the only non-singular idempotent matrix, all others are singular.
Matrix Decomposition or Factorization
Any matrix A can be decomposed into two comformable matrices B and C, i.e., represented by the a
product of matrices B and C. However, there are innite number of possible decompositions. So there
is no unique decomposition. Even so, there is a particular representation called LU decomposition,
which is useful in systems analysis:
A = LU (72)
Here L is the lower triangular matrix and U is the upper triangular matrix.
Rank of a matrix
The rank of a matrix A is the maximum number of linearly independent row vectors of a matrix A
and is denoted by
rank (A) . (73)
The rank of a matrix A is also the maximum number of linearly independent column vectors of a matrix
A. Therefore,
rank
_
A
T
_
= rank (A) (74)
Example 6 rank
_
_
_
_
0 1 1
1 0 1
1 1 0
_
_
_
_
= 3, rank
_
_
_
_
0 1 1
1 0 1
1 1 1
_
_
_
_
= 2
2.3 Eigenvalues and eigenvectors
2.3.1 Denitions
Denition 1 Let A be an n n matrix. A nonzero vector v is called an eigenvector of A if there is
a number such that
Av = v
The number is called an eigenvalue associated with the eigenvector v.
13
-6 -4 -2 0 2 4 6
-6
-4
-2
0
2
4
6
x vectors
y vectors
Figure 7: Multi-variable mapping
Denition 2 The set of eigenvalues is called the spectrum of A. The largest absolute value of the
eigenvalues is called the spectral radius of A.
Denition 3 The set of all eigenvectors corresponding to an eigenvalue of A together with 0, forms a
vector space, called the eigenspace of A corresponding to this eigenvalue, and is denoted by E ().
Denition 4 The problem of determining the eigenvalues and eigenvectors of a matrix is called an
eigenvalue problem.
2.3.2 Physical meaning of eigenvalues and eigenvectors
The matrix multiplication can be interpreted as a mapping between vectors. e.g.,
y = Ax (75)
is a function which maps a vector x to another vector y, a multi-variable function.
e.g.
y =
_
y
1
y
2
_
= Ax =
_
5 3
3 5
_ _
x
1
x
2
_
Now we pick up x vectors which are inside the unitary circle and see what are the corresponding y
vectors.
From the gure, it can be seen that the vectors within the unitary circle have been mapped to the
elliptic area. The two eigenvalues of matrix A are:(8, 2) and the corresponding eigenvectors are:

1
= 8 : v
1
=
_
1
1
_

2
= 2 : v
2
=
_
1
1
_
(76)
14
This means that, for any vector x which has the same direction of v
1
the corresponding y would be
exactly the same direction and be amplied to 8 (
1
) times larger. Similarly, for any vector x which
has the same direction of v
2
the corresponding y would be exactly the same direction and be amplied
to 2 (
2
) times larger. Therefore, the lengths in these principal directions of this ellipse are 8 and 2.
2.3.3 Characteristic Equation
Let A be an nn matrix. The function p () = det (AI) is called the characteristic polynomial
of A and the equation det (AI) = 0 is called the characteristic equation of A.
Theorem 1 is an eigenvalue of a matrix A if and only if det (AI) = 0. Each solution of the
characteristic equation has a corresponding eigenvector x:
(AI) x = 0 or Ax = x
Why? From the denition:
Ax = x (77)
(AI) x = 0 (78)
From the Crammer rule, this homogeneous linear system has a nontrivial solution if and only if the
corresponding determinant |AI| is zero. If x is an eigenvector of A corresponding to the eigenvalue
of ,then kx (k = 0) will also be an eigenvector of A :
Ax = x (79)
then:
Akx = kAx = k (x) = (kx) (80)
2.3.4 Computing Eigenvectors
Let A be an n n matrix. To nd the eigenvectors of A
1. Compute the determinant of AI. This determinant is a polynomial of degree n.
2. Find the roots of this polynomial. The n roots are the eigenvalues of A.
3. For each eigenvalue solve the equation (AI) x = 0. The solutions x to this equation are the
eigenvectors.
Example 7 A =
_
1 2
3 4
_
Then
AI =
_
1 2
3 4
_

_
0
0
_
=
_
1 2
3 4
_
(81)
|AI| =

1 2
3 4

= ( 1) ( + 4) 3 2 =
2
+ 3 10 = 0 (82)

1
,
2
= 5, 2 (83)
To solve the eigenvectors, we need to nd the nontrivial solutions to the following equations:
For = 5
_
1 + 5 2
3 4 + 5
_
x =
_
6 2
3 1
_
x = 0 (84)
x = [1, 3]
T
(85)
15
For = 2
_
1 2 2
3 4 2
_
x = 0 (86)
x = [2, 1]
T
(87)
2.3.5 Some properties of eigenvalues
Trace:
Let A be an n n matrix with eigenvalues
1
,
2
, ...,
n
. Then the sum of the eigenvalues is the
sum of the diagonal entries of A:

1
+
2
+... +
n
= a
11
+a
22
+... +a
nn
(88)
The sum of the diagonal entries of A is called the trace of A, denoted tr A.
tr (A) =
n

j=1

j
(89)
In our previous example:

1
+
2
= 3 = tr (A) (90)
Determinant:
The product of the eigenvalues equals the determinant of A:
n

j=1

j
=
1

2

n
= det A (91)
In our previous example:

2
= 10 = det (A) (92)
A singular matrix has at least one zero eigenvalue
The eigenvalues of a diagonal or triangular matrix are identical to the elements on the main
diagonal.
Given any non-singular matrix T, the matrices A and

A = TAT
1
have identical eigenvalues. In
this case

A and

A are called similar matrices.
2.3.6 Matrix Diagonalization
Let A be an n n matrix that has n linearly independent eigenvectors. Let S be the matrix whose
columns are the eigenvectors of A. Then S
1
AS is a diagonal matrix with the eigenvalues of A along
its diagonal:
S
1
AS = =
_

1
0 0
0
2
0
0 0
.
.
.
.
.
.
0 0
n
_

_
16
Proof: Assume S = [v
1
, v
2
, v
n
] , where v
1
, v
2
, v
n
are the eigenvectors corresponding to the
eigenvalues of
1
,
2
.
n
. Then
S
1
AS = [v
1
, v
2
, v
n
]
1
A[v
1
, v
2
, v
n
] (93)
= [v
1
, v
2
, v
n
]
1
[
1
v
1
,
2
v
2
,
n
v
n
] (94)
= [v
1
, v
2
, v
n
]
1
[v
1
, v
2
, v
n
]
_

1
0 0
0
2
0
0 0
.
.
.
.
.
.
0 0
n
_

_
(95)
=
_

1
0 0
0
2
0
0 0
.
.
.
.
.
.
0 0
n
_

_
(96)
Any matrix with distinct eigenvalues can be diagonalized since the n eigenvectors corresponding
to these eigenvalues will be independent.
The diagonalizing matrix S is not unique. Any column of S can be multiplied by a nonzero
constant to produce a new diagonalizing matrix.
The equation SA = S only holds if the columns of S are the eigenvectors of A. Other matrices
S will not produce a diagonal matrix .
Not all matrices are diagonalizable since not all matrices possess n linearly independent eigenvec-
tors.
If the eigenvectors x
1
, ..., x
k
correspond to dierent eigenvalues
1
, ...,
k
, then those eigenvectors
are linearly independent.
Let A be an n n matrix with eigenvectors
1
, ...,
n
. Then the eigenvalues of A
k
are
k
1
, ...,
k
n
,
the k
th
powers of the eigenvalues of A. Each eigenvector of A is also an eigenvector of A
k
. If S
diagonalizes A, it also diagonalizes A
k
:

k
=
_
S
1
AS
_ _
S
1
AS
_

_
S
1
AS
_
= S
1
A
k
S
If A and B are diagonalizable, they share the same eigenvector matrix S if and only if AB = BA.
17

Das könnte Ihnen auch gefallen