1.1 General Control Problem 1.2 Introduction To Multivariable Control System 1.3 Mathematical Tools
1.1 General Control Problem 1.2 Introduction To Multivariable Control System 1.3 Mathematical Tools
1 General Control Problem
1.2 Introduction to Multivariable Control
System
1.3 Mathematical tools
3
Controller
System Analysis
2
Controller WE FOCUS ON
Selection THESE 3 STEPS
1
Understanding
the plant
3
(1) Understanding the plant refers to studying the system plant to be controlled
and obtaining initial information about the control objectives.
Modeling the system and simplifying the model if necessary. Analyzing the
resulting model to determine its properties, deciding which variables are to
be controlled outputs and manipulated variables.
What sensors and actuators will be used. where will they be placed. Deciding
on performance specifications based on the overall control objectives
(2) The second important step in control problem is Controller selection or
design.
The control selection or design refers to selecting the control configuration,
deciding on the type of controller to be used and Designing the controller
4
(3) The third step in control system problem is analysis of the
integration of controller and plant.
Control system analysis refers to analysis of the resulting
integrated control system to see if the specifications are
satisfied. If they are not satisfied modify the specifications or
the type of controller.
Simulating the resulting controlled system either on computer
or pilot plant.
(4) The last and fourth step of general control system is Implementation
issue.
Implementation of designed control system refers to choosing hardware
and software and implementing the controller.
Finally testing and validating the control system and tuning the
controller on line if necessary.
Addis Ababa University, Institute of Technology
5
Control Engineers should have these stages of control
Engineering in mind while studying any control
techniques.
n s n n 1s n 1 ... 0
For SISO system:‐ G ( s) m
s m 1s m 1 ... 0
For MIMO System we have matrix transfer function where gij is transfer function
between input j and output i.
g11 ( s ) g12 ( s ) ... g1n ( s )
g ( s ) g 22 ( s ) ... g 2 n ( s )
G ( s ) 21
... ... ... ...
g m1 ( s ) g m 2 ( s ) ... g mn ( s )
Addis Ababa University, Institute of Technology 7
State space form:
Where;
x Ax Bu x is nx1 state variable vector,
y is output of vector mx1
y Cx Du U is an lx1 manipulating signal vector
A is nxn system matrix,
B is nxl input matrix,
C is output matrix and
All matrices having constant elements for linear system.
Addis Ababa University, Institute of Technology
9
• In nonlinear systems we can not have transfer function or state space
representation using constant matrices.
U u1 u 2 . . u m
W w1 w 2 ... w l T
T
y1 W w1 w2 . . wl
T
u1
u y
2 2
U . Y . Y y1 y2 . . yn
T
SYSTEM
.
.
um yn
Addis Ababa University, Institute of Technology
Inputs (actuating signals) Controlled outputs state variables
• Feed water through valve control • Steam flow rate
• Fuel flow rate through valve control • Steam temperature
• Air flow through guide‐vane • Steam pressure
Internal Variables
• Furnace temperature
• Furnace vacuum
• Dram level
• …
Reference Inputs
MIMO System Transfer Function G(s)
Multivariable and interaction between variables
Multivariable system has interaction. One output is function of a multi input
variables. The figure shows a three input three output system.
Y1 is function of u1, u2 and u3, while y2 is function of u1, u2 and u3.
In general an output is a function of two or more input variables in MIMO
systems
Problems arising from Control loop
interaction
‐ Closed loop system my be destabilized
‐ Controller tuning become more difficult
General lxm system
y1 g11 g13 u1
y g
g12
g22 g23 u2
Y G ( s )U ( s )
2 21
y3 g31 g32 g33 u31
y1 g11 . . . g1m u1
. . . . . . .
ݕଵ ൌ ݃ଵଵ ݑ ݏଵ ݃ଵଶ ݏ ݑଶ ݃ଵଷ ሺݏሻݑଷ . . . . . . .
ݕଶ ൌ ݃ଶଵ ݑ ݏଵ ݃ଶଶ ݏ ݑଶ ݃ଶଷ ሺݏሻݑଷ . . . . . . .
yl gl1 . . . glm um
ݕଷ ൌ ݃ଷଵ ݑ ݏଵ ݃ଷଶ ݏ ݑଶ ݃ଷଷ ሺݏሻݑଷ 15
Addis Ababa University, Institute of Technology
Introduction
Manipulation of transfer function Matrix for MIMO System
z G2u1
u1 G1u
z G2G1u
G G2G1
y G1v
v u G2 y
G1G2 y G1u
G ( I G1G2 ) 1G1 ( I G1G2 ) y G1u
For appropriately sized matrices
y ( I G1G2 ) 1G1u
( I G1G2 ) 1G1 G1 ( I G2G1 ) 1
- +
ym + n
It can be shown that the output y is expressed as; T – is complementary
y I GK GK r I GK Gd d I GK GK n
1 1 1 sensitivity function
T S T S – is sensitivity function
It is clear that, complementary sensitivity function, T, has to be unity for y to track r,
while sensitivity function, S, has to be zero for the disturbance effect on output y to be
zero. Also, it is
S T I
Addis Ababa University, Institute of Technology 17
For a unity feedback system shown above, determine the expression for;
(a) ery
(b) u
(c) Investigate the following considering L = GK;
1. Performance, good disturbance rejection requires large loop gain: L large
2. Performance, good command following: L large
3. Stabilization of unstable plant: L large
4. Mitigation of measurement noise on plant outputs: L small
5. Small magnitude of input signals: K small and L small
6. Physical controller must be strictly proper: K→0 at high frequencies
7. Nominal stability (stable plant): L small (because of RHP-zeros and time
delays
8. Robust stability (stable plant): L small (because of uncertain or neglected
dynamics)
Addis Ababa University, Institute of Technology 18
(1) Shaping of Transfer function in frequency response
• Shaping the open loop transfer function gain as function of frequency
• Shaping of closed‐loop transfer function as a function of frequency
considering sensitivity, complementary sensitivity, … in which we can
consider optimization.
(2) Signal based using time domain
• LQG which considers disturbance
• H∞
(3) Numerical Optimization
• Multi‐objective optimization like rise time, stability margins
• Model predictive control …
Addis Ababa University, Institute of Technology 19
Introduction
Transfer Function …
External input Signals: Important signals of
r (reference signal ) interest in control:
• Output signal, z
d u (disturbance at input )
• Command signal, u
d (disturbance at output ) • Error signal, e
n (measurment noise )
z Gu d
u d u K (r n z )
z Gd u GKr GKn GKz d
( I GK ) z Gd u GKr GKn d
z I GK 1GKr I GK 1GKn G I KG 1 d u I GK 1 d
Exercise: Determine the transfer functions with error and control
signal u as output.
20
Addis Ababa University, Institute of Technology
Scope of course in terms of Control targets
In general, a control target can be one or more of the following.
Common targets
• Regulation (disturbance rejection)
The controller manipulates system control variable such that the controlled
variable remains constant at the set point.
• Reference tracking
Controller manipulates the system control variable such that the controlled
output tracks desired value as function of time.
Other targets of control systems (Process Control) are:‐
• Generation of sequential procedures (for start up or shut down)
• Adaptation (change adjust some tunable parameters)
• Fault detection (avoid process damage or provide reconfiguration)
• Supervision (changing the operating condition structure or components)
• Coordination (providing the set point)
• Learning “intelligent control” (extracting some knowledge from experience)
Addis Ababa University, Institute of Technology 21
Scope of course in terms of Control targets
22
Addis Ababa University, Institute of Technology
Scope with respect to input‐output relation:‐ linear/nonlinear,
continuous/discrete
• This course is limited to Linear and continuous system. Discrete
and nonlinear systems are not treated.
• Scope with respect to domain of analysis:‐ both time domain and
frequency domain analysis are used.
Addis Ababa University, Institute of Technology 24
Some Matrix properties are very important in multivariable
control analysis and design. For example, matrix norms to
indirectly size the vector signals and transfer functions and are
extensively used.
2.2 Mathematical Background
‐ Matrix Basics
‐ Eigenvalues and Eigenvectors
‐ Singular value Decomposition
‐ Relative Gain Array
‐ Matrix Norms
‐ Factorization of Sensitivity Function
‐ Linear Fractional Transformation
Addis Ababa University, Institute of Technology 25
Introduction
1.3.1 Matrix Basics
A. Definition
• Matrix is a
‐ The elements can be constant real or complex numbers
‐ Variables of time or variables of frequency
• A matrix with m rows and n columns is said to be an mxn matrix or of
order mxn. If m = n the matrix is said to be .
General expression of matrix:
Transpose of matrix A is a matrix whose
rows and columns are replaced by columns
aij of A a ji of AT
and rows of matrix A, respectively.
Identity Matrix
Matrix vector is a matrix with one row or one column
Column vector Row vector The identity matrix has
x1
elements which are 1s
x on the diagonal and
2 zeros on off‐diagonal.
X ... Y y1 y2 ... ym 1 ym
It is usually denoted by
x
n 1 I.
x n
1 0 .. 0 0
0 1 0 0
...
I ... 0 ...
... 0
0 ... 0 0 ... ...0 1
Zero matrix has all its 0 0 ... 0 0 0 ... 0 1
elements equal to zero. 0 ... 0
27
Addis Ababa University, Institute of Technology
Introduction
B. Operation on Matrix Operation
Two matrices A and B can be added if they are of the same size, having
the same number or rows and columns. Additions and subtractions are
effected element wise.
c11 ... ... d11 ... ...
C A B ... cij aij bij ... D A B ... d a b
ij ij ij ...
... ... ... ... ... ...
Two matrices can be multiplied if the columns of the first matrix and the
rows of the next matrix are equal.
If matrix a is mxn matrix and matrix B is an nxp matrix, the product AXB is an mxp matrix.
Introduction
Direct division of matrices is not done. Rather
matrix in denominator means inverse of the
B
matrix if it has inverse. D BC 1 29
C
Inverse of a square matrix A, if it exists, is defined as
adj ( A)
A1
det( A)
The product of a matrix and its inverse is always the identity matrix of equal
size to A. 1 1
A A AA I
If A is non singular so is its transpose. A A
T 1 1 T
A matrix is said to be if it doesn’t have inverse.
Addis Ababa University, Institute of Technology
Introduction
Matrix Inverse
Inverse of a square matrix A if it exists is defined as
adj ( A)
A1
det( A)
A matrix is said to be singular if it doesn’t have inverse.
The product of a matrix and its inverse is always the identity matrix of equal
size to A. A1 A AA 1 I
If A is non singular so is its transpose. A A
T 1 1 T
Matrix Identities
Multiplication of any matrix by the identity matrix (with appropriate size)
results the original matrix.
AI IA A
IF A, B, C are of appropriate size, then;
A( BC ) ( AB )C A( B C ) AB AB A B C AC BC
Transpose and multiplication
AB T BT AT AB 1 B 1 A1 30
Addis Ababa University, Institute of Technology
Introduction
A matrix is symmetric matrix if elements Aij=Aji.
Orthogonality
If A is a matrix and x is a vector and Ax=0, then the vector x is orthogonal to each row of
the matrix.
Matrix conjugate for matrix A having complex numbers is matrix with all
elements of A conjugated.
a11 ... ... a ij j
A ... a ij ... a ij j
... ... ...
Matrix conjugate transpose is matrix resulting from first conjugating elements and then
transposing the matrix.
AH AT
Hermitan Matrix is a square matrix such that aij is the complex conjugate of aji for all
elements of aij of the matrix i.e. a matrix in which corresponding elements with respect
to the diagonal are conjugates of eachother. The diagonal elements are always real
numbers.
Unitary Matrix:‐ A matrix is said to be unitary if its hermitan is its inverse.
A H A1
Trace of square matrix A, (tr(A)) is sum of diagonal elements. tr ( A) aij i j
i 31
Addis Ababa University, Institute of Technology
Introduction
Example
Hermitan matrix
T
2 i 1 2i 2 i 1 2i 2 i 5i 0
A 5i 0 5 i A H 5i 0 5 i 1 0 5
0 5 5i 0 5 5i 2i 5 5i
Unitary matrix
1 1 i 1 i H 1 1 i 1 i
A
2 1 i 1 i 2 1 i 1 i
A
H 1 4 0 H 1
4 0 4
AA A A
32
Addis Ababa University, Institute of Technology
Introduction‐Mathematical Background
Eigenvalues and Eigenvectors
Definition: A.1
Let A be a square nxn matrix. The eigenvalues i, i=1,…,n, are the n
solutions to the nth order characteristic equation
det( A I ) 0
The right Eignevector ti corresponding to the eigen value i is the
nontrivial solution (ti≠0) to
( A i I )t i 0 At i i t i
The corresponding left Eigenvectors qi satisfy
When we just say eigenvector we mean the right eigenvector.
33
Addis Ababa University, Institute of Technology
Introduction
Lemma A.1: Matrix inversion Lemma.
If A1, A2, A3 and A4 are defined and are of appropriate size.
A1 A2 A3 A4 1 A11 A11 A2 ( A4 A11 A2 A31) 1 A4 A11
Lemma A.2: Inverse of partitioned matrix.
If A11 and X 1 exist A
A 11
A12
A21 A22
1 1 1
A11 A12 A11 A11 A12 X 1 A21 A11
1 1
A11 A12 X 1
A
21 A22
1
X A21 A11 1
X 1
Where; X A A A1 A 1
22 21 11 12 If A22 and Y 1 exis then
1
A11 A12 Y 1 Y 1 A12 A22
1
A 1
21 A22 1
A22 A21Y
1 1 1 1
A22 A22 A21Y A12 A22
1
Y A11 A12 A22 A21
34
Addis Ababa University, Institute of Technology
Introduction‐Mathematical Background
Schur’s Formula for partitioned matrix:
A11 A12 1
det det( A11 ). det( A22 A21 11 A12 )
A
A21 A22
1
det( A22 ). det( A11 A12 A22 A21 )
Addis Ababa University, Institute of
35
Technolgoy
Introduction‐Mathematical Background
Eigenvalues and Eigenvectors
Example: 1 1 1 0 1 1 0
A 2 3 2 det(I A) det 2 3 2
1 1 2 1 1 2
That means the eigenvalues of the matrix are;
1, 2, 3
Eigenvectors corresponding to each eigenvalue are solutions of Ax=x;
1 1 0 x1 x1 x1 x2 x1 x2 0 x1 1
2 3 2 x 1 x 2 x 3 x 2 x x 2 x 2 x 0 x x
1 2 2 1 2 3 2 1 3 1 3 x2 0
1 1 2 x3 x3 x1 x2 2 x3 x3 x1 x3 0 x3 1
1 1 0 x1 x1 x1 x2 2 x1 x1 x2 x1 2
2 2 3 2 x 2 x 2 x1 3 x2 2 x3 2 x2 x2 2
2 2 x3 1
1 1 2 x3 x3 x1 x2 2 x3 2 x3
x1 x2 3 x1 x1 1
1 1 0 x1 x1 x2 2
2 3 2 x 3 x 2 x1 3 x2 2 x3 3 x2
3 2 2 x3 3
x1 x2 2 x3 3 x3
1 1 2 x3 x3
36
Addis Ababa University, Institute of Technology
Introduction‐Mathematical Background
Eigen Values and Eigenvectors
Eigenvalues of A and A transpose are the same.
Eigenvalues are invariant under similarity transformation.
DAD 1
Inverse of A exists if all eigenvalues of A are non zero.
Eigenvalues of Hermitan matrix are real.
Positive Definite:‐ Hermitan matrix is positive definite if all its
eigenvalues are positive.
Eigenvalues of loop transfer function i(L(j)) are known as
characteristic loci.
Addis Ababa University, Institute of Technology 37
Introduction‐Mathematical Background
Eigen values and Eigenvectors
Some Properties of Eigenvalues of state matrix:
For a linear system given by x Ax Bu
State x can be transformed to another set of states q by the
Eigen vector T such that
x Tq q T 1
x q T 1 ATq T 1Bu
The system is stable if the real part of the Eigen values are
negative.
Addis Ababa University, Institute of Technology
Introduction‐Mathematical Background
Singular Value Decomposition (SVD)
The singular values decomposition of an lxm matrix is
U is unitary and orthognal matrix UU T I
A U V V is
H
unitary and orthognal VV T I
Sigma is diagonal given by
1
l ≥ m
0
1 Diagonal matrix of real non negative singular values.
1 diag1 2 ... , k ; k min(l , m )
This factorization is called Singular Value Decomposition. MATLAB can be used to ge
1 2 3 0.3863 0.9224
0.4287 0.8060 0.4082
A U , S ,V svd ( A) U
9.5080
S
0 0
V 0.5663 0.1124 0.8165
4 5 6 0.9204 0.3863
0 0.7729 0
0.7039 0.5812 0.4082
1 2
0.2298 0.8835 0.4082 9.5255 0
A 3 4 U , S ,V svd ( A) U 0.5247 0.2408 0.8165 S 0 0.5143 0.6196 0.7849
0.8196 0.4019 0.4082 V
5 6 0 0 0.7849 0.6196
Addis Ababa University, Institute of Technology 39
Introduction‐Mathematical Background
Singular Value Decomposition
Singular values of matrix A is given by;
i ( A) i ( A H A) i ( AA H )
Where i(x) denotes eigenvalues of x.
MATLAB can be used to determine the singular values of A using
command svds ( A)
( A H ) ( A);
( A) ( B ) ( AB ) or ( A) ( B ) ( AB )
( AT ) ( A);
( A) ( B ) ( AB )
( AB ) ( A) ( B )
Addis Ababa University, Institute of Technology 40
Introduction‐Mathematical Background
Singular Value
A 0
max ( A), ( B )
0 B
Pseudo‐inverse (Generalized inverse)
Consider the set of linear equations
y Ax ylx1 Alxm , xmx1
y Ax 2 is min imized x 2 x12 x 22 ... x m
2
The solution is given in terms of the pseudo‐inverse (Moore‐Penrose
generalized inverse) of A:
x A y
1 H 1 r
A Vr r U r v i uiH
i 1 i ( A)
Addis Ababa University, Institute of Technology 41
Introduction‐Mathematical Background
Singular Value Decomposition
Pseudo‐inverse exist for any matrix.
AA A A A AA A
MATLAB command for pseudo‐inverse of A is pinv(A).
Addis Ababa University, Institute of
42
Technolgoy
Introduction‐Mathematical Background
Rank of a matrix: is equal to the number of non‐zero singular
values of the matrix. Let rank(A) = r, then the matrix A is called
rank deficient if r < k=min(l,m), and we have singular values i=0
for i=r+1,…k, A rank deficient square matrix is a singular matrix
(non‐square matrices are always singular)
Singular Value of a 2x2 matrix
Let,
2
b tr ( A A) a ij , c det( A H A)
H
i, j
Now the sum of the eigenvalues of a matrix is equal to its trace
and the product is equal to its determinant, so
1 2 b, 12 c
b b 2 4c b b 2 4c
( A) ; ( A)
2 2
1 2 ( A) (5.465
Example: A
3 4 ( A) 0.366
Addis Ababa University, Institute of Technology
Introduction‐Mathematical Background
Singular Value Decomposition
Condition number is the ratio of the largest singular value to the smallest singular
value. If the ratio is high the system said to be ill‐conditioned.
1 0 10.1 0
Example:‐ 1 0
G
RGA(G )
0 1
G
0 0.1
CN 101
10 1
G is poorly conditioned when CN is large number, usually > 10, small change in the
model for this process can make it very difficult to control.
Input output selection can be effected in such a way that the CN is less than 10.
Example:‐ 0.5714 0.3766 0.7292 1.618 0 0 0.0541 0.9984 0.0151
W V T
0.6035 0.4093 0.6843 0 1.143 0 0.9085 0.054 0.0068
0.5561 0.8311 0.0066 0 0 0.0097 0.006 0.0154 0.999
Addis Ababa University, Institute of Technology 44
Introduction‐Mathematical Background
Relative Gain Array
• For a square matrix A of size mxm, the RGA is defined as
RGA( A) A( A1 )T
The product is element wise multiplication
• For a matrix A given by
1 2 1 2 1 2 3
A ; A 1.5 0.5 ; RGA( A) 3 2
3 4
• RGA is a measure of sensitivity to element by element
uncertainty in the matrix
• Properties RGA
‐ The sum of elements in each row and in each column is 1.
‐ RGA of diagonal matrix is 1
‐ RGA is scaling invariant
• RGA can be generated to any matrix using the pseudo
inverse. However most properties above may not be defined.
Addis Ababa University, Institute of
45
Technolgoy
Introduction‐Mathematical Background
Singular Value Decomposition
2.4376 3.0241 0.4135
RGA is 1.2211 0.7617 0.5407 Preliminary pairing:‐ y1‐u2, y2‐u3, y3‐u1
2.2165 1.2623 0.0458
Norm is a single number used to express indirectly the size or
magnitude of a vector or matrix or system.
The most common norm is the Euclidean norm, which is
distance between two points.
Addis Ababa University, Institute of Technology 47
Introduction‐Mathematical Background
Matrix Norms
Vectors in n dimensional space
• A vector having n elements is said to be n‐dimensional
vector a1
.
a .
• Vector norms .
n
Three types of vector norms a
i) 1‐norm sum of absolute values of elements
a 1 ai
i
ii) 2‐norm is square root of sum of square of elements
2 2
a
2
ai
i
iii) ∞‐norm is the maximum of the absolute value of elements
a max a max a i Addis Ababa University, Institute of Technolgoy 48
i
Introduction‐Mathematical Background
Matrix Norms
Similar to vector norms, the followings are matrix norms
i) sum matrix norm A
sum ij
a
i, j
ii) Frobenius matrix (Euclidean) norm 2
A F
a ij tr ( A H
A)
i, j
iii) Max element norm
A max max aij
i, j
iv) Induced norm for a system with input w, output z and
transfer function G, induced norm is defined as
Gw p
G ip max
w 0 w p
p 1 / p
Where w p i w i denotes the vector p norm
;
Addis Ababa University, Institute of
49
Technolgoy
Introduction‐Mathematical Background
Matrix Norms
Signal and system norm;
1‐norm in time (integral absolute)
e (t ) 1 ei ( ) d
i
2‐norm in time
2
e (t ) 2
2
ei ( ) d e (t ) 2 e i ( ) d
i i
∞‐norm in time
e (t ) max max ei ( )
i
Power‐norm or RMS norm
1 T 2
e (t ) power lim i
e ( ) d
T 2T T i
50
Addis Ababa University, Institute of Technology
Introduction‐Mathematical Background
Matrix Norms
• System performance can be evaluated or controller can be designed by taking
various input signal types and output norms. The most important ones are H2
and H∞ norms. If (g(t) is impulse response of G(s))
‐H2 is defined as G ( s ) 2 g (t ) 2
H∞‐norm is defined as e (t ) 2
G ( s ) max G ( j ) max
d (t ) 2
L1‐norm e (t )
L1 norm : g (t ) 1 g (t )dt max d ( t )
d (t )
Based on the norms used for input
and output, the system d 2 d d
G (s ) usually
performance can be evaluated by e2
g (t ) 1 usually
H2, H‐infinity or e G (s ) 2
L1 norms e pow 0 G (s ) G ( s)
51
Addis Ababa University, Institute of Technology
Introduction‐Mathematical Background
Matrix Norms
Modal Matrix and Similarity transformation
Consider Matrix A, a vector x and constant ;
Ax x
This equation says that multiplying x by A is equal to multiplying x by a constant.
Vector x of A is known as Eigenvector whileis known as Eigen value.
Matrix M formed from Eigenvectors is called Modal matrix.
Eigenvalues of matrix A are obtained from the equation;
det( A I ) 0
52
Addis Ababa University, Institute of Technology
Introduction‐Mathematical Background
Linear Transformation
• A function L which maps vector space V to another vector
space W is said to be linear if
L(ru sv ) rL(u) sL(v ) Where r and s are scalars
• The above definition of linear function (linear
transformation) respects additively and homogeneity
This property helps to express any vector as a sum of two other
vectors.
Addis Ababa University, Institute of
53
Technolgoy
Introduction‐Mathematical Background
For a vector signal with m rows we can define the mxm matrix
(covariance matrix). Example Rz for z(t) as follows.
Rz z (t ) zT (t )dt
If Rz is known for a signal z, we can calculate the sizes of a number
of related signals. If we, for example, form a scalar signal using a
linear combination of the components of z, s(t)=Lz(t).
2 T T T T T
s2 Lz (t ) z (t ) L dt L z (t ) z (t )dtL LRz L
The size of the kth component of Z is given by the kth diagonal
element Rz.
2
s 2 coresponds to sum of diagonal elements in matrix Rz
2
s 2 tr ( Rz )
Addis Ababa University, Institute of
54
Technolgoy
Introduction‐Mathematical Background
Matrix Norm
Matrix B is similar to matrix A if
B P 1 AP
Consider a matrix given below. The eigen value is and it has one Eigenvector.
The matrix is called Jordan block.
1
J 2
0
Generally an n by n matrix of the form has n Jordan cells with Eigenvalue and
single Eigen vector.
1 0 . . . 0
0 1 . . . 0
. 0 . . . .
0 . . . . 1
0 . . . 0 0
Addis Ababa University, Institute of
55
Technolgoy
Introduction‐Mathematical Background
Matrix Norm
Norm of Inxn = 1,
Norm of Unitary matrix =1
If U is unitary, norm
AU UA A
Addis Ababa University, Institute of
56
Technolgoy
Introduction‐Mathematical Background
Matrix Linear fractional Transformations
Like scalar differential equation matrix differential equations may
arise when modeling MIMO systems
A second order vector differential equation is given by;
A0 x (t ) A1 x (t ) A2 x(t ) 0
Taking Laplace transform
A0 x ( s ) sA1 x ( s ) s 2 A2 x ( s ) 0
In SISO systems the transfer function is ratio of two polynomials in s
N(s) and D(s) .
N ( s)
G ( s)
D( s)
Similarly, in MIMO systems the transfer function is ratio of two
vector polynomials N(s) and D(s) .
G ( s ) N ( s )D ( s ) 1
This is called matrix fraction description.
Addis Ababa University, Institute of
Technolgoy
57
Introduction‐Mathematical Background
Linear Diophantine Equation
P11 P12
P
P12 P22
The lower fractional transformation Fl is the transfer function R1 resulting
from wrapping positive feedback K around the lower part of P.
z p11w p12u
v p21w p22u
Eliminating v and u
z R1w Fl ( P , K ) w Fl ( P , K ) P11 p12 K ( I p22 K ) 1 p21
[ p11 p12 K ( I p22 K ) 1 p21 ]w
In words, R1 is written as a lower LFT of P in terms of the parameter K.
Similarly,
58
Addis Ababa University, Institute of Technology
Introduction‐Mathematical Background
Linear Diophantine Equation
Fu ( P , ) P22 p21( I p11) 1 p12
An interconnection of LFT yields an LFT
An important property of LFTs is that any interconnection of LFTs is again an LFT. R is
written in terms of lower a lower LFT of K’, which again is a lower LFT of K, and we want
to express R directly as an LFT of K. we have
R Fl (Q, K ' ) K ' Fl ( M , K ) R Fl ( P , K )
Addis Ababa University, Institute of
59
Technolgoy
Introduction‐Mathematical Background
Linear Diophantine Equation
Fu ( P , ) P22 p21( I p11) 1 p12
The lower fractional transformation Fl is the transfer function RI resulting
from wrapping positive feedback K around the lower part of P.
Eliminating v and u
z R2 w Fu ( P , ) w
[ p11 p12 K ( I p22 K ) 1 p21 ]w
Interconnection of LFTs
p p12
P 11
p21 p22
Q11 Q12 M11 ( I Q22 M11 ) 1Q21 Q12 ( I M11Q22 ) 1 M12
M 21 ( I Q22 M11 ) 1Q21 M 22 M 21Q22 ( I M11Q22 ) 1 M12
60
Addis Ababa University, Institute of Technolgoy
Introduction‐Mathematical Background
Reference
‐ Sigurd Skogestad, Multivariable Feedback Control
analysis and design, Wiley 1996
P. Albertos A. Sala, Multivariable Control
Systems: Engineering Approach, Springer
Torkel Glad and Lennart Ljung, “Control Theory
multivariable and nonlinear methods, Taylor and
Francis, 2000
Addis Ababa University, Institute of
61
Technolgoy