Control Systems

Yusup Isaev
PREFACE

Control system engineering is an exciting and challenging field and is a

multidisciplinary subject. This book is designed and organized around the concepts
of control system engineering as they have been developed in the frequency and
time domain for an introductory undergraduate course in the control system for
engineering student of all disciplines.
The author has tried to set out the material in simple and readily available
form. A scope of knowledge of higher mathematics necessary for understanding
the content corresponds to the syllabus for earlier stages of tuition at higher
technical education institutions. This book includes the coverage of classical
methods of control systems engineering: Laplace transforms and transfer functions,
block diagrams and spectral signal representation, root locus analysis and design,
Routh-Hurwitz stability analysis, frequency response method of analysis including
Bode, Nyquist, steady-state error analysis, second order systems, phase and gain
margin and bandwidth, phase plane method and dynamic systems phase portraits,
harmonic linearization method.
Book content provides the foundations which help solve problems quickly
and effectively find practice tasks. The author focused on engineer calculation
methods using applied program tools as Mathcad, Electronics Workbench,
MATLAB - Simulink. Most programs used in the book are designed by the author,
though standard application programs as MATLAB - Simulink are also used in
some cases.
The book contains plenty of explanatory examples. Tasks given in appendix
help fasten the understanding of the material. This book is addressed to the wide
range of readers: students and specialists resuming studies in the field of control
theory because of expanding range of problems concerned with automatic
processes, students and specialists who want to refresh their knowledge by
studying a part of the book that has not been include into engineering specialities
curriculum
I would appreciate being informed of errors, or receiving comments about the
book. Please send to the authors’ e-mail to [email protected]
CONTENT
CHAPTER 1. Linear systems…………………………………………. 7
1.1. Linear systems description and analysis with the help of 9
differential equations…………............…………………………….........
1.1.1. General notions on automatic control…………………………….. 9
1.1.2. Differential equations. Linearization ……………………………. 9
1.1.3.Component definition of automatic system control and 12
mathematic description of linear component properties....………………
1.2. Systems description and analysis with the help of impulse response 14
functions………………………………………………………………... 16
1.2.1. Block transfer function...……………………………………….... 16
1.2.2. Impulse response characteristic……………………………...…… 17
1.2.3. Impulse response function(weight function – Green’s function)…
1.3. Spectral signal representation………….…….……………………. 18
1.3.1 Bode plot……………………………………………..……………. 22
1.3.2. Correlation between component input and output variables in 26
steady-state……………...……………………..…………………………
1.3.3. Steady and minimal phase components frequency characteristics 27
peculiarities…………………...…...……………………………………..
1.4. Some typical blocks of automatic control systems ………………. 28
1.4.1. Amplifying block.………………….……………………..............
1.4.2. Aperiodic inertial component of the first period..……………….. 30
1.4.3. Oscillatory,aperiodic and conservative components…………...... 30
1.4.4. Astatic differentiated component ……….…………….................. 30
1.4.5. Differential component………….……….…………….................
1.5. Block diagrams………..………….……….……………............... 33
1.5.1. Получение передаточной функции системы автоматического 38
регулирования по передаточным функциям звеньев…….…….......... 39
1.5.2. Blocks in series……………………………………………............ 43

1
1.5.3. Blocks in parallel…………….……………………………............ 43
1.5.4. Feedback control system ……………………….…………………
1.5.5. Block diagram reduction……….………………………………… 44
1.5.6. Summing point moving……….………..………………………… 44
1.5.7. Branching point moving……..……………………………………
1.5.8. Example of diagrams reduction rules……………………………. 45
1.6. Control system generalized equivalent circuit……………………. 46
1.6.1. Automatic control system property in steady state……………….. 46
1.6.2. Proportional controller (P controller)..…………………………… 47
1.6.3. Proportional derivative controller(PD controller).………………. 48
1.6.4. Integrated controller (I controller)………….……..………………
1.6.5. Proportional integrated controller (PI controller)………………… 50
1.6.6. Proportional integral-derivative controller (PID controller)………
1.7. Transient processes calculation………………….....…………….. 54
1.7.1. Transient process calculation in state space ……………………... 54
1.7.2. Convergence of matrix differential equation to scalar equations… 57
1.7.3. Transient processes calculation with Impulse response function 57
use ……………………………………………………………………….. 58
1.7.4. Analytical method of transient process calculation……………….
1.8. Automatic control linear system stability……………..…………... 59
1.8.1. The Routh-Hurwitz stability criterion(algebraic stability criterion) 61
1.8.2. The Routh-Hurwitz algebraic criterion application………………. 61
1.8.3. Mikhailov stability criterion………………………………………
1.8.4. Nyquist stability criterion ………………………………………... 68
1.8.5. Logarithmic stability criterion………………...…………………..
1.8.6. Application of Nyquist criterion and transient processes quality 74
criteria…………………………………………………………………… 80
1.8.7. Analysis of controller type and parameters on stability in 83
frequency domain ………………………..…….……………………......

2
1.9. Stability areas determination………...………………………….... 85
1.9.1. D decomposition method………………….………………………
1.9.2. Root locus analysis………….……….…………………………… 88
1.9.3. Poles placement method…….……….…………………………… 90
CHAPTER 2. Non-linear dynamic systems………………………….. 92
2.1. Phase space and dynamic systems phase portraits ………………… 102
2.1.1. Phase portraits construction with surface use……………………..
2.2. Phase plane method……......………………………………………. 105
2.2.1.Servomotor equation with ideal relay characteristic……………….
2.2.2. Servomotor equation with real relay characteristic……………….. 107
2.3. Lyapunov straight method………………………………………… 112
2.4. Harmonic linearization method………...………………………….. 112
2.5 Self-exciting oscillations analysis algorithm……………………….. 115
Literature……………………………………………………………….. 118
Appendix 1…………………………………………………………......... 121
Appendix 2………………………………………………………….........
Appendix 3…………………………………………………………......... 121
Appendix 4……………………………………………………………….
Appendix 5………………………………………………………………. 131
Appendix 6………………………………………………………………. 140

140

3
CHAPTER 1. LINEAR SYSTEMS
1.1. Linear systems description and analysis with the help of differential
equations
1.1.1. General notions on automatic control
Automatic control theory and automatic control system are parts of engineering
cybernetics.
Engineering cybernetics studies general principles of technological and
manufacturing processes dynamic system control.
Automatic regulation is defined as the process of keeping up the required value
of some physical variable in technical devices without human involvement with the
help of regulators.
Automatic control is changing process according to some law some physical
variable in technical devices without human.
Technological process of energy production, transmission and distribution is
referred as the ordered interaction of power engineering objects: turbo and
hydrogenerators, transmission lines, transformers and other machinery. Directional
influence on these objects is called control process. Cooperating electric power
objects which are exposed to intended influences are called control objects.
Control systems are based on feedback principle which is so that the state of
controlled object is characterized by the same parameters х1 , х2 , ..., хn
(Figure 1) which is compared with desired values and if current state of
parameters is different from the desired one then the system establishes
actuating signals х р1 , х р 2 , ..., х р m , which return the state of system to the desired

one.
Feedback mechanisms can help today’s increasingly complex computer
systems adapt to changes in workloads or operating conditions. Control theory
offers a principled way for designing feedback loops to deal with unpredictable
changes, uncertainties, and disturbances in systems.

1
 For power plants disturbances occur as unexpected change of fuel quantity,
station capacity give to electric system by other power plants and capacity assumed
by consumers etc

xd1 xdn

 
x1
Object
xn

actuator

Figure 1

Because of the fact that object characteristics are given parameters than the
necessary requirements to control system should be provided by the corresponding
controller characteristics selection.

Contol by disturbance means that if there is a certain correlation between the

disturbance actions and regulated variable than the object can be influenced to
prevent the expected controlled variable deviation.

1.1.2 Differential equations. Linearization

At CS investigations two problems are solved

1. Analysis problem – CS structure and parameters are given, its properties

should be determined.

2. Design problem – CS should be constructed according to given properties.

2
 f(t) Both problems are not simple and they can be
solved with the help of mathematics. CS is
y(t) x(t)
divided into subsystems “bricks”. Each
component is described by non-linear equation,
Figure 2
differential in most cases.

Let’s analyze one of these blocks which is described by non-linear equation:

   d 
 d2
F ( y, y, x, x, x)  f , where x  x, x  2 x. - dynamics equation
dt dt
y (t )  input value , x (t )  output value , f (t )  disturbance .

Equation always has equilibrium state with coordinates :

  
x  x0 , y  y0 , f  f 0 , x  x  y  0 .

 F ( y ,0, x,0,0)  f ,  statics equation.

Investigation of this equation is not an easy task. So it is reduced – i.e. linearized to

neighbourhood of equilibrium point, decomposing in Taylor series.

 F    F    F   F  
 F     y  ...  f
F ( y0 ,0, x0 ,0,0)    x      x      x    y  
 x 0   x 0   x 0  y 0 y
 0
 
x  x(t )  x0 , y  y (t )  y0 ,  x  x,....
Ignoring the components of higher order of vanishing subtract from the obtained
equation statics equation and obtain:

 F    F    F   F  
 F 
  x      x      x    y      y  f  f 0  f
 x 0   x 0   x 0  y 0  
  y 0

 F 
Divide linearized equation by invariable   and obtain:
 x 0

 F   F   F 
      F   
      y 
  x 0    x 0   y 0 0  y  f
x   x  x y  
 F   F   F   F   F 
         
 x 0  x 0  x 0  x 0  x 0
3
   F   F 
     
 2   x 0 


 x 0
T2  F , T1  F
    
   
  x 0  x 0

  F   F 
  y   
 
 0 , k     x 0 ,
T22  x  T1 x  x  k1y  k2  y  k3f где k1   
  

 F   F 
2
    
  x 0  x 0

k3  1
  F 
  
 x 0




d
If derivation procedure is set down by p  then the equation can be rewritten in
dt
the form of:

T 2
2
p 2  T1 p  1 x   k1  k 2 p  y  k3f

from which the required variable can be found

x 
 k1  k2 p  y 
k3
f  W ( p)y  W f ( p)f
T
2
2
p  T1 p  1
2
T p  T1 p  1
2
2 2

 k1  k2 p  k3
Formulae W ( p)  , W f ( p)  are called transfer
T
2
2
p 2  T1 p  1 2
T 2 2
p  T1 p  1
functions of deviation of defined variable and of disturbance

The last reduction is to divide all the components into basic variables
x  x / x 0 , y  y / y 0 , f  f / f 0

T 2
2
 
p 2  T1 p  1 x  k1  k2 p y  k3 f ,

 y0  y0  f0
where k1  0 k1 , k 2  0 k 2 , k3  0 k3
x x x

4
1.1.3 Control system component definition and mathematic description of
linear component properties
To analyze control system component properties their block diagram is used, in
which separate elements are depicted as blocks connected to each other according
to substitute elements connection.
Separate component is characterized by a
definite mathematical connection between input
Figure 3 ( xin ) and output ( xout ) variables (Figure 3).
System components obtain the property of action direction, i.e. ability to one way
signal passing from input to output.
Connection between output and input variables of linear component (or system) is
depicted by linear differential equation with constant coefficients.
d nx d n1 x d 2x dx
a0   a   ...  a   an1   an  x 
dt n 1
dt n 1 n2
dt 2
dt (1)
m m 1 2
d y d y d y dy
 b0  m  b1  m1  ...  bm2  2  bm1   bm  y
dt dt dt dt
Linear differential equations have solution, which consist of two components
x  xc  x (2)
First component ( xc ) is a complementary solution of homogeneous differential
equation, which is obtained from (1) by making its right part equal zero.
In general homogeneous differential equation has the following solution

x c  C1  e p1t  C2  e p2t  ...  Cn  e pnt (3)

p1, p2, …, pn are characteristic equation roots which are given by

a0  p n  a1  p n1  ...  an2  p 2  an1  p  an  0 . (4)

Coefficients C1, C2, …, Cn are determined by initial conditions.
Forced component ( x  ) is a particular solution of inhomogeneous equation (1)
when the right part of equation is not zero.
If in steady state y is not a function of time then forced component can be
calculated by making all the derivatives equal zero, the following result is obtained

5
 an  x  bm  y ;
bm
x   y .
an

1.2 Systems description and analysis with the help of impulse response
functions

1.2.1 Block transfer function

The transfer function is based on application of operator method, namely the
Laplace transform. The formula of the Laplace transform for a function of time x(t)
is given by

X ( p )   e  pt  x(t )  dt  L  x(t ) , x(t )  L1  X ( p )  . (6)
0
where p is a complex variable, and L is Laplace operator
The transfer function is defined as the ratio of the Laplace transform of the
output хвых ( p ) to the Laplace transform of the input хвх ( p ) under the assumption
that all initial conditions are zero
X ( p)
W ( p)  Вых . (7)
X Вх ( p)
The transfer function can be obtained from differential equation if the Laplace
transform is applied to it
a0  p n  X ( p)  a1  p n1  X ( p)  ...  an1  p  X ( p)  an  X ( p) 
(8)
 b0  p m  Y ( p)  b1  p m1  Y ( p)  ...  bm1  p  Y ( p)  bm  Y ( p).
From (8) it is easy to define that
b  p m  b  p m1  ...  bm1  p  bm A( p )
W ( p )  0 n 1 n1  . (9)
a0  p  a1  p  ...  an1  p  an B( p )
It should be noted that denominator of transfer function coincides with the left part
of characteristic equation mentioned above.
Reverse to the Laplace transfer function is Green’s function (weight function –
impulse response function)
w(t )  L1 W ( p )   W ( p )  L  w(t ) 

1.2.2 Impulse response characteristic (impulse response function)

Let’s consider two important functions:
the Dirac delta function  ( x  x0 ) and  ( x  x0 ) - The Heaviside step function

6
 The Dirac delta function property:
 at t  t0 
 (t  t0 )  
0 at t  t0
,   (t  t ) f (t )dt  f (t )

0 0

The Heaviside step function property:

0 at t  t0
 (t  t0 )     (t )   (t )
1 at t  t0


  ( x  x )dx   ( x  x )

0 0

The Laplace transform of these functions

1
L  (t )  1, L  (t ) 
p

Impulse response function h(t) describes the

reaction of the output in response to input
change in the form of unit step function. This
is system’s impulse (block reaction) to some
external change. The unit step function is the
Heaviside function  (t )

Figure 4
Differential equation solution with unit step function  (t ) in the right part is
impulse response function h(t ) :

dx(t )
 x(t )   (t )  equation solution is x(t )  h(t )
dt
,

then by external change f (t ) we obtain:

7
 t
dx (t )
 x(t )  f (t )  equation solution is x (t )  h(t ) f (0)   h( ) f (t   ) d
dt 0

At zero initial conditions:

t
x(t )   h( ) f (t   ) d
0

Differential equation solution with impulse right part  (t ) is Green’s function or

weight function:
dx(t )
 x(t )   (t )  equation solution is x(t )  w(t ) .
dt

Impulse response function w(t) describes the reaction of the output in

response to input change in the form of delta function.

By arbitrary external change f (t ) we obtain:

t
dx (t )
 x(t )  f (t )  equation solution is x (t )  h(t ) f (0)   w( ) f (t   )d
dt 0

At zero initial conditions:

t
x(t )   w( ) f (t   ) d
0

1.2.3. Impulse response function (weight function-Green’s function)

Relation between transfer and impulse response functions can be seen using
the following correlations:

 (t )   (t )  w(t )  h(t )

t t
 (t )    (t )dt  h(t )   w(t )dt
0 0

Remember that the Laplace transform of integral and derivative has the form of:

t  F ( p)
L  f (t )  F ( p) L   f (t )dt   , L  f (t )  pF ( p)
0  p

Hence the impulse response function has the form of

8
  dh(t )  1
H ( p)  L  h(t ) , W ( p)  L   , H ( p )   W ( p) (11)
 dt  p

Relation between dynamic characteristics can be seen in diagram shown in

Figure 5

Figure 5
Example 1: Differential equation is given

d 2x dx dy (t ) dx(0) dy (0)
5 2  4  3x  y (t )  2 , x(0)  y (0)  0,   0.
dt dt dt dt dt
Determine the transfer function of differential equation

1. Write the equation in operator form :

5 p 2 X  4 pX  3 X  Y  2 pY   5 p 2  4 p  3  X  1  2 p  Y

2. Calculate the transfer function: W ( p ) 

1  2 p 
5 p 2
 4 p  3

Example 2: Differential equation is given

d 2x dx dx(0)
0,1  10  100 x  f (t ), x (0)  0,  0.
dt 2 dt dt

Calculate the equation using MathCAD.

9
Determine the transfer function of differential equation

And w(t)-Green’s function (weight function)

2 2
Z( p )  0.1 p  10 p  10

1
W ( p ) 
Z( p )

1 invlaplace  p (  50.)  t
w1( t )   .2582 e  sinh ( 38.73 t)
2 2 float  4
0.1 p  10 p  10

 88.729833462074168852 1
p  Z( p ) solve  p     
 11.270166537925831148 p
1

d 1 p 1 t 1 p 0 t (  11.27)  t (  88.73)  t
z( p )  Z( p ) w( t )  e  e w( t ) float  4  .1291 e  .1291 e
dp  1
zp  0
zp

t  0  0.01 4 

0.083

0.062
w1( t )
0.042
w( t )
0.021

0 0.088 0.18 0.26 0.35

t

1 invlaplace  x (  50.)  t (  50.)  t

h1( t )   .1000e-1  .1000e-1e
  cosh ( 38.73 t )  .1291e-1e
  sinh ( 38.73 t )
0.1x2  10x  102x float  4

d 1 1 p 1 t 1 p 0 t
z( p )  Z( p ) h ( t)   e  e
dp Z( 0) p z p
1 1   p z p
0 0  
(  11.)  t1 (  89.)  t1
h ( t1) float  2  .10e-1  .11e-1 e  .15e-2 e

t  0  0.01 4 

Determine h(t) impulse response function of differential equation

10
 0.0098

0.0073
h1( t )
0.0049
h( t )
0.0024

0 0.088 0.18 0.26 0.35

t

Check w(t) = h'(t)

0.083

d 0.062
h1( t )
dt 0.042
w( t )
0.021

0 0.088 0.18 0.26 0.35

t

Calculate the transient process by external influence f(t)=20sin( t)

  20 f( t)  20 sin   t

t

Fo( t)   w t    f  d

0

 x
1 
d
2
d 2  
0.1 x( t)  10 x( t)  10  x f( t) D( t  x)   2  N  100
2 d t 10 10 f( t)
dt  x 
1
x 
0

 0.1 0.1 0.1 

 0   0
x  rkfixed   0 5   N  D i  0 N t  x
 
0 

0.12

0.065
Fo t i

 x1  i 0.01
0 0.11 0.22 0.33 0.44
0.045

0.1

ti

11
1.3 Spectral signal representation

Any time function can be depicted as sum of harmonic functions.

a0 N N
X (t )    ak cos  kt   bk sin  kt    Ak cos  kt  k  ,
2 k 1 k 0

  2 / T , T is function period. Expansion coefficients are determined by

formulae:

 k 2 t   k 2 t 
T /2 T /2
2 2
ak   X (t )cos   dt , bk   X (t )sin   dt ,
T T /2  T  T T /2  T 

0 при a  0
b  
Ak  ak2  bk2 , k  arctg  k    при a  0
 ak   / 2 при a  0


If it is written using Eiler’s formula

e jt  cos  t   j sin(t ), e  jt  cos  t   j sin(t ) ,

then expression for series is the following :

a0 N N
1
X (t )    ak cos  kt   bk sin  kt    Ak exp  jkt , Ak   ak  jbk 
2 k 1 k  N 2

Function expansion into a Fourier series with MathCAD help

Example 1: Expansion of a triangular pulse into a Fourier series

12
 2  1
T  0.05 t  0 0.01 T  2 T   f 
T T

t 2 T
f( t)  if 0  t 
T 2
1.2
 1  t   2 if T  t  T
  1
 T 2
0.8
0 otherwise f( t ) 0.6
0.4
0.2
0
0 0.017 0.033 0.05
t

 1 
 0.405 
T  
2   0 
M  5 k  0 M a    f( t)  cos    k t dt a 
Част ь рису нк а с идент ифик ат ором от ношения rId192 не найдена в файле.

A  i a
k T   0.045  k k
0
 0 
 
 0.016 

Two forms of expansion

a M A M
 akcos  kt 
0 0
F( t)   F1( t)    A  sin    k t  arg A   
2 2  k   k  
k 1 k 1

Expansion check

arg  A 
A 
 k
k deg
1.2
1 90
F( t ) 0.405 -90
0 -90
f( t ) 0.6 0.045 -90
F1( t ) 0 -90
0.016 -90

0 0.05 0.1
t

Generator assembly which gives triangular alternating impulses in program

Electronics Workbench. The required impulse form we assembly as

13
superposition of sinusoidal sources with corresponding frequencies, amplitudes
and phase shifts.

Example 2: Trapezoidal impulse expansion in a Fourier series

2  1
T  0.05 t  0 0.01 T  2 T   f 
T T

t 3 T
f( t )  if 0  t 
T 3
1.2
T 2 1
1 if t T
3 3 0.8
 1  t   3 if T  2  t  T f( t ) 0.6
 
 T 3 0.4
0 otherwise 0.2
0
0 0.017 0.033 0.05
t

14
  1.333 
 0.456 
T  
2   0.113 
M  5 k  0  M a   
k T 
 
f( t )  cos   k t dt a
 0  k
A  i a
k
0
 0.028 
 
 0.018 

Two forms of expansion

a M A M
 k   
0 0
F( t )   a  cos   k t F1( t )    A  sin    k t  arg  A   
2 2  k   k  
k 1 k1

Expansion check

arg  A 
A 
 k
k deg
1.2
1.333 90
F( t ) 0.456 -90
0.113 -90
f( t ) 5.636·10-5 -90
0.6
F1( t ) 0.028 -90
0.018 -90

0 0.05 0.1
t

Generator assembly which gives trapezoidal alternating impulses in program

Electronics Workbench. The required impulse form we assembly as
superposition of sinusoidal sources with corresponding frequencies, amplitudes
and phase shifts.

15
1.3.1. Bode plot

Bode plot is the ratio between complex representation of output variable and
complex representation of input variable which is being changed according to
potential law.
y (t )  Ay  sin  t . (16)

By y(t ) change according to (16) at linear component output in steady state the
output variable will change according to potential law with the same frequency, but
different amplitude and phase.
x (t )  Ax  sin( t   ) . (17)

Complex representation (16) and (17)

y  Ay  e j t ; x  Aх  e j ( t ) . (18)

Hence, W ( j ) which stands for Bode plot equals

16
 jt
x Aye e
W ( j )   jt
 e j  A( )  e j ( ) . (19)
y Aх e

In (19) А() and () functions are Bode magnitude plot and Bode phase plot.

Bode plot (19) which is a complex variable has the algebraic form of
W ( j )  U ( )  jV ( ) . (20)

In (20) U() and V() are real and imaginary frequency characteristics.

Bode plot is interconnected to the Transfer function.

Mind that formula for derivative of n-order from от y and x are equal

dk y
k
 ( j ) k  Ay  e j t  ( j ) k  y;
dt ( 21)
dkx
k
 ( j ) k  Aх  e j t  ( j ) k  x,
dt

By substituting (18) in the differential equation (9) we obtain

b0  ( j ) m  b1  ( j ) m1  ...  bm1  ( j )  bm R( j )
W ( j )   (22)
a0  ( j ) n  a1  ( j ) n1  ...  an1  ( j )  an Q ( j )

Bode plot has its real and imaginary parts.

W ( j )  A( )e j ( )  U ( )  jV ( )
U ( )  A( )cos( ( )), V    A( )sin( ( ))
 V ( )  0 при U    0
A( )  U ( ) 2  V ( ) 2 ,  ( )  arctg  
 U ( )   при U    0
As shown in correlation (9) and (22) that formula
for Bode plot is obtained from substituting in
transfer function p  j .

Figure 6

17
To solve practical problems of control systems properties Bode plot is plotted on
complex planes U() and V() as locus which is a geometric part of vector end
W(j) at frequency change from 0 to  (Figure 3).

1.3.2. Correlation between component input and output variables in steady

state.

Correlation in steady state can be obtained if we assume all the derivatives

equal zero in transfer function p  0 .
Hence

bm
x  y   W (0)  k. (23)
an
bm
Coefficient k  is called component coefficient of amplification
an
1.3.3. Steady and minimal phase components frequency characteristics
peculiarities.
The system is described by A( ) and e j ( ) or U ( ) and V ( ) variables. But
for some types of components (steady component) there exists the correlation
between the functions which formed these pairs and one of such functions can be
used to describe the system. To determine the weight function for instance the
following ratio should be used:
 
1 1
w(t ) 
  U   cos t  dt

or w(t )  
  V   sin t  dt


On integrating these formulae we obtain the formulae for transfer function:

1

U   1

V  
h (t )   sin t  dt or h(t )   cos t  dt .
 
  


These correlations are for steady components in which denominator poles of

transfer function have negative real parts.

18
Мinimal phase component is a component which have poles and zeros with
negative real parts. That means they have minimal phase shift at frequency 
compared to other components.
Example:
k
W ( p) 
Tp  1

Calculate Bode plot

k
W ( j ) 
jT   1

Calculate its real and imaginary parts

k 1  jT   k  jkT  k kT 
W ( j )  U ( )  jV ( )     j
1  jT  1  jT   1  T 2 2 1  T 2 2 1  T 2 2
k kT 
U ( )  , V ( )   ,
1  T 2 2 1  T 2 2
k 2  k 2T 2 2 1  T 2 2 k
A    U    V   
2 2
k 
1  T  
2 2 2
1  T  
2 2 2
1  T  
2 2

 V ( ) 
    arctg    arctg  T     arctg T  
 U ( ) 


When  is changed,  is changing from 0 to  .
2

Let’s analyze the block with transfer function

k
W ( p) 
Tp  1

Do the same procedure

19
Calculate its real and imaginary parts:

k 1  jT   k  jkT  k kT 
W ( j )  U ( )  jV ( )     j
1  jT  1  jT   1  T 
2 2
1 T 
2 2
1  T 2 2
k kT 
U ( )   , V ( )   ,
1  T 2 2 1  T 2 2
k 2  k 2T 2 2 1  T 2 2 k
A    U    V   
2 2
k 
1  T  
2 2 2
1  T  
2 2 2
1  T  
2 2

 V ( ) 
    arctg      arctg T     , arctg T    (0,  / 2)
 U ( ) 


When  is changed,  is changing from  до  . Thus the second
2
component makes greater delay.

1.4. Some typical components of control systems

1.4.1. Amplifying component
Let’s compute transfer function of inertia-free component
U1 U 2 U U
I1  , I2  , I1  I 2  1   2
R1 R2 R1 R2

U2 R
  2
U1 R1

Component transfer function:

U 2 ( p) R
W ( p)   2 k
U1 ( p ) R1

U1 U U2 U U U2
I1   , I2  1 , I1  I 2   1  1
R1 R2 R1 R2

20
 U 2 U1 U1 U 2  1 1
  ,    U1 Component transfer function:
R2 R2 R1 R2  R2 R1 

U2  1 1
   U1 ,
R2  R2 R1 

U 2 ( p)  1 1   R 
W ( p)      R2   1  2   k ,
U1 ( p )  R2 R1   R1 

1.4.2. Aperiodic inertial component of the first period

Consider the example of aperiodic inertial component. Electrotechnical

diagram shown below can be such components.

Develop the formula for inertial component with capacity:

dU C dU
iC  C , U C  iC R  U  U C  RC C  U
dt dt

Let’s represent for time invariable T  RC :

dU C
UC  T U
dt

U C  TpU C  U  1  Tp U C  U

In our case:

X 1
1  Tp  X  Y W 
Y 1  Tp

21
We can obtain the same formula through diagram resistance having complex
representation:

1
Z ( )  R  ,
jC

1
Find current I ( )  U / Z ( ) , voltage on the required element U C  I ( ) ,
jC
then substitute all the values and obtain:

1 U 1 U 1 U
U C  I ( )    ,
jC Z ( ) jC  1  jC  jCR  1
 R  jC 
 

UC 1 1
 W ( j )   W ( p) 
U  jCR  1  pT  1

Thus transfer function of inertial component is:

k
W ( p) 
 pT  1

1  t /T
And its weight function(Green’s function): w(t )  ke
T

To calculate its transfer function h(t ) its impulse response function W ( p) with
1/ p multiplier can be used

W ( p) 1 k A(0) A( p) pt k
   e k et /T  k 1  et /T 
p p  pT  1 B(0) pB( p) 1
T
T

Check

k  t /T
w(t )  h(t )  e
T

22
For the second example do the same calculations taking into consideration time
constant equals T  L / R :

UR 1 1
 W ( j )   W ( p) 
U  j L / R  1  pT  1

Plot graphic responses dependences:

1 100

0.8 80

0.6 60
h( t ) w( t )
0.4 40

0.2 20

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0 0.01 0.02 0.03 0.04 0.05 0.06
t t

Bode plot is described by formulae which we have already analyzed:

k 1  jT   k  jkT  k kT 
W ( j )  U ( )  jV ( )    j
1  jT  1  jT   1  T  1  T 
2 2 2 2
1  T 2 2
k kT 
U ( )  , V ( )   ,
1  T 2 2 1  T 2 2
k 2  k 2T 2 2 1  T 2 2 k
A    U    V   
2 2
k 
1  T  2 2 2
1  T   2 2 2
1  T  
2 2

 V ( ) 
    arctg    arctg  T     arctg T  
 U ( ) 


When  is changed,  is changing from 0 до  .
2

Plot the locus for Bode plot at k=1 and T=0.1c using MathCAD

k
T  0.1 k  1 W ( p  k)  A    k  W  j    k U   Re A    k  V   Im A    k 
T p  1
arg  A    k  3
     0  .1 10
deg

23
 0.5 1
0.8
0.25
0.6
V(  ) A (   k)
0 0.25 0.5 0.75 1 0.4

0.25 0.2

0.5 0 10 20 30 40 50

U(  )

0 10 20 30 40 50

30
(  )
60

90

1.4.3. Oscillatory, aperiodic and conservative components

Parallel series RLC circuit can be the

L R oscillatory component. Let’s calculate the voltage
y=U C x=UC
of capacitance

dU C di
iC  C , U L  L L , iC  iL , U C  iC R  U L  U
dt dt
d 2U C dU C
L 2
 RC  Uc  U
dt dt

Let’s write down the formula for this equation

d 2U C dU C
LC 2
 RC  Uc  U   LCp 2  RCp  1Uc  U
dt dt

Calculate transfer function:

24
 U C ( p) 1 1
W ( p)    2 2 , T2  LC , T1  RC
U ( p )  LCp  RCp  1 T2 p  T1 p  1
2

Or through resistance:

Z ( p)  R  pL  1 / Cp, I ( p )  U ( p) / Z ( p )
1 U ( p) 1
U C ( p)  I ( p)   U ( p) /  Lp 2  RCp  1
pC R  pL  1 pC
pC
W ( p )  U C ( p ) / U ( p )  1 /  LCp 2  RCp  1

Calculate Green’s function – weight function:

W ( p )  A( p ) / B( p )  1 /  LCp 2  RCp  1 ,
2
 RC  R 2C 2  4 LC  R  RC  1
B ( p )  0,  p1,2  ,     2
2 LC 2L  2L  L
2
R 1  RC 
  ,    , p1,2     j
2L L2  2 L 

Then the weight function equals:

 A( p1 ) p1t 
w(t )  2Re  e 
 B  p1  

Now transfer function can be calculated:

W ( p) 1
 A( p ) / pB ( p ) 
p p  LCp  RCp  1
2

A(0)  A( p1 ) p1t   2   2 t    

h(t )   2Re  e  1 e sin  t  arctg   
B  0  B  p1       

Take into consideration the second order component behaviour in MathCAD.

25
In program determine the circuit impedance and transfer function after that. Both
impulse response function and Green’s function are determined with the help of
Laplace transform.

6
L  0.1 C  100 10 R  10

1 1
Z( p )  R  p  L  W ( p ) 
C p Z( p )  C p

solve  p ( 50.)  312.2 i 100000.

p  Z( p )   W ( s ) simplify 
float  4 ( 50.)  312.2 i 2
100. s  s  100000

2 d
B( s )  100. s  s  100000 A ( p)  100000 B'( p )  B( p )
dp

invlaplace  x (  50.)  t
w( t )  W ( x)  320.3 e  sin ( 312.2 t )
float  4

W ( x) invlaplace  x (  50.)  t (  50.)  t

h ( t)   1.  1. e  cos ( 312.2 t)  .1601 e  sin ( 312.2 t)
x float  4

Determine the same functions through theory analysis

 A  p2 p 2 t   A  p 2 p 2 t  A ( 0)
w2( t)  2 Re e h2( t)  2 Re e 
 B' p2   B' p 2  p 2  B( 0)
   

250 1.6
1.4
200
w( t ) 1.2
150 h( t ) 1
d 100 0.8
h( t ) h2 ( t ) 0.6
50
dt 0.4
0 0.01 0.02 0.03 0.04 0.05 0.06 0.2
w2( t ) 50
100 0 0.01 0.02 0.03 0.04 0.05 0.06

150 t

In general oscillatory component equation has the form of:

d 2x dx
a2 2  a1  a0 x  b0 y .
dt dt

26
or

a2 d 2 x a1 dx b
2
 x 0 y
a0 dt a0 dt a0

d 2x dx
 2  2  x  ky
dt dt

At static mode we have :

b0 x
x y  x  ky  W (0)   k
a0 y

Characteristic equation:

d 2x dx
 2
 2  x  ky
dt dt
X k A( p )
 p 2
 2 p  1 X  kY  W ( p )   
Y  p  2 p  1 B ( p )
2

Denominator fraction B ( p ) has two roots which can be joined(united) complex,

p1,2    j , or right half plane(RHP) or left half plane p1 , p2  0 . If the

characteristic equation which is a differential equation describing the second order

component has left half plane roots then the component is called aperiodic
component of the second power.
Calculate Bode plot of oscillatory component:
k
W ( )  U ( )  jV ( ) 
 j 2  1   2 
 
k 1   2  k 2
j ,
4   1  
2 2

2 2
4   1  
2 2

2 2

k 4 2 2  1   2 
2
k
A( )  U ( ) 2  V ( ) 2  
4 2 2  1   
2 2
4 2 2  1   2 
2

27
Let’s consider the example of frequency characteristics of locus assembly with
different values of damp coefficient  through transfer function with MathCAD
help.

k
k  0.1   0.01   0.002 W  p   
2
  p  2  p  1
solve  p ( 5.)  21.79 i
A      W  j    U     Re A     
2
  p  2  p  1  
float  4 ( 5.)  21.79 i
arg  A    k 
V     Im A         k 
2
  0  0.01 10  0.5
deg

0.3 0.2 0.1 0 0.1 0.2 0.3


V  
 0.1

 2 0.2
V(   )
0.3
V(    2)
0.4

0.5


U  

  U(   )  U(    2)
 2

0.45
0 10 20 30 40 50

   0.36  
A       
 2  0.27 60
 2
A (   ) (   )
0.18
A (     2) (    2) 120
0.09

0 10 20 30 40 50 180



One more type of second order component is conservative (self-vibrating)

component:

28
 d 2x
  x  ky
dt 2
X k
 p 2
 1 X  kY  W ( p )  
Y  p 2  1

All mentioned above components are static components, because at

substituting in transfer function W ( p), p  0 we obtain a certain final number and
formula is the same for all the components:

1. Amplifying (noninertial, static) component: W ( p)  k , W (0)  k

k
2. Aperiodic(Inertial)component: W ( p)  , W (0)  k
Tp  1

k
3. Oscillatory component: W ( p)  , W (0)  k
 p  2 p  1
2

4. Aperiodic component of the second order:

k
W ( p)  , W (0)  k
 p  2 p  1
2

k
5. Conservativе (self-vibrating) component: W ( p )  , W (0)  k
 p2  1

1.4.4. Astatic integrated component

Let’s analyze the example of integrated

component:
C
R
dU 2
U1
i1  U1 / R, iC  C , i1  iC
U2 dt
t
dU 2 1
RC 0
C  U1 / R  U 2 (t )   U1 (t )dt
dt

29
 t
dx
 ky  pX  kY  W ( p )  k / p  x  k  ydt
dt 0

Green’s function (weight function):

W ( p)  k / p, w(t )  L1 W ( p)  k

Transfer function:

W ( p)  1 
W ( p )  k / p, h(t )  L1    L1  2   kt
 p  p 

Check:

w(t )  h(t )  kt   k

Determine frequency response and bode plot :

W ( j )  k / j , W ( j )  U ( )  jV ( )   jk /  , U ( )  0, V ( )   k / 
A( )  W ( j )  k /  ,  ( )   / 2  90o

k
k  1 W ( p )  h ( t)  t k w( t)  k    arg W  j  
p

1.2 0.06
1
w(t)
0.8 0.04
d 0.6 h(t)
h(t)
dt 0.4 0.02
0.2

0 0.012 0.024 0.036 0.048 0.06 0 0.012 0.024 0.036 0.048 0.06
t t

30
 0 200 400 600 800 1000
30
()
60
deg
90

120

7. Real integrated component possess some inertia

k
Tp  1 pX  kY  W ( p) 
Tp  1 p

The examined component is not simple because it has two blocks/components in

series - integrated and intertial(aperiodic).

k k 1
W ( p)    W ( p )W2 ( p )
Tp  1 p p Tp  1 1

1.4.5. Differential component.

Impulse response function of differential component has the form of:

dy
xk , X  kpY  W ( p)  kp
dt

Let’s find transfer function:

W ( p)  1  p k 
h(t )  L1 
p   L  p   k (t )
   

Now Green’s function can be obtained:

31
 w(t )  h(t )  k (t )

At random external influence we obtain:


x (t )  

f ( )h(t   ) d  kf (t )

Bode plot:

W ( j )  jk  U ( )  jV ( ), U ( )  0, V ( )  k
A( )  W ( j )  k ,  ( )   / 2

Let’s tabulate all transfer function and their properties in Table 1

Table 1

Component The impulse response Weight function Transfer function

type function
1. Inertia-
free
(amplifying W ( p)  k
component)

2. Ideal
differential
component W ( p )  Tд  p

3.Real Tд  p
differential W ( p) 
component
Tд  p  1

4.Integral 1
component Wint ( p ) 
Tint  p

32
5. Inertial K
component W ( p) 
of the first
T1  p  1
degree

6. Inertial Characteristic equation roots

component p1    j ; p2    j
of the
second
power K
W ( p) 
(oscillatory) T1  p  T2  p  1
2

7.Inertial Characteristic equation roots

component p1  1 ; p2   2
of the
second
degree K
W ( p) 
(aperiodic) T1  p  T2  p  1
2

Things to remember:
1. All the components can be divided into two groups which differ a lot from
each other.
 Static components are components which in steady state (at p=0) have a
certain connection between output and input variables. These components
have stable equilibrium. All components given in Table 1 are static
components but integral one.
 Astatic components have neutral equilibrium
Integral component is astatic in Table 1. At
p  0 : Wи (0)   . This means that at xвх  1(t )
theoretically is xвых   .In input influence is
applied on limited time then output variable remains
constant, i.e. equals the integral of input variable,
Figure 7
for this time (Figure 7).
2. There is an interconnection between the transfer function W ( p) and impulse
response function h(t ) capable of transient process h(t ) . When substituing
33
 p  0 in impulse response function W (0)  h() we obtain steady state in
time domain. When substituting p   in impulse response function
W ()  h(0) we obtain transient process beginning.
3. Frequency response locus always starts from real axis because real part of
frequency response U ( ) is an even function , and imaginary partV ( ) is an
odd function: W ( j )  U ( )  jV ( )  W (0)  U (0) , V (0)  0 .
4. Frequency characteristics are determined by the following factors:
 oscillation indicator M  Amax ( ) / A(0) – characterises system tendency
to oscillations :the higher it is the less qualitative is the system (1.1< M
<1.5)
 Resonance frequency is the frequency at which magnitude frequency
response is at its hight(at this frequency oscillations have maximum
amplification)
 System bandwidth – interval from   0 to 0 at which the following

condition is fulfilled A(0 )  0,707 A(0)

 Frequency cutoff A(ср )  A(0) indirectly characterizes the duration of

transient process, the following correlation is true Ty  (1  2)2 / ср

More

The greater the peak of resonance frequency the more intensive are the oscillations
34
 T  2 /  p

1.5. Block diagrams

1.5.1. Control system transfer function calculation by blocks transfer
functions

A control system consists of a number of components connected to perform

a desired function. Once a component (system or subsystem) is reduced to a
mathematical model, it can be represented as a block, with the component
operation described by the mathematical function. It is convenient and useful to
represent the elements of a control system by blocks. The properties of the block
are contained in the transfer function represented by the Laplace transform. Many
control systems involve several blocks representing the controller actions, the
plant, feedback elements, and other functions. A block diagram is a pictorial
representation of an entire control system in terms of all the elements and their
transfer functions.

When constructing structural diagrams of control systems the following symbols

illustrated in diagrams below are helpful as well
 Signal branching sign

 The summing point also known as a summing junction is the block used to
represent the addition/subtraction of signal as shown in Figure below

35
 or

х4 = х1 + х2  х3.

1.5.2. Blocks in series

xвх(p) xвых(p)
W1(p) W2(p) ... Wn(p)

Figure 8

From transfer function definition it follows that

Wэкв ( p )  W1 ( p )  W1 ( p )  ...  Wn ( p ) . (24)

1.5.3. Blocks in parallel

Figure 9

Parallel subsystems have a common input and output formed by the algebraic sum
of the outputs from all the subsystems.

Hence equivalent transfer function will be equal to

Wэкв ( p )  W1 ( p )  W1 ( p )  ...  Wn ( p ) (25)

36
1.5.4. Feedback control system

Figure 10

Closed loop control systems are divided into two categories , based upon the nature
of the feedback signal: negative-feedback closed loop and positive-feedback
closed loop systems. The system consists of one block in the forward path and one
block in the feedback path. Hence we can write:
хвх1 ( p )  хвх ( p )  хос ( p ),
(26)
where, хос ( p )  Wос ( p )  хвых ( p ).

In (26) sign "+" corresponds to positive feedback, and sign"" corresponds to

negative feedback.

The transfer function of reduced form of a closed loop control system can be
obtained from
хвых ( p )  W1 ( p )   хвх ( p )  Wос ( p )  х вых ( p )  . (27)

From (27) calculate the required transfer function

хвых ( p ) W1 ( p )
Wэкв ( p )   . (28)
хвх ( p ) 1  W1 ( p )  Wос ( p )

Notice that when obtaining the formula (28) sings corresponding to positive and
negative feedback are changed into the opposite ones. In (28)"" sign corresponds
to positive feedback and "+" sign is negative.

1.5.5. Block diagram reduction

Block diagram reduction help reduce separate elements of block diagram to three
block types described above.

Block diagram reduction is quite simple and illustrated below.

37
1.5.6. Summing point moving

Original diagram

х2 ( p )   х1 ( p )  W1 ( p )  х3 ( p )   W2 ( p )
.
 х1 ( p )  W1 ( p )W2 ( p )  х3 ( p )W2 ( p )

Moving a summing point to the output

Moving a summing point to the input

It is clear that the relation between х2 ( p ) and х1 ( p ) in diagrams b and c is the

same as in а diagram.

38
1.5.7. Branching point moving

Equivalence at branching point moving is to keep the relation between input

value х1 ( p ) and branching moving point value.

Original diagram

х3 ( p )  W1 ( p )  х1 ( p ) .

Moving a summing point to the output

Moving a summing point to the input

х1(p) х2(p)
W1(p) W2(p)

W1(p)

х3(p)

It is clear that relation in b and c diagrams is the same as in а diagram

39
 х3 ( p )  W1 ( p )  х1 ( p ) .

1.5.8. Examples of diagrams reduction rules

1. Blocks in series:

2. Blocks in parallel:

3. Closed loop system reduction with negative feedback:

4. Closed loop system reduction with unit negative feedback:

W1 ( p)W p ( p )
1  W1 ( p )W p ( p )

5. Application of block diagrams reduction rules:

40
 W1 ( p) W2 ( p)
1  W1 ( p)Wp1 ( p) 1 W2 ( p)Wp2 ( p)

W1 ( p)W2 ( p)
1  W1 ( p)Wp1 ( p)  W2 ( p)Wp 2 ( p)  W1 ( p)Wp1 ( p)W2 ( p)Wp 2 ( p)  W1 ( p)Wp1 ( p)W ( p)

6. Block reduction with takeoff point moving ahead of a block:

W1( p)W2 ( p)
1  W4 ( p)W2 ( p)  W1( p)W2 ( p)

7. Block reduction with takeoff point moving:

W1( p)W4( p)  W1( p)W2( p) W5( p)W3( p)

8. Application of superposition principle

Find the relation between input хin and output value хout , and relation between
disturbance f and output value хout :

41
 W2 ( p ) W1 ( p ) хin  f 
хout 
1  W1 ( p )W2 ( p )W p ( p )

1.6. Control system generalized equivalent circuit

Applying block diagram reduction rules any block diagram can be reduced
to two blocks diagram (Figure 6). Consider one of the blocks to be equivalent
controlled object with transfer function Wо ( p ) and the second block to be
equivalent controller with transfer function Wc ( p) .

Figure 11

In shown in Figure 8:

x - controlled variable,

xref - reference input signal,

42
 xd - disturbance,

xc - control response

To figure out how the control signal influences control system behavior let us
represent controller block diagram as three blocks in parallel: inertialess, ideal
differentiated and integral(Figure 9).

Figure 12

Controller transfer function according to diagram in Figure 9

1
Wc ( p )  K р  Td  p  . (29)
Tint  p

The obtained equivalent diagram shows the influence of object parameters and
controller on control system properties in steady state and transient processes.

System properties are determined by the dependence of control response on input –

reference or disturbance.

As all system blocks are taken to be linear, then to obtain the required dependences
we use superposition principle. Hence the following transfer functions appear.

43
 By reference input By disturbance input
Superposition principle

1. Transfer function by reference signal

х( p)
Wref ( p )  ; at хref ( p )  0 .
хref ( p )

2. Transfer function by disturbance action channel

х( p )
Wd ( p )  ; at хd ( p )  0 .
хd ( p )

For analysis use the transfer function for open loop system. To determine the
transfer function break the system at "а" point(Figure 8) let хd ( p )  0 . Then
х( p)
W раз ( p )  .
хref ( p )

Based on block diagrams reduction we obtain

Wо ( p )  Wc ( p )
Wref ( p )  ; (30)
1  Wо ( p )  Wc ( p )

Wо ( p)
Wd ( p)  ; (31)
1  Wо ( p)  Wc ( p)

44
 W раз ( p )  Wо ( p )  Wc ( p ) . (32)

According to superposition principle

х ( p )  Wз ( p )  х з ( p )  Wв ( p )  хв ( p ) . (33)

Analyze the influence of controller transfer function. Consider that transfer

function of the object is stable and is a static component, at
p0  Wоб (0)  K об ( in steady state).

1.6.1. Control system behavior in steady state

The relation between controlled variable and perturbation action in steady

state can be obtained (33) at p  0

х  Wз (0)  х з   Wв (0)  хв  (34)

Let us consider steady state at using controllers of different types

1.6.2. Proportional controller (P controller)

The transfer function of controller in this case is given by

Wр ( p)  K р . (35)

Hence (24) let Tд  0; Tи   .

Consider that Wоб (0)  K об then (33) obtain at p  0

K об  K р K об
х   xз    xв  . (36)
1  K об  K р 1  K об  K р

Double sign (36) is used because disturbance in some cases can increase or
decrease the controlled variable.

45
As(36)shows, when P controller is used for the object described by static unit we
have a mistake in reproduction of reference input signal which equals
 Kоб  K р 
xз   х  хз      1   хз  . (37)
1 K  K 
 об р 

Residual is the difference of the obtained value x from the desired value х з
(reference value).

Some influence on the controlled variable of disturbance remains which is equal to

K об
xв   хв  . (38)
1  K об  K р

From (37) and (38) it follows that, the bigger the gain factor is, the more precisely
is the reproduction of disturbance and the less is the influence of disturbance.

Control systems which possess the after-effect of disturbance are called static
systems. Dependence of controlled variable form disturbance action is shown in
Figure 10.

In some engineering fields the dependence х ( xв  ) is called control characteristic,

and slope coefficient of this characteristic is called statism coefficient , which is
determined as follows(Figure 10)
х
Kс  .
хв

By letting x and xв some basic values x0 and xв0 statism coefficient can be
determined in relative units
х xв 0
K с*   . (39)
x0 хв

46
 Figure 13. Controller output characteristic and external characteristic

In power supplies there is an analog dependence –

static characteristics of turbine controller (controller
external characteristics). Such a characteristic help
control frequency change when load rise occurs.

 P0
s , for dependance  ( P )
P  0

It is more often used as 1 / s for dependence P( ) .

K об
Example: The object with transfer function is given Wоб ( p )  ,
Tоб  p  1
Tоб  0,1 K об  0,5 x з  1 . Calculate transfer function for closed loop at P
controller presence.

WР ( p )WОб ( p ) xЗ x
W ( p)  +
WР(p) WОб(p)
1  WР ( p )WОб ( p) -

Let’s solve the example using MathCAD

0.5
T  0.1 W ( p ) 
p T  1

47
 W(p )k W(p ) invlaplace  p (  10.)  t
W1( k  p )  h ( t )   ( .50000)  e  .50000
1  W(p )k p float  5

invlaplace  p
 .50000 k 
W1( k  p ) 10. 10. (  1.)  ( 10. 5. k)  t
h1( t  k) 
float  5
 e 
p  10.  5. k 10.  5. k 

t  0  .001 6 T

1.2

1
h( t )

h1( t  10) 0.8

h1( t  15)
0.6
h1( t  65)

1 0.4

0.2

0 0.1 0.2 0.3 0.4 0.5 0.6

t

Proportional controller minimizes the static error and decreases transient process
time

1.6.3. Proportional derivative controller(PD controller)

Controller of this type has transfer function

Wc ( p)  K р  Td  p . (40)

As at p  0  Wc (0)  K р , then the qualities of system with PD controller in

steady state regime coincide with P controller system.

The component of transfer function which proportional to the derivative term

influences much the qualities of control system in transient process.

1.6.4. Integral controller ( I controller)

I controller transfer function

48
 1
Wр ( p)  . (41)
Tи  p

Transfer functions on reference signal and disturbance in system with I controller

Wоб ( p )  W р ( p ) Wоб ( p)
Wз ( p )  ; Wв ( p) 
1  Wоб ( p )  W р ( p ) 1  Wоб ( p)  Wр ( p)

Wоб ( p )
Wз ( p )  , (42)
Tи  p  Wоб ( p )
Wоб ( p )  Tи  p
Wв ( p )  . (43)
Tи  p  Wоб ( p )

At p  0 Wз (0)  1; Wв (0)  0 . Hence at given equivalent block diagram the

application of integral controller provides the precise reference signal reproducing .
It also eliminates the influence of perturbation action of controlled variable.

Automatic system control at which the regulated component does not depend on
perturbation action is called astatic.

1.6.5. Proportional integral controller (PI controller)

Contoller transfer function is given as

1
Wр ( p)  K р  (44)
Tи  p

Transfer functions by reference and disturbance signals

Wоб ( p )  ( K р  Tи  p  1)
Wз ( p )  . (45)
Tи  p  Wоб ( p )  ( K р  Tи  p  1)

Wоб ( p)  Tи  p
Wв ( p )  . (46)
Tи  p  Wоб ( p )  ( K р  Tи  p  1)

The same as in system with I controller at p  0 Wз (0)  1; Wв (0)  0 .

Thus, adding to proportional part of integral reference also provides astatism of
control system.

49
 K об
Example: The object with transfer function Wоб ( p )  is given,
Tоб  p  1
Tоб  0,1 K об  1, x з  1 . Find the transfer function for closed loop at presence

P W p ( p )  k п  , I W р ( p )  kи / p  and PI W р ( p )  k п  kи / p  controllers.

Solution:

WР ( p )WОб ( p)
W ( p) 
1  WР ( p)WОб ( p)

K об kп K об kи
P  W ( p)  , P  W ( p) 
Tоб p  1  K об kп  Tоб p  p  K об kи
2

K об  kп p  kи 
PI  W ( p ) 
Tоб p 2  p 1  kп K об   K об kи

1.4 1.1

1.2 1.06

h1 t  0  10 1 h1 t  0  8 1.03

h1 t  20  0 h1 t  20  0
0.8 0.99
h1 t  15  20 h1 t  15  20

h t  0.6 h t  0.95
1 1
0.4 0.92

0.2 0.88

0.85
0 0.21 0.43 0.64 0.86 1.07 1.29 1.5 0 0.21 0.43 0.64 0.86 1.07 1.29 1.5
t t

Увеличенный фрагмент начало процесса

1.6.6. Proportional integral derivative controller (PID contoller)

The transfer function is given as

1
W р ( p )  K р  Tд  p  . (47)
Tи  p

50
It is evident that in steady state regime ( p  0 ) system with PID controller
possesses the same qualities as the system with PI controller.

Example: Let’s analyze voltage control circuit of synchronous generator.

Generator excitation current changes at rheostat resistance change. Rheostat

engine shift is performed by internal combustion engine. Engine is an integrated
component, because turn of its shaft angle     dt is proportional to rotation

speed integral  .Rotation speed  if proportional to voltage given to armature UУ

.When mismatch between U з and U an amplifier input signal occurs, which is
transmitted to excitation winding through amplifier and integrated component.
Excitation winding amplifies or reduces magnetic field till the mismatch
disappears U  U з  U

Block diagram

TУ , TД , TВ , TДН are amplifier, engine, excitation loop or divisor time constants

о, k Н is load coefficient.

51
1.7. Transient process calculation

1.7.1. Transient processes calculation in state-space

State space is one of the most effective methods of transient processes

calculation. The method essence is to transform n order differential equation into
the system of first order n differential equations written in Cauchy form. For
example for RLC series circuit equations in Cauchy form can be written as:

diL R 1 e(t )
  iL  uC  ;
dt L L L
duC iL

dt C

If state vector is considered as x(t )  iL (t ), uC (t ) and write out the

coefficients at unknown variables and given influences in matrix form then the
equations can be written as:

dx(t )
 A  x(t )  B  V(t ) или D(x, t )  A  x  F(t ) ,
dt

y (t )  Cx(t ) - coupling equation

where А is square state matrix of n  n dimension;

V(t) is vector m of given influences;

В is matrix of n  m dimension, elements of which are determined by circuit

parameters and its structure.

52
 For series RLC-circuit matrices are written down:

 R 1
 iL (t )   L  L  1
x(t )   , A , B  L.
u (t )  1  
 C 
 0   0
 C 

The most complicated thing in state space method is A and B state matrices
formation. There exist several effective methods of state space equations
formation, but they all are based on Kirchhoff’s equation system with elimination
of variables which are not state space variables .MathCAD has functions which
help solve equations in symbolic form and this simplifies state space equations
and respective matrices formation. Below we analyze the transient process
calculation by using state space approach in MathCAD.

Example 44. Determine inductance current iL (t ) and voltage on circuit elements

(Figure 34) for the given circuit:
uC1 (t ), uC 2 (t ), u R 2 (t ), u R1 (t ),
u R 2 (t ), u L (t )

R2 L i L (t)
e(t )  Е  20 V , J  1 A
E1(t) C1 R2 C2
iC 1 (t)
J(t)
R1 iC 2 (t) если R1  20  , R2  100  ,
Рис. 34 C1  100  F , C2  50  F ,
L  0,1 H .
Solution. To form state space equations let’s
write down circuit equations according to Kirchhoff’s laws, then use computing
unit Given Find. Let’s perform differential equation integration numerically, using
rkfixed function as shown in MathCAD file below.

53
 UC2
iL C  U'C2  J
2 R
2

iL R  L i'L  UC1  UC2 0 Given

1
Circuit equation written according to Kirchhoff’s laws

E iL  C1U'C1 R2  UC1

 E  R  iL  UC1
2 
 
 R  C 
2 1
 
 R   iL  UC2  J R 

A  UC1 UC2 iL E J  Find U'C1 U'C2 i'L   2 2
 C R 
 2 2 
  iL R1  UC1  UC2 
 
 L 

Matrices formation

A  augment ( A ( 1  0  0  0  0)  A ( 0  1  0  0  0)  A ( 0  0  1  0  0) ) B  A ( 0  0  0  E J)

54
  1 0
1 
 R C   E 
C  
 2 1 1   R2 C1 
 1 1 
A  0  B   J 
C R C  
 2 2 2   C2 
 R 
 1 1 1
  0 
 
 L L L 

6
R  20 R  100 C  100 10 6 C  50 10 L  0.1 E  20 J  1
1 2 1 2

 100 0 1  104   2000 

 
A   0 200 2  104  B   20000
 
   0 
 10 10 200 

State space matrix eigenvalues determination which should equal roots of

characteristic equations

 132.84 
  eigenvals A   183.58  545.458i

 
 183.58  545.458i

1 1
Z( p )   R  L p 
1
 1  C p   1  C p 
R 1  R 2 
 2   2 

solve  p 

( 183.58)  545.46i
p  Z( p )  ( 183.58)  545.46i

float  5  
 132.84 

Roots of characteristic equation and eigenvalues coincide, that means А

matrix is formed correctly. To check В matrix let’s determine forced

components of state variables which should satisfy the equality –А-1 В.

55
 1
 34.545  E R J R 
A  B   45.455 UC1пр   R   E
2 2

  R R R R  R  R 2
 0.545   2 2 1 2 2 1 
UC1пр  34.545

UC2пр  R 
E
R U

J R  R
2 1 
R R R 2 R R R 2 C2пр  45.455
2 2 1 2 2 1

J R
E 2
iLпр   iLпр  0.545
R R R R R R
2 2 1 2 2 1

To determine additional values form the connection matrix С:

UR2 E  UC1

UR1 R  iL
1

UL UC2  UC1  R  iL
1

 1 0 0 
 
C   0 0 R1 
 1 1 R 
 1

Solve differential equation system by Rounge-Kutta method:

0 0
Dt  x  A x  B xo   0  N  400 T  0.05 i  0  N x  rkfixed xo  0  T  N  D  t  x
 
0

 1  
 UR1   x  i
 i    E
 UR2   C  2     0 
 i  x  i  
 UL   3   0 
 i   
 x  i
 


t  x 0 i  0  N

56
 60
50
 x1 
 i 40

 x2  30
 i 20
U R1 10
i

U R2 0 0.01 0.02 0.03 0.04 0.05

i 10

UL 20
i 30
40
50

ti

 x3  2.5
 i
U R1 2
i
R1 1.5

U R2 1
i
R2
0.5

0 0.01 0.02 0.03 0.04 0.05

ti

To determine state space components derivatives the extended state space

matrix should be multiplied by diagonal matrix of capacitances and inductance

 iC1    x1  

 i   1 C 0 0  i
      
 iC2    0 C 0  Dt   x 2  
 i 2 i i
    
 UL   0 0 L    3  
 i x i
  

To animate the transient process it is enough to substitute the number of

iterations on time interval N for FRAME variable. The in diagram editing
window in позиции Traces activate Hide arguments to hide arguments at
axis coordinates. If you wish you can end diagram by black point. Keep in mind
that by animating arguments margins at diagram axis should be fixed.

57
 Diagram fragment for the 70th frame at FRAME=70 and closed arguments.

k  FRAME i  0  k

60 3

50
2.5
40

30 2
20

10 1.5

0 0.01 0.02 0.03 0.04 0.05 1

10

20 0.5
30

40 0 0.01 0.02 0.03 0.04 0.05

50

Below we give the example of transient process calculation in MathCAD,

Electronics Workbench и MATLAB for dynamic system described by equation:
  
x  x  x  1, x(0)  0, x(0)  0

MathCAD

d
x1 x2 A      
 0 1  B   0   A  1 B   1 
     eigenvals ( A )
dt  1 1  1 0
d
x2 x1  x2  1
dt

 0.5  0.866i  0  
  D( t  x)  A  x  B N  100 i  0  N x  rkfixed   0 10 N  D
 0.5  0.866i   0  

 d2xi    x1  

   D  x 0  
  i
  2 
 d3x 
 i  

i


 x i

Electronics Workbench

58
 1.25

0.75

0.5

0.25

0 1 2 3 4 5 6 7 8 9 10
0.25

0.5

MATLAB

59
1.7.2.Convergence of matrix differential equation to scalar equations

It is evident that solution of first order differential equation with one variable
more simple than several variables equation system. Hence it is interesting to
calculate transient processes by substituting space state matrix by scalar equations.
To change matrix equation into scalar equations system keep in mind that state
matrix A has eigenvectors and eigennumbers  connected to each other by
AΛ   Λ correlation.

dx(t )
In matrix state space  A  x(t )  F substitute the variables changing
dt
x  Λy . As a result we obtain:

dx dy
 Ax  F  Λ  AΛy  F 
dt dt
dy
  Λ 1  A  Λ  y  Λ 1F.
dt

For eigenvectors the equality is true.

 1 0 
 

Q  Λ AΛ 
1 2 ,
 . 
 
 0 n

That is why state space equation is written in the form of independent scalar first
dy
order equation   λ  y  Λ 1F .
dt

As a result we obtain the system of independent differential equations

60
  dy1
 dt  1 y1  b1

dy  dy2   y  b
1
 λ  y  Λ F   dt 2 2 2
.
dt .........................

 dyn   y  b
 dt n n n

At zero initial conditions these equations are solved as follows:

b1 b2 bn
y1 (t )  

1  e  ,
1t
y2 (t )  
2
1  e  , ...,
2t
yn (t )  
n
1  e .
nt

After scalar equations one can go on to the initial variables

 x1   y1 
M    
 x2  y
x  Λy   Λ 2  .
L1  ...   ... 
L2    
R1 R2  xn   y3 
E
i L1 i L2 C
Example 47. Let’s analyze the diagram
Рис. 38 after switching with zero initial conditions
shown in Figure 38. Determine inductance currents and voltage on capacitance for
the following numerical data:

Е  100 V ; С  40 mcF ; R1  70 ;

R2  50 ; L1  0, 2 H ; L2  0,1 H ; M  0,05 H .

Solution.Write down equation system for instant current values and voltages
according to Kirchoff’s laws taking into account the mutual inductance:

d d
L1 iL1 (t )  M iL 2 (t )  iL1 (t )  R1  iL 2 (t )  R1  E ;
dt dt
d d
L2 iL 2 (t )  M iL1 (t )  iL 2 (t )  R2  uC (t )   iL 2 (t )  iL1 (t )   R1  E ;
dt dt

61
 d
C uC (t )  iL 2 (t ).
dt

Using computing unit Given Find, find state space matrix А and independent
influence F.

 1   
 0 0   
 C   0 
 M R1  L2  R1  M  R1  L2  R2  M  R1  M    E  L2 
A , F 2 
  L1  L2  M  L1  L2  M  L1  L2  M 2   
2 2
  L1 L2 M 
 L1  R1  L1  R1  M R2  L1  R1  L1  R1  M   E  M 
   2 
  L1  L2  M  L1  L2  M 2  L1  L2  M 2   
2
 L
 1 2 L M 

With the help of eigenvals( A) function calculate the  eigenvalues of А state

space matrix. Matrix A eigennumbers coincide with characteristic equation roots,
that is why to check the matrix one should make up the formula for input
equivalent circuit resistance (figure 39):

L1 -M L2 -M Z ( p) 
 L1  M  p  R1  Mp  
 L1  M  p  R1  Mp .
M R
E 1
  L2  M  p  R2  0
R1 C Cp
iL1 iL2

Рис. 39 Further procedures and numerical calculations

are given in MathCAD.

5
R1  70 R2  50 L1  0.2 L2  0.1 M  0.05 C  40 10
E  100

State space matrices and independent variables

 0 0 10   0 
 
A   2.857 200 57.15  F   571.5 
 3  285.7 
 11.43 600 1.171  10   

62
 Matrix A eigenvalues

 79.294 
 
  eigenvals ( A )    106.354 
 3
 1.186 10 

Check А matrix

 1186. 
solve p 
z( p) 
 L1  M p  R1  M p
  L2  M p  R2 
1 z( p)   106.4 
float 4
 L1  M p  R1  Mp C p  79.29 
 
Matrix А eigenvectors

 0.9988 0.9975 0.9032 

S  eigenvecs ( A) S   0.0386 0.0563 0.0275 
 
 0.0317 0.0424 0.4284

  79.29  .53e-15 .578e-15 

1
Q  S A S float  4   .84e-17  106.4  .96e-16 
  .1e-16  .1e-15  1186. 
 

Find a new matrix of G => S-1F sources for у variable:

 141.978   uC 
  d
 
1 4
G  S F G   3.155  10  y Q y  G  iL1  S y
 4
dt  
 3.172  10   iL2 

Solution for a new у variable

y1 ( t) 
G0
 1  e  t 
0
y2 ( t) 
( G) 1
 1  e t
1
y3 ( t) 
( G) 2
 1  e t
2
0 1 2

(  79.29 )  t
y1 ( t) float 4  43.44  43.44 e

(  106.4 )  t
y2 ( t) float 4  (  5.066 )  5.066 e

63
 (  1186. )  t
y3 ( t) float 4  (  .5006e-1 )  .5006e-1 e

 y1 ( t) 
 
x( t)  S  2 
y ( t)
Get back to state space variables:  
y3 ( t)
 

(  79.29 )  t (  106.4 )  t (  1186. )  t

x( t) 0 float 4  100.0  397.4 e  297.5 e  .1081 e

(  79.29 )  t (  106.4 )  t (  1186. )  t

x( t) 1 float 4  1.429  15.37 e  16.80 e  .3287e-2 e

(  79.29 )  t (  106.4 )  t (  1186. )  t

x( t) 2 float 4  .893e-18  12.60 e  12.66 e  .5129e-1 e

1
  
min 


3
100
2.4   0.013
80 x  ti 11.8
60 x
x  ti 0 x  ti 21.2
40 x T  5 
0.6
20
0 T  0.063
0 0 0.013 0.025 0.038 0.05 0.063
0 0.013 0.025 0.038 0.05 0.063
ti
ti
N  100

T
i  0  N t  i
i N

Example 48. Determine inductance current iL(t) and voltage on capacity uC(t)
when breaking the circuit at Figure 40 with given elements parameters : Е=50 V;
R=10 Ω; L=0,05 H; С=150 mcF.

64
 Solution. Write down the equations for

R diagram after commutation:

L
E C u L  2 RiL  uC ; iC  CduC / dt.
R R
It follows:
Figure 40.
diL (t ) 1 duC (t ) 1
 (2 RiL  uC );  iL ;
dt L dt C

 2R 1

 L 
L  0
A ; F   .
 1 0   0

 C 

E
Initial conditions: iL (0)   2,5 A, uC (0)  RiL (0)  25 B.
2R

Further calculations are performed in MathCAD.

Given data and initial conditions

6 E E
L  0.05 C  150 10 R  10 E  50 iL0  uC0 
2 R 2

  2 R 1 
L L   400 20 
A   A  - state space matrix
 1   6.667  10
3
0 
 C 0 
 

Determine eigenvalues and eigenvectors:

 200  305.505i 
  eigenvals( A)    
 200  305.505i 

 0.03  0.046i 0.03  0.046i 

  eigenvecs( A)    
 0.999 0.999 

Substitute the variables and obtain the scalar equation:

65
 x  Λ  y  y  Λ 1  y

  1 0   y1   y10  i   y10   12.519  35.514i 

dy         1  L0    
dt  0    y2   y20   uC0   y20   12.519  35.514i 
 2        

T
T  0.05 N  100 i  0  N ti  i
N

 iL   y t  
y1 ( t)  y10 e
 0t
y2 ( t)  y20 e
 1t  i     1 i 
 uC   y2  ti 
  i  

Write down the final calculation and plot diagrams

 (  200.00 )  t (  200.00 )  t 
 y1  t   float 5  2.5000 e cos  305.51 t   3.2732 e sin 305.51 t  

   
 y2  t   complex  (  200.00 )  t (  200.00 )  t
  cos  305.51 t   70.920 e sin 305.51 t  
 25.000 e 

3 50

2 40
30
1
iL uC
i i
20
0 0.013 0.025 0.038 0.05 10
1
0 0.013 0.025 0.038 0.05
2 10

ti ti

1.7. 3. Transient processes calculation with Impulse response function help

Let state space equation of arbitrary circuit has the form of

dx(t )
 A  x(t )  B  e(t ) .
dt

Laplace Transform will make its solution easier and let move up from differential
equations to algebraic. At zero initial conditions we have

66
 IpX( p)  A  X( p)  B  E( p)   Ip  A   X( p)  B  E( p)
X( p)   Ip  A  B  E( p),
1

where I is a unit matrix.

If we take external action as Dirac delta function, than for multidimentional

transfer function we obtain:

X( p )
  Ip  A  B .
1
W( p) 
E( p )

To obtain multidimentional impulse response function h (t ) , one should

perform reverse Laplace transform for transfer function W( p). If at circuit input

is emf source in the form of unit step(скачок)  E ( p )  1 / p  , then after reverse

Laplace transform W( p) / p we obtain impulse response function h(t ) .

MathCAD help perform character direct and inverse Laplace transform.

Example 49. Determine voltage on capacitor and inductance current for

diagram on Figure 41 knowing that initial conditions are zero.. Initial conditions:
Е=30 V; R=20 Ohm; L=0,1 H; C=100 mcF.
R
L Solution.Write Kirchhoff equation and form
E C
R state space matrix and vector of output influence
R

 duC  duC
 C dt  iL  R  uC  RC dt  E
  

 RC duC  u  i R  L diL
 dt
C L
dt

Given

 C u'C  iL R  uC  RC u'C E RC u'C  uC iLR  Li'L

67
  1 3 iLR  uC  E 
  
 2 L 
A iL  uC  E  Find i'L  u'C 
 1 iLR  uC  E 
  
 2 R C 

The impulse response function W( p) in MathCAD can be easily obtained as

shown below when state matrix and vector of output influence are known.

6
E  10 L  0.1 C  100 10 R  10

 3  R 1   1 
2 L 2 L 2 L  250 
A    F    1 eigenvals( A)   
 1 1  1 1   400 
 2 C 2 RC   2  RC 
   

1
W ( p)  ( identity( 2) p  A) F

 5.0000  p  500.

2500.0 
 2 2 
 p  650. p  .10000e6 p  650.00 p  .10000e6 
W ( p) float 5 
 25000. p  150. 
 2  500.  
2
 p  650.00 p  .10000e6 p  650. p  .10000e6 

Now we can obtain analytical dependence for transient functions of

inductance current hL (t ) and capacitor voltage hC (t ) , using inverse Laplace
transform

W ( p) 0 invlaplace p (  400. )  t (  250. )  t

hL( t)   .5000e-1 e  .1000 e  .5000e-1
p float 4

W ( p) 1 invlaplace p (  400. )  t (  250. )  t

hC ( t)   ( 2.500 ) e  2. e  .5000
p float 4

To obtain analytical dependence of inductance current iL (t ) and voltage on

capacitor uC (t ) transient functions should be multiplied by emf Е.

68
 T
iL( t )  h L( t )  E u C( t )  h C( t )  E T  0.1 N  100 i  0  N ti  i
N

0.6 10
0.48 8
0.36
iL ti uC t i
6
0.24 4
0.12
2

0 0.025 0.05 0.075 0.1

0 0.025 0.05 0.075 0.1
ti
ti

If input influence is time function e(t ) , then to determine inductance current

and voltage on capacitor one can apply Duhamel’s integral:

de( ) de( )
t t
iL (t )  hL (t ) E (0)   hL (  t ) d ; uC (t )  h(t )C e(0)   hC (  t ) d .
0
d 0
d

If, for instance, external influence in e(t )  10cos(200t ) example,than in

MathCAD the following should be written:

t

iL( t)   hL t     e    d  hL( t) e ( 0)
d
e ( t)  10 cos ( 200 t)
 d
0

t

uC ( t)   hC  t     e    d  hC ( t) e ( 0)
d
 d
0

0.3 10

0.15 5

iL t i uC t i
0 0.025 0.05 0.075 0.1 0 0.025 0.05 0.075 0.1
0.15 5

0.3 10

ti ti

69
 The example below shows differential equation system solution of DC
motor with independent excitation in MathCAD with state space method
application in Electronics Workbench with structural simulation application.

DC motor with independent excitation

 diа Разрешим  diа Uа 1 ke

Uа  iа Rа  Lа dt  F, F   Су уравнения относительно  dt  L  T iа  L 
 
а а а

M  M  J d , Md   Су производных и запишем  d  1 [k i  M ]

 d с
dt  dt J м а с

ke   Сe , k м   См их в нормальной форме:

Given

Ua ia Ra  La i'a   ke

ia kм M c  J '

  Ua  ia Ra   ke 

 
La
A  ia    M c  Ua  Find i'a  '    
 ia kм  M c 
 
 J 

70
  Ra ke 
 
La La
A  augment ( A ( 1  0  0  0)  A ( 0  1  0  0) ) A    B  A  0 0 M  0  A  0 0  0  U 
c a
 kм 
 0 
 J 
 Ua 
 
La
B 
 Mc 
 
 J 

Given parameters

3
Pн  0.17 Uн  110 Rоя  5.84 La  128 10   47.5 n  750 J  0.004

5
Pн 10  n
Iн  Rдоп  4.40 Ra  Rоя  Rдоп Ra  10.24 Iн  3.254 н  н  78.54
  Uн 30

3
10  Pн Uн  Iн Ra La J Ra
kм  ke  kм  0.665 ke  0.976 Ta  Ta  0.013 Tм 
н Iн н Ra kм  ke

Tм  0.063

Tм
 5.045 T  1. U  200 M c( t)  Iн ke if 0  t  T 0.4 M c( t)  Iн ke
Ta
Iн ke 2 if T 0.4  t  0.8 T

Iн ke 0.5 otherwise

 Ra ke   U 
    L 
La La
 B( t)   a   Iпр 
A  
1
 A  B( T) D( t  x)  A  x  B( t) N  200
 kм   Mc( t)   пр 
     
0
 J   J 

 Iпр   4.775   1562.5   0   0  T  N  D t  x0 i  0  N

  B( 0)    x  rkfixed  
 пр   154.762  794.16965  0  
 

71
 20 200

16 160

12 120

8 80

4 40

0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

Substituition method of matrix equations into scalar equations

 0.33 0.13   58.203

  eigenvecs A       eigenvals A    
 0.944 0.992  21.797

 7.059  103 
b  
1
 B( 0) b  
 3
 y 1( t) 
b
0    t  1
 e
0
y 2( t) 
b
1    t  1
 e
1
0 1
 5.919  10 

72
 (  21.80)  t1 (  58.20)  t1
y 1( t1) float  4  162.0 e  162.0 y 2( t1) float  4  416.7 e  416.7

      1 i 
 ii  y t 
   y 2 ti 
 i  

20 200
160
i 15
i
i 120
  x
1 10   i 80
x
2
i
5 40

0 0.1 0.2 0.3 0.4 0 0.08 0.16 0.24 0.32 0.4

ti ti

1.7.4. Analytical method of transient processes calculation

Let differential equations system for state space vector x   x1 , x2 , x3 ...xn  has

dx
the form of  Ax  B .
dt

The following formula will be the solution of this system:

t
x(t )  x(0)e   e A ( t  )B( ) d
At

where x(0)   x1 (0), x2 (0), x3 (0)... xn (0) is initial conditions vector; eAt is matrix
exponent.

Matrix exponent eAt , argument of which is square matrix At of n order, can be

presented in the form of final row of n components
e At   0 (t )I  1 (t ) Α   2 (t ) A 2  ...   n-1 (t ) A n-1 ,

where k are row coefficients - time domain.

73
 Coefficients of  0 ,1 , 2 ... n1 row can be found by solving , n system of
algebraic equations written as state matrix own numbers А:

 0  11  12 2  ..  1n 1 n 1  e 1t

 t
 0  21  2  2  ..  2  n 1  e 2
2 n 1

 ,
. . . . . . . . . . . .
      2  ..   n1  e nt
 0 n 1 n 2 n n 1

1 1 12 ... 1n 1    0   e t  1

    
1 2 22 ... 2n 1   1   e t 
or matrix form: 
2

.  .
. . . . .   ..   .. 
    
1 n 1  
n n ... n   n1 
2  ent 
  

Equation system solution according the given decomposition coefficients has

the following form:

1
  0  1 1 12 ... 1n1   e t  1

     
 1   1 2 22 ... 2n1   e t  2

.
 ..   . . . . . .  .. 
   
n 1 
 
 ent 

 n1  1 n n ... n 
2
 

After substituting the obtained coefficients into matrix exponent

decomposition we obtain the required solution.

Example 50. Determine capacitance current and voltage on capacitance in

circuit at Figure 42, if elements parameters are given:

E  20 V ; R  10 ; C  50 mcF ; L  0,15 H .

Solution. Write equation system according to

Kirchhoff law and form state space matrix and

R vector of output influence:

R
E C
L
R
74
Рис. 42
 diL
L  iL R  uC ;
dt

diL  du 
L  iL R   iL  C C  R  E;
dt  dt 

 R 1 

 L  0 
L ;
A  F E .
 1 1   
   L 
 C RC 

All the following procedures are given in MathCAD.

Given

L i'L  iL R uC 
L i'L  iL R  iL  C u'C R  E

  iLR  uC 
 
 L 
A  iL  uC  E  Find i'L  u'C  
  iLR  uC  E 
 
 R C 

A  augment( A ( 1  0  0)  A ( 0  1  0) ) B  A ( 0  0  E)

 R 1 
 0 
L L 
A    6
B R  10 E  20 C  50 10 L  0.15
 1 1   E 
 C RC   RC 
 

 R 1 
 66.667 6.667   0 
A  
L L 
A
  F 0 
 1 1  4 3
 F 
 E   4

 2  10 2  10   RC   4  10 
 C RC 
 

75
  E 
 iL0   3 R   iL0   0.667 
Initial conditions        
 uC0   E   uC0   6.667 
     
 3 

matrix А eigenvalues

 138.285 
 
3
  eigenvals( A)
  1.928  10 

Matrix exponent decomposition coefficients

 1 0 
1  e 0t 
 ( t)     
 1  1   1t 
e 

(  138.29)  t (  1928.4)  t
 ( t) 0 float 5  1.0773 e  .77250e-1 e

(  138.29)  t (  1928.4)  t
 ( t) 1 float 5  .55863e-3 e  .55863e-3 e

Matrix exponent

1 0 
I  identity( 2 ) I  Exp( t)   ( t) 0 I   ( t) 1 A
0 1 

 1.04 e(  138. ) t  .400e-1 e(  .193e4 ) t .372e-2 e(  138. ) t  .372e-2 e(  .193e4 )  t 
Exp( t) float 3 
 
 (  138. )  t (  .193e4 )  t (  138. )  t (  .193e4 )  t 
 ( 11.2) e  11.2 e ( .400e-1 ) e  1.04 e 

Analytical solution

  iL0   t
iL( t)  Exp( t) 
    Exp t    F d
  uC0   0
   0 0

  iL0   t
uC ( t)  Exp( t) 
    Exp t    F d
  uC0   1
   1 0

76
 (  138.3 )  t (  1928. )  t
iL ( t) float 4  (  .3588 ) e  .2575e-1 e  1.000

(  138.3 )  t (  1928. )  t
uC ( t) float 4  3.855 e  7.191 e  10.00

T  0.03 t  0  0.01 T  T

1 15
0.8 12
0.6 9
iL( t ) u C( t )
0.4 6
0.2 3

0 0.006 0.012 0.018 0.024 0.03 0 0.006 0.012 0.018 0.024 0.03
t t

1.8. Automatic control linear system stability

When electric power system runs small disturbances always occur. These
perturbations can occur because of power change, load change, short circuit or
breaking of circuit. They make electrical or mechanical system unstable, it means
that steady state mode breaks. In this case automatic system controls switch on,
they perform according to the situation to balance the system. Such disturbances
should not break down system stable performance. Never let disturbances increase.
The system should be stable at small perturbations, i.e. it should keep static
stability.

Static stability is system ability to regain the initial mode after small
disturbances or the mode close to initial mode (if disturbance remains).

To state the system stability it is no need to calculate the roots of characteristic

equation because indirect indicators which show roots real parts planes from which
we state the system stability. These indirect indicators are called stability criteria.

77
 2
z( p )  p  a  p  a
1 2

a  26 a1  5.2 a.1  0
1

a  6975.25 a2  3031.76 a.2  1936

2

100 3

2
50
1

15 10 5 0 5 0 0.05 0.1 0.15 0.2

1
50
2

100 3

Formula for characteristic second order polynomial is the same for all three
transient processes. But coefficients in polynomial are different and of different
planes. In the first case the coefficients are right half plane and the transient
process slows down – the process is stable. In the second case the coefficients are
of different planes and the magnitude grows much, in the third case one of
oscillations grows unrestrictedly - the process is unstable, in the third case one
of the coefficients, and the coefficient in the first power equals zero, hence
oscillations magnitude remains stable – the process is in the margin of oscillatory
stability.

1.8.1.The Routh-Hurwitz stability criterion(algebraic stability criterion)

All the coefficients of the characteristic equation of stable system must have
the same sign (the necessary condition of the Routh-Hurwitz stability criterion). If
78
 one of the coefficients turns zero then the system is unstable. If all the coefficients
are right half plane(RHP) then all the roots(if there are roots on the left half
plane)will be left- half-plane(LHP). Complex roots can be right-half-plane.

If the free component coefficient an  0 turns zero it means there is a zero root , i.e.
the system is in the range of dead-beat stability. If some intermediate coefficient
turns zero it means some imaginary roots appear, i.e. the system is in the range of
pendulum stability.

The necessary condition of stability criterion is the Routh –Hurwitz matrix which
is formed as follows:

For characteristic polynomial

D( p)  a0 p n  a2 pn1  a3 p n2 ......  an

Form the matrix using the following pattern.

 Write diagonally all the coefficients starting from a1 to an .

 Complete the matrix from the bottom to the top in order of index increase
 In places where indices are outside the boundaries put zeros.
 a1 a3 0 0
a a2 a4 0 
D( p)  a0 p 4  a1 p 3  a2 p 2  a3 p  a4  A 0
0 a1 a3 0
 
0 a0 a2 a4 

If all the diagonal minors of matrix are right half plane then the system is stable –
this condition is enough for system stability.

T
a  ( 1 2 3 4 5 )

4 3 2
z( p )  a  p  a  p  a  p  a  p  a
0 1 2 3 4

79
The necessary stability criterion is met(All the coefficients are right half plane)

Computing the coefficients

2

( 1.288)  .8579 i 1

solve  p ( 1.288)  .8579 i
shows   z( p) 
float  4  .2878  1.416 i 
 .2878  1.416 i  2 1 0 1 2
 
1

that the system is unstable

 a1 a
3  0 0
 
 a0 a a 0 
2 4
Form the Routh-Hurwitz matrix A    A  60
0 a a 0
1 3 
0 a a a 
 0 2 4

 a1 0 
a 2 2  a  a  a  a 2  2 3  a  2  a   
1 1 2 0 3 3 1 a a 
 0 4

3  12 4  3 a 4  60 The last minor is left half plane(LHP) so the system is
4

unstable.

Example 1. MathCAD

Example 2.

T 3 2
a  ( 1 2 3 4 ) z( p )  a  p  a  p  a  p  a
0 1 2 3

The necessary criterion is met.

 a1 a3 0 
A   a0 a2 0  A 8 a 2 2  a  a  a  a 3  a  2 3  8
  1 1 2 0 3 3
 0 a1 a3 
 

The necessary Routh-Hurwitz criterion is met thus the system is stable.

80
Computing the roots proves the system to be stable.

solve  p 
1.651 
  z( p )  ( .1747)  1.547 i

float  4  
( .1747)  1.547 i

2 1 0 1 2

Example 3.

T 3 2
a  ( 1 1 4 4 ) z( p )  a  p  a  p  a  p  a
0 1 2 3

The necessary criterion is met (all the coefficients are right half plane)

The first diagonal minor is right half plane, the last but one equals zero

Thus polynomial is in the range of pendulum stability

 a1 a3 0 
solve  p  1. 
A   a0 a2 0  A 0 z( p )   2. i 
  float  4  
 0 a1 a3  ( 2.)  i
 

a 1 2  a  a  a  a 2  0 3  a  2 3  0
1 1 2 0 3 3

Example 4.

T 3 2
a  ( 1 1 4 0 ) z( p )  a  p  a  p  a  p  a
0 1 2 3

81
The necessary criterion is met (all the coefficients are right half plane)

The first diagonal minor is right half plane, the last minor equals zero. Thus the
roots of polynomial is on aperiodic stability boundary.

 a1 a3 0 
solve  p  
0
A   a0 a2 0  A  0   z( p )  ( .5000)  1.937 i

  float  4  
 0 a1 a3  ( .5000)  1.937 i
 

a 1 2  a  a  a  a 2  4 3  a  2 3  0
1 1 2 0 3 3

2 1 0 1 2

Example 5.

T 3 2
a  ( 4 1 4 1 ) Z( p )  a  p  a  p  a  p  a
0 1 2 3

The necessary criterion is met (all the coefficients are right half plane)

The first diagonal minor is right half plane. Thus the roots of polynomial is on the
oscillatory stability boundary

 a1 a3 0 
1 1 0 solve  p .25000

A  identity ( 3) A  a0 a2 0  A  4 4 0   Z( p )   1. i 
    float  5  
 0 a1 a3  0 1 1  ( 1.)  i 
 

82
 3

Im(  )
1 0.5 0 0.5
1

Re(  )

1.8.2. The Routh-Hurwitz algebraic criterion application example

We are given a characteristic polynomial for back pressure system control of

turbine with two amplifying components [5]:

 1 2 p 3   1   2   p 1 2  p 2   1   2   p  1 p   p  1  0
a0 p 3  a1 p 2  a2 p  a3  0

Here 1  T1 / T ,  2  T2 / T - relative time constants  p - inequity degree of

backpressure regulation

Using the Routh-Hurwitz criterion we obtain:

 2  0,   p 2 1 2  1   2    1   2   p   1   2   1 2   0
2

Let’s plot/construct stability borders for turbine backpressure system with

MathCAD help.

     
2 2
f 1 2      1 2 1  2   1  2  1  2  2 1

83
   1
  
2
 
 1   221  2   1  22 1  2  41 2  41 2  
1 2 2 4 2 2 4 3 2 2 3
  2 2     
 2 1 2  12  
f 1 2  solve    
  1

 2 

 1   221  2   1  22 1  2  41 2  41 2   
1 2 2 4 2 2 4 3 2 2 3
     
 2 122  122 
   

 1 
 2 
  1  2 
1 
 
2 2  4 2 2 4 3 2 2 3
 2 2 1  2  1  2 2  1  2  4 1  2  4 1  2 
 2 2   1     ( 1)
2 1  2  1 2
 

30
N  50 i  0  N j  i i 
N
i  2 d
i j  
  i   j

Self regulation influences stability much. As  p grows, stability area decreases and

collapses at  p  0,125 . При  p  0,125 the system is stable at any time

invariables of servomotor.

84
1.8.3. Mikhailov stability criterion

Let’s analyze the characteristic polynomial of n power:

D ( )  a0  n  a1 n 1  a2  n  2  ...an 1  an .

The polynomial can be depicted as :

D (  )  a0    1    2  ...    n  .

Where amplitude of complex function has the form of

| D ( ) || a0    1    2  ...    n  |,
arg  D ( )   arg    1   arg    2   arg    3   ...arg    n 

j n- number of roots,

j  1 j  3 m-polynomial right half plane roots

1 3
j  2
j  4 Let’s put p  j in polynomial
4
2 0

arg D( j))  arg j 1  arg j 2  ...arg j n 

arg D( j)  (n  m)  m  (n  2m)   

j
If all the roots are left half plane m  0 :
j i
arg  D ( j )    n      

At frequency change from zero to infinity obtain:

n
j arg  D( j )   0 
2

Thus Mikhailov criterion for stable system describes a curve (the Mikhailov
hodograph) of characteristic polynomial which starts on the positive part of real
semi-axis, generates an anti-clockwise motion throughТаким n quadrants.

85
Alternative definition of Mikhailov criterion For stable system it is required the
zeros of real and imaginary parts of characteristic polynomial shift, and their
gross number equals n .

Пример. System impulse response function has the form of

1 1
2
p  p2 p 1
K

Fig.

Determine at which К values the system is stable, unstable or is in stability limit.

1 1
W1 ( p)  , W2 ( p )  , W3 ( p)  K
p2  p  2 p 1
K
W ( p)  3
p  2 p2  3 p  K  2

Solution is given in MathCAD

1 1 W 1( p )  W 2( p )  W 3( p )
W 1( p )  W 2( p )  W 3( p )  K W ( p ) 
2 p1 1  W 1( p )  W 2( p )  W 3( p )
p p2

K 3 2
W ( p  K)  D( K  p )  p  2 p  3 p  2  K
3 2
p  2 p  3 p  2  K

U K    Re D K  i   V K    Im D K  i  

U K   complex  ( 2)    2  K
2
V K   complex    3  3
  0 0.01 3

86
 4 System is unstable
3
2 System is in stability
V( 6   )
1 limit
V( 4   )
8 6 4 2 0 2 4 6 8
V( 2   ) 1
2
3 System is stable
4

U( 6   )  U( 4   )  U( 2   )

Mikhailov stability criterion

8 8

4 4
U( 6   ) U( 4   )

V( 6   ) 0 0.5 1 1.5 2 2.5 3 V( 4   ) 0 0.5 1 1.5 2 2.5 3

4 4

8 8

 

System is unstable System is in stability limit

4
U( 2   )

V( 2   ) 0 0.5 1 1.5 2 2.5 3

4

System is stable

1.8.4. Nyquist stability criterion

Nyquist criterion help determine closed loop

WР ( p)
system stability by open loop system stability

87
 WР ( p ) M ( p) / D( p) M ( p)
WЗ ( p )   
1  WР ( p ) 1  M ( p ) / D ( p ) D ( p )  M ( p )

Characteristic polynomial of closed loop system has the form of

 M ( p) 
D( p)  M ( p)  D( p) 1    D ( p ) 1  W  p  
 D ( p ) 

Let’s use Mikhailov’s criterion

 arg  D( j )  M ( j )   arg D( j )   arg 1  WР ( j ) (*)

0  0  0 


If closed loop system is stable then the result should equal n .
2

According to Mikhailov criterion characteristic polynomial argument equals


 arg D( j )     (system may be unstable too) (**)
0  2

Taking into consideration (*) и (**) we obtain the formula for variable argument
1  W ( j ) formula:

  
 arg 1  WР ( j )  n   arg D( j )  n   
0  2 0  2 2

As   n    Н then we finally obtain :

    
 arg 1  WР ( j ) 
0  2
n
2
   
2
n
2
 n    Н      2П  Н 
2

For stable system it is enough that frequency response locus of open loop system
WР ( j ) at frequency change  from 0 up to  covers the point (-1, j) for angle


 2П  Н
2
88
 
 arg 1  WР ( j )   2R  N  if the sum of number of right-2R and
0  2
neutral-N roots is equal zero then the angular is equal 0, i.e. the loop should not
cover the point (-1, j).

Magnitude and phase criterion: System will be stable if locus of open loop system
WР ( j ) does not cover the points
2
jV ( )
with  1, j 0  .
coordinates
h
1
h - minimum section of real axis,
U () which characterizes the distance
2 1  0 1 2 3
between a critical point and
nearest point of root locus с
1
crossing with real axis , is called
modulus stability margins.
2
Minimal angle  formed by

3
the radius passing through the
point of locus intersection with
unit radius circle (with the center at the beginning of coordinates) and negative
part of real axis is called phase stability margin.
Nyquist – Mikhailov stability criteria let state control system stability with
feedback by bode plot locus of closed loop. Criteria can be used in these cases when
system differential equation (or separate components) is unknown but designer
has the corresponding research characterictics.
Let’s analyze some examples with the help of MathCAD
1. Example of stable open loop and stable closed loop systems

W 1( p ) W 2( p )

10 1
W1 ( p )  W2( p )  W p ( p )  W1 ( p )  W2 ( p )
p2 2
p  2 p  4
89
 solve  p  
2.
 

denom W p ( p )  float  5
 ( 1.)  1.7321i
 
Open loop system is stable
( 1.)  1.7321i 


U   Re W p  i   
V   Im W p  i  

10 2
W ( p )    0  .01 10 t  0 .01 2 
3 2
p  4 p  8 p  18

With Trace function help(right click in plot field)determine phase and magnitude
stability margins.

z  
 0  Phase stability margin
  arg z 1
     59.826
 0.5  j 0.86  deg

x0  0.43 Magnitude stability margin 1  x0  0.57

0.5
V(  )

sin( t )
1.5 1 0.5 0 0.5 1 1.5
Im( z )
0.5
0

1.5

U (  )  cos( t )  Re( z )  x0

90
 Determine analytically cutoff frequency – frequency at which locus crosses the
unit circle

2
complex  8
    
V   Im W p j  
simplify
 10  
 2  4 2 
 4        4   16
2
complex  2
    
U   Re W p j  
simplify
 ( 40) 
 2  4 2 
 4        4   16
2
 8
 
V   10  
 4   2     4  4  2  16
  
2
 2
 
U   ( 40) 
 4   2     4  4  2  16
  

 V  2  U  2 1  solve    1.817 
C         ( .9085)  1.574 i
 C  1.817
  0  float  4   0
  ( .9085)  1.574 i
91
2. Example of stable closed loop system and unstable open loop system

W 1( p ) W 2( p )

10 5
W1 ( p )  W2( p )  W p ( p )  W1 ( p )  W2 ( p )
p2 2
p  2 p  4

solve  p 
2. 

denom W p ( p )   ( 1.)  1.732 i


V   Im W p  j    
U   Re W p  j   
float  4  
( 1.)  1.732 i
2
  0  .01 10

Locus covers the point

4 3 2 1 0 1 2 3 4 5 6 7
1
(-1, j0), hence closed
loop system is unstable
V(  ) 2

0
3

U(  )   1

W1( p )  W2( p ) 50
W ( p )  simplify 
1  W1( p )  W2( p ) 3 2
p  4 p  8 p  58

Roots check proves the obtained result

92
 solve  p 
4.831 
denom( W ( p ) )  .417  3.441 i 

float  4  
 .417  3.441 i 

3. Example when open loop system is unstable, closed loop system is stable:
WР ( p )

1.5
T  1 W p ( p )  Open loop system is unstable
T p  1

complex complex
1.500 

U   Re W p  j   simplify 
2

V   Im W p  j   simplify  ( 1.500) 
2
float  4 1.   float  4 1.  

2
  0  .01 10

0.25

2 1.5 1 0.5 0
V(  ) 0.25 Locus starts from U(0) = -1,5 and
0 0.5
moves to zero, hence closed loop
0.75
system is stable.
1

U(  )   1

4. Example, when open loop system has astaticism of the first order closed loop
system is stable:

W 1( p ) W 2( p )

1 5 5
W1( p )  W2( p )  W p ( p )  W1 ( p )  W2 ( p ) W p( p) 
p 2
p  2 p  4 2
p  p  2 p  4 

93
 solve  p   Open loop system is in aperiodic stability
0

denom W p ( p )   ( 1.)  1.732 i

float  4   margin
( 1.)  1.732 i

complex
10.

U   Re W p  j   simplify 
4 2
float  4   4.   16.
complex 2
  4.

V   Im W p  j   simplify  5.
float  4
 4
   4.   16.
2 
2
  0  .01 10

2 1.5 1 0.5 0 0.5 1

V(  )
1 Locus starts with negative imaginary axis
sin( t ) and moves to zero around the point (-1, j0)
2
below, hence closed loop system is stable.
3

U (  )  cos( t )

5. Example when open loop system has astaticism of the first order а closed loop
system is unstable:

4 5
W1( p )  W2( p )  W p ( p )  W1 ( p )  W2 ( p )
p 2
p  2 p  4

94
 solve  p  
0
 

denom W p ( p )  float  5
 ( 1.)  1.7321i
 
( 1.)  1.7321i 

W1( p )  W2( p ) 20
W ( p )  convert  parfrac  p 
1  W1( p )  W2( p ) 3 2
p  2 p  4 p  20

solve  p  
2.9463
  
denom( W ( p ) )
float  5
 .47313  2.5621i
  
U   Re W p  i   
V   Im W p  i  
 .47313  2.5621i 

2
  0  0.01 10

4 3 2 1 0
1
V(  ) 2
0 3

4
5
6

U(  )   1

Example. When open loop system has astaticism of the first order closed loop
system is at stability margin.

W 1( p ) W 2( p )

1 1
W 1( p )  W 2( p )  W p ( p )  W 1( p )  W 2( p )
p 2
p p1

95
 W p  j complex 
1
 
    1
2
i

    1    2  1  2

2  2
2 2

2
1 1 1
U   V   
1   
2
2

2

 1   2 2   2

solve  p  1. 
  0.1 0.1  0.01 10
2
1  W p( p)   1. i 
float  5  
( 1.)  i

V(  )

0 8
1.33 0.67 1.0001 10

U(  )   1

System is at stability margin

Example.

W 1( p ) W 2( p )

2 1
W 1( p )  W 2( p )  W p ( p )  W 1( p )  W 2( p )
p1 2
p p1

2 3
complex ( 2)    1  i   2 i 
 
W p i 
simplify
 2
 1   2     4   2  1
  

96
 2 2
2   1  2
 
U1   ( 2) 
 1   2     4   2  1
 
V1   2  
 1   2     4   2  1
     

 1 
 2 
2
o    2 solve     
2
 1 
 2 
 2 

 1.414 
o     
U1 o 1  0.667 
V1 o 1  0  
h  1  U1 o 1  h  0.333
 1.414 

1  1.2  

arg W p  1 i      19.669
deg

2
  0.1 0.1  0.01 10

solve  p1  
1.8106
1  W p ( p1)  
 ( .9474e-1)  1.2837i
float  5  
( .9474e-1)  1.2837i 

0.5
V1(  )

V1 o1 1.5 1 0.5 0 0.5 1 1.5

sin( t )
0.5
V1( 1)

1.5

U1(  )   1  U1 o1  cos( t )  U1( 1)

Determine modulus and phase stability margins(on diagram)

97
 arg ( 0.94211  i 0.3333)  
 19.483
deg

1.8.5. Logarithmic stability criterion

For the system which is stable being open-loop to be stable when being
closed-loop it is enough at frequency  , for which L( )  0 , difference between
the number of positive and negative phase response crossing  ( ) through the
straight line  ( )   equal zero

Example 1.Investigate closed loop system stability using logarithmic magnitude

phase frequency response if transfer function of open-loop system has the form of:

98
 960
W ( p ) 
3 2
p  26. p  208. p  480.

simplify solve  p  10. 

denom ( W ( p ) )
3 2
 p  26. p  208. p  480 denom( W ( p ) )   4. 
float  3 float  5  
 12. 

Solution. Open-loop system is stable , all the poles are left half plane
960 simplify 3 2
W ( p )  denom ( W ( p ) )  p  26. p  208. p  480
3 2 float  3
p  26. p  208. p  480.

solve  p  10. 
denom( W ( p ) )   4. 
float  5  
 12. 

Determine the real and imaginary parts of magnitude frequency response

complex 2
13.   240.
U   Re W  i    ( 1920.) 
simplify 4 2 6
260.   18304.   230400.  

complex 2
  208.
V   Im W  i   simplify  960. 
4 2 6
float  5 260.   18304.
  .23040e6  

Determine the frequency at which the locus crosses the unit circle – crossover
frequency

 W  i   1 solve  
0     5.1618
  0  float  5
 

Determine the frequency at which locus crosses the real axis

 V 1 0  solve  1  W  0 j 

1     14.422 Z    t  0 .01 2 
 1  0  float  5  W  1 j 

Plot locus

99
 1

0.5

Im( W( i  ) )

sin( t ) 1 0.5 0 0.5 1 1.5 2

Im( Z ) 0.5

1.5

Re( W( i  ) )  cos( t )  Re( Z )

Plot logarithmic magnitude frequency response and logarithmic phase frequency

response in one figure.

L   20 log W  i   2
  0.1 0.1  0.05 10 y  360 360  1  0

1
   if arg  W  i    0  arg  W  i     2  arg  W  i     deg

20
0 1
12
4 60

0.1 41 10 100120 (  )
L(  ) 12
 180
20 180
L( 1)
Phase stability
28  
 0
36 240
margin
44
300
52
60 360
Modulus
  1       0 stability
margin

The graph shows that magnitude frequency response crosses L( )  0 earlier
than phase frequency response crosses the axis   180 O thus the system is
stable.

100
  
 0  180  77.196 -phase stability margin

L( 1)
20
1  10  0.805 - modulus stability margin

Compare to magnitude phase frequency response locus(see the graph above)

Example 2. Investigate the closed loop system stability according to logarithmic

magnitude phase frequency response if the transfer function of open loop system
has the form of:

( .50000)  .86603i 

11  15 p solve  p ( .50000)  .86603i

W ( p )  denom ( W ( p ) ) 
p4  12.p3  22.p2  21.p  10. float  5 

1. 

 10. 

Solution . The open–loop system is stable, all the poles are left half plane.
Determine the real and imaginary parts of magnitude phase frequency response.

complex ( 557.)  2  110.  191. 4

U   Re W  i   
simplify 2 6 8
  100.   100.  

complex 2 4
( 154.)    127.  5. 
V   Im W  i   simplify  ( 3.)  
2 6 8
float  5   100.   100.  

W (p) ( 11.)  15. p

w( p )  simplify  ( 1.) 
1  W ( p) 4 3 2
p  12. p  22. p  6. p  21.

Determine the frequency at which the locus crosses the unit circle –crossover
frequency

 W  i   1  solve  
0    1.2370
   0  float  5
 

Determine the frequency at which the locus crosses the real axis

101
  V 1 0  solve  1
 W  0 j 
1   1  0   .92088 Z    t  0 .01 2 
  float  5  W  1 j  
 U 1  0 

0.5

Im( W( i  ) )

sin( t ) 1.5 1 0.5 0 0.5 1 1.5

Im( Z ) 0.5

1.5

Re( W( i  ) )  cos( t )  Re( Z )

Plot magnitude phase frequency response and phase frequency response in one
figure
L   20 log W  i     0.1 0.1  0.05 10 y  360 360  1  0
1
   if arg W  i    0  arg  W  i     2  arg  W  i     deg
   if    180  5       360   
10
5
60

0.1 1 10120 (  )
5
L(  )
 180
10 180
L( 1)
15  
 0
240
20
300
25
30 360

  1       0

The figure shows that magnitude phase frequency response crosses L ( )  0

later, than phase frequency response crosses the axis   180 O thus the system
is unstable.

Compare magnitude phase frequency response to locus (see figure above)

102
1.8.6. Application of Niquist criterion and transient processes quality criteria

To evaluate the quality of the transient process the following quantity

characteristics are used(Fig 14)

1. Maximum overshoot
xПm
xПm*    xПm*100% .
x 2  x1

2. Setting time.

The time during which the output variable differs from the steady value more than
on five per cent.

Fig 14.
3. Oscillations

Full-wave oscillations number of output value change form regulation time.

tр
M  .
Tр

Automatic control system which is in stability limit has bad transient process
quality parameters. In this case the transient process represents undamped
oscillations and that is why
хПm*  1; t р  ; M   .

103
.It is evident the closer the system to stability margin the worse is transient process
quality. Because of this fact we have indirect parameters of transient process
quality using stability criteria.

Let us analyze such criteria based on Nyquist criterion. Figure 15 shows magnitude
and phase characteristic of open loop system which is stable being closed.
As it is shown that the system can be in stability
margin because of magnitude change or
magnitude and phase characteristic vectors
rotation to the point with coordinates 1; j 0 .
As indirect parameters of transient process
quality we have the variables shown at Figure
15:
"с" –modulus stability margin
"" – phase stability margin
Figure 15

1.8.7. Analysis of controller type and parameters on stability in frequency

domain.

Plotting open loop system magnitude and phase response on complex plane
allows us examine the influence of controller type and parameters on stability
margins.

By analyzing remember that object transfer function is known as well as its

magnitude and phase characteristic.

Cotroller type and its parameters influence can be detected when taking the usual
the magnitude and phase locus vector being open at some k frequency in the
following form

104
 W раз ( jk )  Wоб ( jk )  W р ( jk ) 
 K 
 Wоб ( jk )   K р  K д  jk  и   (51)
 jk 


j 90 K и e  j 90
 Wоб ( jk )  K р  Wоб ( jk )  K д  k  e  Wоб ( jk )  .
k

Formula (50) provides the rule of open loop system magnitude and phase
characteristic visualizing.

Let us analyze special cases

1 P controller(proportional controller) ( K р  0; Kд  0; Kи  0 ).

Wраз ( jk )  Wоб ( jk )  Wр ( jk ) 

 Wоб ( jk )  K р

Wоб ( jk )  0 Ak ; Wраз ( jk )  0 Bk  0 Ak  K р

Figure 16
As shown above in system with P controller Kр gain сauses static accuracy
increasing.At the same time as Figure 16 shows that Kр gain can cause stability
margins decrease, at some big Kр value can break stability.

2.PD controller ( K р  0; Kд  0; Kи  0 ).

105
 W раз ( jk )  Wоб ( jk )  W р ( jk ) 
 Wоб ( jk )   K р  K д  jk  

 Wоб ( jk )  K р  Wоб ( jk )  K д  k  e j 90

Wоб ( jk )  0 Ak ;
Wраз ( jk )  0 Ak  K р  Ck Bk ; .

Ck Bk  0 Ak  Tд  k  e j 90

Figure 17

Derivative component does not influence system properties in steady state

operating conditions, but Figure 17 shows that at it increases stability margins..

3. I controller ( K р  0; Kд  0; Kи  0 ).

W раз ( jk )  Wоб ( jk )  W р ( jk ) 


Kи K и e  j 90
 Wоб ( jk )   Wоб ( jk ) 
jk k

Wоб ( jk )  0 Ak ;
1 
Wраз ( jk )  0 Bk  0 Ak   e j 90
Tи  k

Figure 18

106
In steady state operating conditions in system with I controller static errors don’t
occur . But integral controller makes stability margin decrease as all the vectors
Wраз ( j ) rotate 90º to point side 1; j 0 .

4. PI Controller(proportional integral controller) ( K р  0; Kд  0; Kи  0 ).

W раз ( jk )  Wоб ( jk )  W р ( jk ) 

 K 
 Wоб ( jk )   K р  и  
 jk 

K и e  j 90
 Wоб ( jk )  K р  Wоб ( jk )  .
k
Wоб ( jk )  0 Ak ; W раз ( jk )  0 Ak  Ak Bk ;

e  j 90
Ak Bk  0 Ak 
Tи  k

Figure 19
Integral component of regulation law turns to complete static errors elimination
but on equal terms decreases stability reserves (by using components Bk Ck ////)
Correlation of controller type and its parameters influence both system property in
steady state operating conditions and on stability margins makes it easy to
determine that improving accuracy of the system and increasing stability margin
are controversial

To solve this contradiction is to change structure of the system adding same

subsidiary connections.

PID controller can be used to illustrate this example. Steady-state operating

conditions of such a system are described above and it is seen that static errors
occur in it. Figure 20 illustrates how three components influence magnitude and
phase frequency response position in closed loop state.

107
 W раз ( jk )  Wоб ( jk )  W р ( jk ) 
 K 
 Wоб ( jk )   K р  K д  jk  и  
 jk 
 Wоб ( jk )  K р
Wоб ( jk )  0 Ak ;

 Wоб ( jk )  K д  k  e j 90
W раз ( jk )  0Ck  Ck Dk  Dk Bk ; 
K и e  j 90
 Wоб ( jk )  .
0Ck  0 Ak  K р ; k

e  j 90
Ck Dk  0 Ak  ;
Tи  k

Dk Bk  0 Ak  Tд  k  e j 90

Figure 20

From Figure 20 it’s evident that stability reserves by integral component(vector

Ck Dk ) can be eliminated by simultaneous use of differential component(vector
Dk Bk )

To remember.
1. At amplification coefficient increase in proportional controller transient
process time increase static error decreases, but overregulation increases.
Uncontrolled coefficient’s increase makes the system unstable.
2. Integral controller decreases the static error, but worsens system dynamic
characteristics, i.e. increases transient process time increase, increases
overregulation and regulation time.
3. Differential controller increases system stability, decreases overregulation
and regulation time, i.e. improves system dynamic characteristics, doesn’t
influence the system in steady-state.

108
1.9. Stability areas determination

1.9.1. D-decomposition method

The D-decomposition method is used to investigate the influence of one or two

parameters which are coefficients of system characteristic equation on CS stability.
Let’s analyze D –decomposition method by one parameter.
Write down the characteristic polynomial and change v varying coefficient

P( j )
A( j )  P( j )  vQ( j )  v    U ( )  jV ( )
Q( j )
Let’s plot the locus with axis U ( ), jV ( ) . The locus divides the complex plane
into subareas with the certain range of varying parameter change. Selecting in
subarea the arbitrary value of the investigated parameter v substitute its value into
the characteristic polynomial and find polynomial roots. If n-left roots of n order
characteristic polynomial are in area then this area is called D(n)- area. If n-m of
left roots but m of right roots m then the area is called D(n-m) area.
Analyze the example using MathCAD
D-decomposition method

3 2
D( p  k)  p  p  p  k
2  3
k      i    
2 3
U     V      

D i   k complex    2  k  i 3  

T
k  ( 0.5 i 0.5  0.25 1.25 )

.64780 ( .76205)  .53915 i

solve  p   solve  p 

D pk   ( .17610)  .86072 i


D pk   ( .33671)  1.0658 i

0 float  5   1 float  5  
( .17610)  .86072 i  .98763e-1  .52661 i

109
 solve  p  
1.1118
  

D pk 
2 float  5
 .55901e-1 1.0589i
 
 .55901e-1 1.0589i 

  1.5 1.5  0.001 1.5

0.6

0.4

0.2
V(  )

Im( k) 0 0.25 0.5 0.75 1 1.25 1.5

0.2

0.4

0.6

U(  )  Re( k)

T
ORIGIN  1 T  ( 0.04 0.08 0.25 ) T  0.2 T   0.4

 T p 0 0 0 1 
 
 1 T1 p  1 0 0 0 
 
A p  T   0 1 T p  1 1
2
0 
 
0 0 1 T p 0
 3 
 0 0 0 1 T  p  1 
 


A p  T 
collect  T
 5 3 2 4
 .160e-3 p  .880e-1 p  .49 p  p  .680e-2 p  T  1. 
float  3

1
( p ) 
.160e-3 p 5 3 2
 .880e-1 p  .49 p  p  .680e-2 p
4 

110
 T
  20 20  0.01 20 T  ( 0.05 0.25 i 0.08  0.025 )

 25.614 
 
( 10.227)  7.4396i

solve  p  
T  0.05 A  p  T  0  ( 10.227)  7.4396i

1  1 float  5
 1.7839  5.2278i 
 
 1.7839  5.2278i 

 25.134 
 
( 8.5027)  4.9867i

solve  p  
T  0.25 A  p  T  0  ( 8.5027)  4.9867i

2  2 float  5
( .18014)  3.1945i 
 
( .18014)  3.1945i 

( 25.146)  .37205i 
 
( 11.589)  4.6065i

solve  p  
T  0.025  0.08i A  p  T  0  ( 7.4132)  7.4857i

3  3 float  5
( .77487)  6.0487i 
 
 2.4230  2.7975i 

1.9.2. Root locus analysis

111
 Root locus is defined as characteristic polynomial roots trajectories at one
parameter change of automatic system control from 0 to 
Let characteristic polynomial has the form of:

A( p)  P( p)  Q( p)  0

Where P( p), Q( p) are polynomials of n and m power,  is the investigated

parameter.
Let polynomial A( p ) has the form of:
n m
A( p )    p  zi      p  qi   0
i 1 i 1

At   0 polynomial A( p ) has n roots zi , but at    has only m roots qi .

Root locus will have n branches which start from n roots zi and end in m qi
roots . The rest n-m end in eternity. Asymptotes for branches ending in eternity
come from one point – asymptotes point. Let’s write down the formula of
asymptote center determination.

n m

 z  q
i i
x0  i 1 i 1
nm

Angles between beams of the obtained star are determined by formula:

  2 /  n  m
The nearest angle to real axis equals  / 2 .
Root locus branches separation points from real axis are function extremum
points.

P( p)
 f ( p )  
Q( p)

112
Critical value of the required parameter at which the system is at the stability
margin is determined at branches points of root locus crossing with imaginary axis.
Critical value can be obtained by solving the equation:

A( j )  P( j )  Q( j )  X (,  )  jY ( ,  )  0

Solving system of equations

 X ( ,  )  0
 ,
Y ( ,  )  0

we obtain кр , кр .

Example: Transfer function of open loop system is given as

k
W ( p, k )  , determine at which k values closed loop system
p  p  1 p  3 p  4 

will be stable.
Use MathCAD to solve the problem

Root locus construction

k W ( p  k)
W ( p  k)  W с( p  k) 
p  ( p  1)  ( p  3)  ( p  4) 1  W ( p  k)

k
W с( p  k) simplify 
4 3 2
k  p  8 p  19 p  12 p
Characteristic polynomial of a closed loop system
4 3 2 4 3 2
C( p  k)  k  p  8 p  19 p  12 p P( p )  p  8 p  19 p  12 p Q( p  k)  k

Root calculation at fixed value of system k -parameter


L( k)  polyroots ( k 12 19 8 1 )
T 
 0 
4
The beginning points of root locus branches p  P( p) solve  p   
 3 
 1 
 

 i
100
Roots trajectories calculation in locus N  500 i  0  N x 
i
 i  0.1 Lc  L x
i
N

113
 3
 p
k
k 0
Asymptotes center point calculation x0  x0  2
4

Locus branching points from real axis

solve  p  2. 
    .4188 
d
C p k
dp float  5  
 3.5812

Auxiliary vector of plotting

T
  
i   

i  
 
i   3
 
i   5 
Z  x0 x0  10 e   10 e    x x 10 e    x 10 e    x x 
4 4 4 4
 0 0 0 0 0

Approximate critical value of the required parameter k=xj

 3.99141  1.21066i
3.99141  1.21066i
L x    j  127 x  25.5
j  0.00859  1.21066i j
 0.00859  1.21066i
 

Im Lci 
 0
Im Lci  1
 1
Im Lci 
 2
Im Lci 
 3
5 4 3 2 1 0 1
Im L xj 

Im( Z )
1

Re Lci   Re Lci 1  Re Lci 2  Re Lci 3  Re L xj   Re( Z )
 0

Find critical parameters

C j    k complex  k    19   i ( 8)    12 

 
4 2 3

114
 
k  4  19 2 solve   k   0 0 
     1.2248 26.250
 3  float  5  
( 8)    12    1.2248 26.250

1.9.3. Poles placement method

One of the simplest methods of controller synthesis is poles placement of
closed loop system transfer function which determine its dynamics namely
operation speed and damping of oscillation (see chapter 6.8). The main idea that
you should place these poles at given points of complex plane with the help of
specially chosen controller. This problem can be solved with the help of linear
equations system.
Let transfer function be polynomials relation
N ( p) n p  n0
Wоб ( p )   2 1
D ( p ) p  d1 p  d 0
Let’s choose controller type

N R ( p ) a1 p  a0
R( p)  
DR ( p ) b1 p  b0

where a0 , a1 , b0 , b1 are unknown coefficients to be determined

Characteristic polynomial of closed loop system is equal to
 ( p )  N ( p ) N R ( p )  D ( p ) DR ( p )   n1 p  n0  a1 p  a0    p 2  d1 p  d 0   b1 p  b0 
 b1 p 3   n1a1  d1b1  b0  p 2   n0 a1  n1a0  d1b0  d 0b1  p  n0 a0  d 0b0

Suppose we want to choose the transfer function of controller to place polynomial

roots ( p ) at given points to keep the identity:

( p)  p3   2 p 2  1 p1   0

where  0 , 1 ,  2 are given numbers. Making the coefficient with the same power
of p in the last two equations we obtain:

115
 p 3 : b1  1
p 2 : n1a1  d1b1  b0   2
p : n0 a1  n1a0  d1b0  d 0b1  1
p 0 : n0 a0  d 0b0   0

or matrix form as follows :

0 0 1 0  a0   1 
n 1    
 1 0 d1  a1     2 
 n0 n1 d0 d1  b1   1 
    
0 n0 0 d 0  b0    0 

Equation solution can be written as:

1
 a0   0 0 1 0 1
a  n 0 d1 1   
 1 1  2
 b1   n0 n1 d0 d1   1 
     
 b0   0 n0 0 d0   0 

Square matrix in this formula (it is called Sylvester matrix) needs to be reversible.

116
CHAPTER 2. NONLINEAR DYNAMIC SYSTEMS

2.1. Phase space and dynamic systems phase portraits

Figure 1 Stable system Unstable system

Any electromechanic system is a dynamic system. Elements forming the

system can be nonlinear, hence differential equations describing dynamic systems are
nonlinear.
To investigate nonlinear systems and make the understanding of complex
dynamic taking place in them phase space is used in which one can construct phase
portraits (see Figure 2). Each dynamic system has its phase portrait.
On phase portrait there are special points – equilibrium points, which help
foretell dynamic system behavior without solving differential equations. These
equilibrium points can be stable and unstable. If dynamic system is in the range of
stable equilibrium point, then small disturbances will not make the system unstable
(see Figure 1). If equilibrium point is not stable, then disturbances persist and make
the system ”unstable” (see Figure 1).

117
 equilibrium point of "saddle" type

equilibrium point of "centre" type

equilibrium points of "centre" and "saddle" type,

the line separating solutions –"separatrix"

Figure 2. Dynamic system phase portraits

118
 Let’s analyze phase space method relating to dynamic system of the second
order.

 dx
 dt  F1 ( x, y )
 , (1)
dy
  F ( x, y )
 dt 2

where F1 ( x, y ) and F2 ( x, y) are nonlinear functions of their own arguments.

To get phase trajectory time need to be eliminated. To do this divide the second
equation by the first one

dy F2 ( x, y )
 (2)
dx F1 ( x, y )

At equilibrium points time derivative turns zero, then we obtain the expression
0/0 which has no defined value and is called an indeterminate form:

dy F2 ( x, y ) 0
  (3)
dx F1 ( x, y ) 0

The points at which there are uncertainties are called singular points. In singular
points on phase plane solutions can bifurcate. Let us describe equilibrium state points
search algorithm.
At equilibrium state the given variables x and y do not change, hence
derivatives equal zero:

0  F1 ( x, y )
 (4)
0  F2 ( x, y )

By solving the given nonlinear equations system we find equilibrium state points
x0 , y0 . After determining equilibrium state point coordinates one need to determine
point type. To determine point type expand F1 ( x, y ) and F2 ( x, y) functions in

119
equilibrium state points neighborhood x0 , y0 ,and analyze the first three expansion
terms:

 F1 F1
 F1 ( x, y )  F1 ( x0 , y0 )  x  x  x0   y  y  y0   ...
 (5)
 F ( x, y )  F ( x , y )  F2  x  x   F2  y  y   ...
 2 2 0 0
x
0
y
0

Put down coefficients at linear terms and obtain Jacobian matrix:

 F1 F1 
 x y   a a 
A ( x0 , y0 )      11 12  (6)
 F2 F2   a21 a22 
 x y 
 

Jacobian matrix elements ai , j are constant variables.

Next step is to determine point character .

To do this one should find eigenvalues of Jacobian matrix  by solving characteristic

equation.

I  A  0 . (7)

where I is identity matrix.

It is easy to determine eigenvalues by using MathCAD eigenvalues(A) function
application, several variants are acceptable.
1. Both eigenvalues are real and positive:
1  0, 2  0

Then, equilibrium state point with coordinates x0 , y0 is

Phase curves around called an unstable node
unstable node

120
 2. Both eigenvalues are real and negative:
1  0, 2  0

Then equilibrium state point with coordinates x0 , y0 is

Phase curves around
called a stable node.
stable node

3. Both eigenvalues are real and have different

signs:
1  0, 2  0 или 2  0, 1  0

Then equilibrium state point x0 , y0 is called a saddle or

Phase curves around
saddle point a saddle node.

4. Eigenvalues are complex conjugated numbers

which have positive real parts:
1    j , 2    j ,   0

Then equilibrium state point x0 , y0 is called an unstable

Phase curves around focus
unstable focus

5. Eigenvalues are complex conjugated numbers

which have negative real parts:
1    j , 2    j ,   0

Then equilibrium state point x0 , y0 is called a stable

Phase curves around focus
устойчивого фокуса
6. Eigenvalues are imaginary conjugated numbers:

121
 1  j , 2   j

Then equilibrium state point with coordinates x0 , y0 is

called a centre.
Phase curves around the
centre

Let’s analyze some examples of equilibrium state points determination and determine
their type using MathCAD.
ORIGIN 1

Example 1. Dynamic system is given

d 2 2
x x  y  17
dt

d
y x y  4
dt

Find equilibrium points of dynamic system

2 2
x  y  17 0

x y  4 0

 4 1 
 x  y  17 
2 2
 x 1 4 
z    solve     
 x y  4  y 1 4 
4 
 1 

1 2 1 0

xoz yo  z I   
0 1
1

xoz Expand nonlinear functions in the neighborhood of equilibrium state
points.
2 2
x  y series  x 4  y 1  2  ( 17)  8 x  2 y

 8 2 
a   
 1 4 
122
  8.449  point with coordinates x=-4 и y=1 is a stable node
eigenvals ( a)   
 3.551 

x y series  x 4  y 1  2  4  4 y  x

2
a   I  30  12   

2 2
x  y series  x 1  y 4  2  ( 17)  2 x  8 y

 2 8 
a   
 4 1 

 7.179  point with coordinates x= -1 и y= 4 is called a saddle node.

eigenvals ( a)   
 4.179 

x y series  x 1  y 4  2  4  y  4 x

2
a   I  ( 30)  3   

2 2
x  y series  x 1 y 4  2  ( 17)  2 x  8 y

 2 8 
a   
 4 1 

 7.179  point with coordinates x= 1 и y= -4is called a saddle

eigenvals ( a)   
 4.179 

x y series  x 1 y 4  2  4  y  4 x

2
a   I  ( 30)  3   

2 2
x  y series  x 4 y 1  2  ( 17)  8 x  2 y

x y series  x 4 y 1  2  4  4 y  x

 8 2 
a   
 1 4 

 8.449  point with coordinates x= 4 и y= -1 is called an unstable

eigenvals ( a)   
 3.551 
node .
2
a   I  30  12   

123
Phase portrait of dynamic system

Example 2. Dynamic system is given

d 2 2
x x y 5
dt

d 2 2
y x  y  13
dt

Find equilibrium points of dynamic system

2 2
x y 5 0

2 2
x  y  13 0

 3 2 
 x2  y 2  5 
z    solve   x    3 2 
 2 2  y 3 2 
 x  y  13   3 
 2 

1 2 1 0
xo  z yo  z I   
0 1

Expand nonlinear functions in the neighbourhood of equilibrium state points

2 2
x  y series  x 3 y 2  2  ( 5)  6 x  4 y

2 2
x  y series  x 3 y 2  2  ( 13)  6 x  4 y

124
  6 4  2
a    a   I  48  10   
6 4 

eigenvals(a)  
 5  4.796i point with coordinates x=3 и y=2is called an unstable focus

 5  4.796i
2 2
x  y series  x 3  y 2  2  ( 5)  6 x  4 y

2 2
x  y series  x 3  y 2  2  ( 13)  6 x  4 y

 6 4 
a  
2
 a   I  ( 48)  2   
 6 4 

 8  point with coordinates x=-3 и y=2 is called a saddle.

eigenvals ( a)   
6
2 2
x  y series  x 3 y 2  2  ( 5)  6 x  4 y

2 2
x  y series  x 3 y 2  2  ( 13)  6 x  4 y

6 4  2
a    a   I  ( 48)  2   
 6 4 

 8  point with coordinates x=3 и y=-2 is called a saddle.

eigenvals ( a)   
 6 

2 2
x  y series  x 3  y 2  2  ( 5)  6 x  4 y

2 2
x  y series  x 3  y 2  2  ( 13)  6 x  4 y

 6 4  2
a    a   I  48  10   
 6 4 

 5  4.796i point with coordinates x=3 и y=2 is called a stable focus.

eigenvals ( a)   
 5  4.796i 

Phase portrait of dynamic system.

125
Example 3. Determine equilibrium state point type of a linearized dynamic
system.

d
x x  4 y
dt

d
y 2x  5 y
dt

A  
 1 4  1 Point type x=0, y=0 unstable node
   eigenvals ( A )  
2 5 3

Example 4. Determine equilibrium state point type of a linearized dynamic

system.

d
x x  y
dt

d
y x  3 y
dt

 1 1   2  Point type x=0, y=0 is a stable node.

A      eigenvals ( A )  
 1 3   2 

126
2.1.1. Phase portraits construction with the help of surfaces.

Let’s analyze some more examples of surface construction which help

determine decomposition structure of phase plane into trajectories. First additional
calculations should be done. Put down Newton’s equation of motion with unit mass
on which force F ( x) influences:

d 2x
 F ( x) (8)
dt 2

Taking into account that force equals potential function gradient with negative sign
the equation can be presented as:

dU ( x) d 2x dU ( x)
F ( x)      (9)
dx dt 2 dx

This equation can be rearranged by multiplying each part by velocity v  dx / dt

dv dU dx 1 dv 2 dU d  v2 
v     0   U   0
dt dx dt 2 dt dt dt  2 

From the obtained formula it follows that formula in brackets equals some constant E

v2 x
 U ( x)  E ( x, v) где U ( x )    F ( x ) dx (10)
2 0

Now let’s investigate the dynamic system having the form of:

 dx
 dt  2 y

 dy  4 x  4 x3
 dt

127
Denote velocity in the first equation by 2y , in the second equation denote the right
part by force influencing the particle with unit mass, then auxiliary potential function

U ( x) has the following form :

U ( x )     4 x  4 x 3  dx  2 x 2  x 4
0

In this case constant energy surface expression can be presented as:

U ( x)  2 y 2  E   2x2  x4  2 y 2   E

Now we can plot the phase portrait of constant energy surface

To solve this equation we use MathCAD

d
x 2y
dt
d 3
y 4x  4x
dt

3
F1( x y )  2 y F2( x y )  4 x  4 x

Find equilibrium state points

 x
 F1( x y )  solve    0 0
 y  0 1
z   1. 0  x  z y  z
 F2( x y )  float  5  
 1. 0 

0 0
x  1  y  0
   
 1  0

Find Jacobian
   
F ( x y ) F1( x y )
 x 1 y   0 2

A ( x y )    Ax y  
  F ( x y )   4  12 x2 0 

F2( x y )  
 x 2 y


Determine point type with coordinates 0, 0

128
 0 2
 0 0
A 1  A x  y A1   
4 0
 
  eigenvals A 1


 2.828 
 Saddle point
 2.828 

Determine point type with coordinates 1, 0

 0 2
 1 1
A 2  A x  y A2   
 8 0 
 
  eigenvals A 2


 4i 
 Centre
 4i 

Determine point type with coordinates -1, 0

 0 2
 2 2
A 3  A x  y A3   
 8 0 
 
  eigenvals A 3

 4i 
  Centre
 4i 

Auxiliary potential functions U(x) is written as

x
 3 2 4
U( x)   4 x  4 x dx  ( 2)  x  x

0

x  1.5 1.5  0.01 1.5

0.5

1.5 1 0.5 0 0.5 1 1.5

U ( x)
0.5

Auxiliary energy function E(x,y) is written as

2
E( x y )  y  2  U( x)

Plot phase portrait

3 2
N  251 i  0  N j  i x  1.5  i y  1  j
i N j N

129
Surface E
i j  i j
 E x  y

Surface limited by separatrix E1

i j
 if E i j
 0 E
i j
0 
Surface behind separatrix E2
i j 
 if E
i j
 0 E
i j
0
Separatrix   10
1
Es
i j  i j  0  0
 if E

Direct differential equations solutions are given below.

All curves on phase plane are closed curves hence the solutions will be
periodic.

 2 x 

1
  x0   0
x x0  y 0  rkfixed
   0  T  N  D
3
D( t  x)  N  10 T  10 t  x( 0  0)
4 x  4 x 3
 0  0   y 0 
 



130
 1.25

2
x(  1.5  0)
2
x(  1.3  0)
2
x( 1.3  0)
2
x( 1.15  0) 1.5 0 1.5
2
x(  1.15  0)
2
x(  1.41425  0)

1.25

1 1 1 1 1 1

x(  1.5  0)  x(  1.3  0)  x( 1.3  0)  x( 1.15  0)  x(  1.15  0)  x(  1.41425  0)

2 2

1
1 2
x(  1.5  0) 0 x(  1.5  0) 0

2 2
0 5 10 0 5 10
t t

Solution behind the separatrix

1.42 1

1 0 2
x(  1.41425  0) x(  1.41425  0) 0

1.42 1
0 5 10 0 5 10
t t

Solution on the separatrix

131
 1 0.5

2 1
x(  1.3  0) 0 x(  1.3  0) 1

1 1.5
0 5 10 0 5 10
t t

Solution within the separatrix

0.5 0.8

2 1
x(  1.15  0) 0 x(  1.15  0) 1

0.5 1.2
0 5 10 0 5 10
t t

Solution within the separatrix

The more is the field covered by phase curve the longer is oscillation period

Motion equation of generator rotor for small frequency change can be written as:
d 2
  PT  PГ (11)
dt 2

Rewrite the equation as the system of first order equations .Normalize the presented
equations by  and taking into account that PГ  Pm sin( ) :

d
 PT  Pm sin( )
dt
(12)
d

dt

Find equilibrium state points of dynamic system and plot phase portrait using
MathCAD

132
PT  0.7 Pm  1

d
  F1( x y )  y F2( x y )  0.7  sin ( x)
dt

PT  Pm sin  
d

dt

Find equilibrium state points

solve  
 x
 F1( x y)    x0 
 y  ( .77540 0 )    zT  0.775  0
z 
 F2( x y)  float  5 y 
x    x
1 0
y  y
1 0
x   y 
 0  2.366  0

Find Jacobian
   
F ( x y ) F1( x y )
 x 1 y   0 1
A ( x y )     
A x y   cos x 0 
 
  F ( x y )   
F2( x y )
 x 2 y


Determine point type with coordinates 0.775 , 0

 0 1
 0 0
A 1  A x  y A1  
 0.714 0
  
  eigenvals A 1

 0.845i 
   Centre
 0.845i

Determine point type with coordinates 2.366, 0

 0 1
 1 1
A 2  A x  y A2  
 0.714 0
  
  eigenvals A 2

 0.845  Saddle point

 
 0.845 

Auxilary potential function U(x) is written as :

x

U( x)   ( 0.7  sin ( x) ) dx float  5  ( .70000)  x  1. cos ( x)  1.

0

x  0.1 0.1  0.01   

133
 0.2
0.1
U ( x)
0.4 0 0.4 0.8 1.2 1.6 2 2.4 2.8 3.2
0.06 0.1

0.2

0.3

Energy auxiliary function E(x,y) is written as:

2
E( x y )  y  U( x)

Plot phase portrait

3.2 1.6
N  251 i  0  N j  i x  0.2  i y  0.8  j
i N j N

Surface E
i j  i j
 E x  y

Surface within the separatrix E1

i j
 if E i j
 0.058 E  0.058
i j 
Surface behind the separatrix E2
i j 
 if E
i j
 0.06 E  0.06
i j 
The separatrix   10
3
Es
i j  i j  0.058  0
 if E

 x   x0   0
D( t  x)    x x0  y 0  rkfixed  0  T  N  D
1
 0.7  sin  x0 
3
N  10 T  35    y 0  
t  x( 0  0)
 

Direct differential equations solutions are given below

Closed curves on phase plane are periodic solutions, open curves are unstable
solutions.

134
 1.25

2
x( 3.2   0.9)
2
x( 0.1  0)
2
x( 0.25  0)
2
x( 1.15  0) 0.4 0 0.4 0.8 1.2 1.6 2 2.4 2.8 3.2
2
x( 3.2   0.774)
2
x( 3.2   0.7)

1.25

1 1 1 1 1 1

x( 3.2   0.9)  x( 0.1  0)  x( 0.25  0)  x( 1.15  0)  x( 3.2   0.774)  x( 3.2   0.7)

Unstable motion behind the separatrix

10 16

1 2
x( 3.2   0.9) 7
x( 3.2   0.9)

0 5 10 15 2 0 17.5 35

t
t

Motion on the separatrix

16 4.4

1 2
x( 3.2  0.774) 7 x( 3.2   0.774) 1.8

0 17.5 35
20 17.5 35 0.79

t t

Motion within the separatrix

135
 2 1

1 2
x(0.1 0) 1 x( 0.1  0)
0 17.5 35

1
0 10 20 30
t t

Motion within the separatrix

1.15

1 2
x(1.15 0) 0.8 x( 1.15  0)
0 17.5 35

0.44
0 17.5 35
t t

2.2. Phase plane technique

2.2.1. Servomotor equation with ideal relay characteristic

Let us apply phase plane technique to detect important features of the process
taking place in nonlinear system of turbine rotation frequency regulation by constant
velocity servomotor described by equations [5]:

c-скорость

F ( )

-c

Ideal characteristic of constant velocity

Servomotor servomotor

136
rotor

d
T ; (13)
dt

control element

     ; (14)

constant velocity servomotor

d
 F ( ) . (15)
dt

Where F ( ) :

F ( )  сsign( ) . (16)

As phase plane coordinates( x , y ) choose

d
x  , y  .
dt

Dynamic system will be described by equations system

 dx
 dt  y
 dy F ( ) где F ( )  с  sign(  x  yTx ),    x  yTx (17)
 
 dt Tx

Find equation of servomotor switching line (vertical segment on plot)

   x  yTx  0, y   x / Tx (18)

Find phase portrait lines equation

In area   0 (negative values F in plot)

137
  dx
 dt  y dy c 2cx
 dy    y2    C1 (19)
  c dx Tx y Tx
 dt Tx

In area   0 (positive values F in plot)

 dx
 dt  y dy c 2cx
 dy c    y2   C2 (20)
  dx Tx y Tx
 dt Tx

Write the equation on vertical segment, equaling coefficients in equation of straight

line and in phase plane

dy c x c 1
 , y     yc (21)
dx Tx y Tx Tx y Tx

Find differential equation on segment

dx x dx x
 y , y    x (t )  C3e  t /Tx (22)
dt Tx dt Tx

Sliding process
Let’s analyze the solution with phase portrait plotting with use of MathCAD.

138
 3
F( x)  sign ( x)  1 T1  0.01 N  10

 x
1 
   x0  
D( t  x)   F x  T1 x   T  0.2 x x0  x1  rkfixed
   0  T  N  D

0 1
  x1  
 T1 

1 0
y ( z)  z z  0.05 0.05  0.01 0.1 p  0.5 0.5  0.001 0.5 t  x( 0  0)
T1

0.12 4
1 2
x( 0.02  4) x( 0.02  4)
0.06 2
1 2
x(  0.03   4) x(  0.03   4)
1 2
x(  0.0   4) 0 0.05 0.1 0.15 0.2 x(  0.0   4) 0 0.05 0.1 0.15 0.2
1 2
x(  0.02  4) 0.06 x(  0.02  4) 2

0.12 4

t t

4 1.2

2 3 0.8
 )
x(0.034
2 2 0.4
x( 0.03  4)
1 F(p)
2 0.5 0.25 0 0.25 0.5
x( 0.0  4)
0.4
2 0.15 0.1 0.05 0 0.05 0.1 0.15
x( 0.01 4) 1 0.8
y(z) 2 1.2
3
p
4

1 1 1 1

 )  x( 0.03  4)  x( 0.0  4)  x( 0.01 4)  z
x(0.034

2.2.2. Servomotor equation with actual relay characteristic.

c-скорость Two vertical segments correspond to

c servomotor switching.

b    x  yTx  b  x   yTx  b
F ( )
b    x  yTx  b  x   yTx  b

-c


139
Plot its slopes on phase plane (see MathCAD document below)

In difference from the above ideal example in this case one more area appears

 dx
 dt  y dy
   0  y  const
 dy  0 dx
 dt

Straight lines are parallel to axis x .Now equilibrium state is not one point but the
segment.

y  0,  b  x  b

3
F( x)  sign ( x)  1 b  0.02 F( x)  sign ( x)  if( b  x  b  0  1) T1  0.01 N  10

 x
1 
   x0   0
D( t  x)   F x  T1 x   T  0.23 x x0  x1  rkfixed
   0  T  N  D t  x( 0  0)

0 1
  x1  
 T1 

1 b 1 b
z  0.1 0.1  0.01 0.1 y1( z)  z  y2( z)  z  p  0.05 0.05  0.00025 0.05
T1 T1 T1 T1

0.12 4
1 2
x(0.02 4) x( 0.02 4)
0.06 2
1 2
x( 0.03  4) x(  0.03  4)
1 2
x( 0.0  4) 0 0.058 0.12 0.17 0.23 x(  0.0  4) 0 0.058 0.12 0.17 0.23
1 2
x( 0.02 4) 0.06 x(  0.02 4) 2

0.12 4

t t

140
 4 1.2

2 3
 )
x(0.034 0.8
2 2
x( 0.03  4) 0.4
2 1
x( 0.0  4)
0.05 0.025 0 0.025 0.05
2
x( 0.01 4) 0.15 0.1 0.05 0 0.05 0.1 0.15 0.4
1
y1(z) 0.8
y2(z) 2
1.2
3

1 1 1 1

 )  x( 0.03  4)  x( 0.0  4)  x( 0.01 4)  z
x(0.034

2.3. Lyapunov straight method Прямой метод Ляпунова

Method is based on scalar functions applications having special properties on
dynamic system solutions and being called Lyapunov functions. Lyapunov functions
help estimate system stability and quality and construct control algorithms providing
the required quality process properties.
For system described by the system of differential equations

 dx
 dt  f1 ( x, y )
 (23)
 dy  f ( x, y )
 dt 2

where functions f1, f 2 are arbitrary and possess any kind of nonlinear nature but
always satisfies the condition f1  f 2  0, при x1  x2  0 , as in steady state all
variables deviations and their derivatives equal zero. A certain function of system all
phase coordinates (23) V ( x, y) can be entered where x, y are variables deviations
from some steady values. The function can be presented in two-dimensional space.
Then in each point of phase space V will possess the definite value and at the
beginning of coordinates will equal zero.

The function V is said to be definite sign function in some area if in any point inside
the area function V has definite sign and turns to zero at the beginning of coordinates.
141
Let us analyze the example of definite sign positive function of the second order
system n = 2

V ( x, y )  x 2  y 2 (24)

it is clear that V >0, и V = 0, only at х = y = 0.

Function V is called constant sign function, if it is of the same sign but it turns to
zero not only at the beginning of the coordinates but in other points of the given area.
Function V is called variable sign function, if it in the given area around coordinates
beginning can have different signs.
Arbitrary function V = V(x, y ) which turns to zero only at х = y = 0, and where х, y
are deviations in which the equation of system motion is written is called Laypunov
function. Let’s determine time derivative of function V. Recollect what gradient
operator stands for:

V ( x, y )  
i j  V ( x, y)  gradV ( x, y)
x x

– direction of fast change of function. Now time

derivative of function f ( x, y) can be written down,
considering x, y depend on t :

dV V V dx V dy V V V
V ( x, y )      f1 ( x, y)  f2 ( x, y)
dt t x dt y dt t x y
dV
  V , v  , v   f 1 , f2 
dt

Laypunov theorem about nonlinear systems stability: if for equations of the

system (23) one can choose the stable Lyapunov function V(x, y), such that its time
derivative function dV / dt was also definite sign function (or constant sign function),
but have the sign opposite to sign V, then the given system is stable.
Example: Let nonlinear system be described by equations

142
  dx
 dt   x  x
3


 dy   y 3
 dt
Let us choose Lyapunov definite sign function of type:
V ( x, y )  x 2  y 2

Find time derivative of Laypunov function

dV V dx V dy V V
   f1 ( x, y)  f2 ( x, y)
dt x dt y dt x y
dV
 2x   x  x3   2 y   y3   2  x2  x4  y 4 
dt
Function dV / dt is definite sign function but it is
opposite to function sign V ( x, y) , hence the system is
Laypunov definite sign function
stable.
V ( x, y )  x 2  y 2
2.4. Harmonic linearization method
One of the main methods of higher order nonlinear systems at present time is
approximative method of harmonic linearization. Let’s analyze the example when the
object with linear transfer function is controlled by the block with nonlinear
characteristic

y  F ( x) . (25)

Suppose harmonic signal enters nonlinear component input

x  Asin(t ) (26)

Output signal will be periodic hence it can be decomposed into

y  F ( A sin(t ))  q( A)sin( )  q( A)cos( )  high harmonics

Fourier series

143
 2
1
y  F ( A sin(t )) 
2  F ( A sin( ))d
0
2
1   1 2 
   F ( A sin( ))sin( )d  sin( )    F ( A sin( ))cos( )d  cos( ) 
 0   0 
 high harmonics

There will be no constant component for odd symmetry of nonlinear characteristic

that is why we have

2
1
2  F ( A sin( ))d  0
0
(27)

Then output signal can be written down


x
y  q( A) x  q( A)  high harmonics (28)


Linear part of closed CS вследствие инерционности является фильтром низких

частот, т.е. высокие гармоники проходят ее со значительно большим
ослаблением, чем первая:


x
y  q ( A)a  q( A)  high harmonics
 (29)
 p d
  q ( A)  q( A)  x, p 
  dt

Such a representation is called harmonic linearization of non-linearity and variables

are determined by formulae

2 2
1 1
q ( A) 
A 
0
F ( A sin( ))sin( ) d , q( A) 
A  F ( A sin( ))cos( )d ,
0
(30)

are called harmonic linearization coefficients.

Transfer function of linear component has the form of

144
  p p
y   q ( A)  q( A)  x,  W ( A, p )  q ( A)  q( A) , p  j
  
(31)
j
W ( A, j )  q ( A)  q( A)  q ( A)  q( A) j


Transfer function does not depend on frequency. It increases the magnitude of input
signal and is called complex amplifying coefficient.
Thus nonlinear element can be substituted by linear element.Its frequency
response depend on amplitude of input signal.
Example. Find amplification complex coefficient of given nonlinear component. We
use MathCAD
F   if   0  1  1 b  0.8
F_   if b    b  0  F   nonlinear component characteristic
4
  1  1  10  1

1.5 1.5
1 1
0.5 0.5
F_ ( ) F_ ( sin( t ) )
1 0.5 0 0.5 1 0 2.36 4.71 7.07 9.42
0.5 0.5
1 1
1.5 1.5

 t
Nonlinear element characteristic and output signal

Find amplification coefficient а for different magnitudes, b=0 as function is

even
2 
1 
a( A )   F( A  sin ( t) )  sin ( t) dt
 A 0

Find amplification coefficients and equivalent sinewaves at different magnitudes.

a( 1)  0.764 f1( t)  a( 1)  sin ( t)

a( 1.5)  0.718 f15( t)  a( 1.5)  1.5 sin ( t)

145
a( 3)  0.409 f3( t)  a( 3)  3 sin ( t)

1.5 1.5

F_ ( sin( t ) ) F_( 1.5 sin( t ) )

f1( t ) 0 4.71 9.42 f15( t ) 0 4.71 9.42

1.5 1.5

t t

1.5

F_ ( 3 sin( t ) )

f3( t ) 0 4.71 9.42

1.5

A_  1.5 f( x)  x A_  B x  1.5 1.5  0.01 1.5 f( x)  if( x  0  f( x)  f( x) )

f( x)  if( bb  x  bb  0  f( x) )

1.5 1
1
0.5
0.5
f( x) f( sin( t ) )
1.5 0.75 0 0.75 1.5 0 2.36 4.71 7.07 9.42
0.5
0.5
1
1.5 1

x t
Nonlinear element characteristic and output signal

Find amplification coefficient а for different magnitudes, b=0 as function is

even.

146
 2 
 1
a( A )   F( A  sin ( t) )  sin ( t) dt
 A  0

Find amplification coefficients and equivalent sinewaves at different magnitudes

a( 1)  0.427 f1( t)  0.427 sin ( t) a( 2)  0.936 f2( t)  0.936 2 sin ( t) a( 3)  1.121 f3( t)  1.121 3 sin ( t)

1 2.5

0.5 1.25
f( sin( t ) ) f( 2 sin( t ) )

f1( t ) 0 1.57 3.14 4.71 6.28 f2( t ) 0 2.36 4.71 7.07 9.42
0.5 1.25

2.5
1
t
t

2
f( 3 sin( t ) )

f3( t ) 0 2.36 4.71 7.07 9.42

2

2.5. Self-exciting oscillations analysis algorithm.

With the help of nonlinear element harmonic linearization the closed loop
system(Fig.a) equates to the system with linear equivalent element(Fig
b).Investigation of system with nonlinear element
results in linear system investigation.

Find characteristic equation

a
W ( p )W H ( A)  1, W ( p)  1 / WH ( A),
W ( j)  U ()  jV (), WH ( A)  q( A)  jq( A)

b From the last equation it follows that:

U ( ) q ( A)  q( A)V ( )  1
U ( ) q( A)  q ( A)U ( )  0

147
 (32)
Where q( A) , q( A) are harmonic linearization coefficients. Equation solution
provides us equilibrium state points П , AП .
At solving the problem Goldfarb diagram is easy to use,its algorithm is given below.
Goldfarb diagram
1. Plot the locus W ( j)
V() 2. Plot the locus 1 / WH ( A)
W(j) 3. Find П , AП solving the equation :

A=∞ 2 U() U ( ) q ( A)  q( A)V ( )  1

1
U ( ) q( A)  q ( A)U ( )  0
A=0 -1/WH(A)
4. If at motion along locus 1 / WH ( A) в
сторону увеличения A the point is
covered by locus W ( j) , then magnitude
A will correspond unstable self-excited
oscillations(point 1).If the point is not
covered by locus W ( j) ,the at magnitude
A stable self-excited oscillations occur
(point 2)
Analyze the example with MathCAD use.
Investigate the system on self-exciting oscillations presence
3
W ( p )  U   Re W  i   V   Im W  i  
3 2
p  3 p  p
complex complex 2
9. 3.   1.
U   U  simplify  V   V  simplify  
2 4  2 4
float  5 7.     1. float  5 7.     1.

148
 4
q ( a) 
 a

Harmonic linearization coefficient

of nonlinear component

Find crossover point, frequency and magnitude

W H( a)  q ( a)

 U   q ( a) 1  solve   


w   V   q ( a) 0   a  ( 1. 1.2732)
  float  5
 0 

 0 a0   w 0  1 a0  1.273

1
  0  .01 10 M ( a)  U( 1)  1 a  0  0.1 2.3
W H( a)

The obtained values of

0.2 parameters 0  1 and a0  1.273
0.1 correspond to stable self-excited
V(  ) oscillations because by point
Im( M ( a) ) 2 1.5 1 0.5 0
moving across the locus in the
direction of increasing
  
Im M a0 0.1
magnitude – a, locus does not
0.2 cover the point

0.3

  
U(  )  Re( M ( a) )  Re M a0

149
Literature
1. Iy.I.Yurevich. Automatic Control Theory. StPb.:BHV- St. Petersburg 2007.
560p.
2. Rao V.Dukkipaty. Control System. Alpha Science International Ltd.
Harrow,U.K. 2005. 698 p.
3. Е.А.Nikulin. Backgrounds of Automatic Control Theory. StPb.:BHV- St.
Petersburg 2007. 640p.

150