Transcript Materialy/06/Lecture1- ICM Introduction

Slovak University of Technology
Faculty of Material Science and Technology in Trnava
Intelligent Control
Methods
Lecture 1:
Introduction. Reasons
for ICM, Basic Concepts
Classic Control Theory:
h(t) +
-
e(t)
w(t) +
u(t)
C(s)
+
v(t) +
S(s)
y(t)
+

C(s)
Image transmission of the controller

S(s)
Image transmission of the controlled
system
2
Classic Control Theory:

Classic (proper):
 60
years old
 Based on the system external description (relation
input – output)
 Continuous systems: Differential equations (linear,
non-linear) ⇨Image transmission
 Examples: RC-unit, liquid level
du2 (t )
RC
 u2 (t )  u1 (t )
dt
U 2 ( s)
1
S ( s) 

U1 ( s) 1  RCs
3
Classic Control Theory:

Modern:
 30
years old
 Based on the system internal description (relation
input – state – output)
x´(t) = A x(t) + B u(t)
y(t) = C x(t) + D u(t)
4
Image Transmission:

Relation between the image of the output
parameter and the input parameter by zero start
conditions
U 2 ( s)
S (s) 
U1 ( s)

X ( s)  L( x(t ))   x(t )e dt
 st
0

Vocabulary for Laplace-transformation available!
5
Image Transmission Estimation:

From differential equations (if available)

As result of the system identification according
to system standardized signals response



Dirac impulse (impulse characteristic)
Unit-pulse signal (transmission characteristic)
Methods for image transmission estimation from
impulse or transmission characteristic available!
6
Control Loop with Negative Feed-Back:
h(t) +
-
e(t)
w(t) +
u(t)
C(s)
+
v(t) +
S(s)
y(t)
+
Image transmission of the controller

Image transmission of the controlled
system
Transmission of control circuit with NF-B:

C(s)
S(s)
Y ( s)
R( s) S ( s)

H ( s) 1  R( s) S ( s)
7
Control Design in Classic Control Theory:

Starting point: Mathematical model
(transmission) S(s) of the system

Design of the controllers with the transmission
C(s) so as the closed control loop has desired
properties
 Feed-back
control quality (output time behavior
should be similar to the desired one)
 Stability of the controlled system
8
Lectures in Classic Control Theory:







Systems, approaches to description (first of all linear
dynamic systems)
System response to normalized input signals, system
behavior appreciation according to response
System stability, determination and criteria
Transmission algebra (global transmission of more
connected systems)
Feed-back control loop
Controllers synthesis, PID-controllers
Feed-back control quality, kriteria
9
Problems of the Classic Control Theory (1):


Mathematical model needed (input-output, inputstate-output)
Model complicated or unsolvable
 Non-linear
 Too

many parameters
Time behavior of systems (models too) varies
(parts mature, pipe-lines foul, supplies falter...)
10
Problems of the Classic Control Theory (2):




Not only deterministic but also stochastic system
behavior
Not all inputs controllable
Control signals have physical restrictions
(valves, supplies, ...)
Time delay (algebraic => exponential equations)
11
From above mentioned results:


Classic control theory supplies and
complements with intelligent methods (soft
computing)
ICT are used (programming languages and environments,
simulation, industrial programmable controllers, AI, NN, GA, fuzzy
sets, ...)

The goal: to create systems
 intelligent
 optimal
 adaptive
 robust
12
It means:

To simplify the mathematical model, or its
replacement with description:
 Linguistic
description (fuzzy sets)
 Modeling and simulation (simulating tools and
environments, NN, ...)

To react on system time behavior changes
 Adaptive


methods
To handle the uncertainty in system behavior
(Bayes probability, fuzzy approach)
To master symbolic (non-numerical) information
13
The mentioned properties allow to be used
in all control levels:
EIS
DSS
MIS
PID-level
Top-level control
(Executive IS, ES, DSS)
MIS, production processes
(systems) control
PID-level (technological
processes control)
14
ICM-lectures structure:
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
Introduction. Classic and modern CT, direction to ICM.
Artificial intelligence.
Problems solution in artificial intelligence systems (resolution
method, state space)
Production systems. Rules chaining as solution method.
Expert systems.
Knowledge base design. Knowledge acquisition in databases.
Uncertainty. Bayes´s and fuzzy approach.
Fuzzy systems, fuzzy control.
Genetic algorithms.
GA in optimizing, control and regulation.
Neuronal nets (NN).
Process modeling with NN.
15
Literature:
1.
2.
3.
4.
Nillson, N.J.: Principles of Artificial Intelligence.
Addison-Wesley, London, New York, 1991.
Man, K.F., Tang, K.S., Kwong, S., Halang, W.A.:
Genetic Algorithms: Concepts and Designs.
Springer Verlag, London 1999.
Karr, C.L., Freeman, L.M.: Industrial Applications of
Genetic Algorithms. Boca Raton, London, New
York, Washington D.C., 1999.
ATP Journal plus 7/2005: Artificial Intelligence in
Practise. Automation, Robotics, Mechatronics.
Advanced Control Techniques, Discrete
Manufacturing Systems.
16