Fuzzy Logic Systems

Module 2

Fuzzy Control System Design

Module 2 Objectives Fuzzy Control Systems Design

• To do a brief review on basics of control.

• To understand the need for fuzzy logic control.

• To understand how to design a fuzzy logic controller.

• Understanding the components of a fuzzy logic controller.

• To understand how inferencing is done in a fuzzy logic controller.

• To be able to understand how rules can be developed in a fuzzy

logic controller.

• To study several types of defuzzification techniques.

• To study the design procedure of a fuzzy logic controller.

• At the end of the module, the student should be able to understand

the basic components of a fuzzy logic controller and how they
2
function.
 Module 2 Contents
• 2.1 Overview of fuzzy logic control
– 2.1.1 Review of Control System Basics
– 2.1.2 Why fuzzy logic control?

• 2.2 Designing a fuzzy logic control system (FLCS)

– 2.2.1 Components of the fuzzy controller
– 2.2.2 Operations of the fuzzy controller
– 2.2.2 Fuzzification
– 2.2.3 Knowledge base
– 2.2.4 Inference Mechanism
– 2.2.5 Defuzzification

• 2.3 Design procedure of the fuzzy logic controller

• 2.4 Summary of Module 2
3

2.1 Overview of Fuzzy Logic Control

2.1.1 Review of Control System Basics

2.1.2 Why fuzzy logic control?

2.1.3 Configuration of a fuzzy logic control

system

4
 2.1.1 Review of Control System Basics
Example of control systems

– Cars, Aeroplanes, Trains, Ships, ...

– Nuclear Plants, Power Stations, ...
– Washing Machines, Refrigerators, ...
– Disk Drives, Stepper Motors, .......
– Chemical Processes, Food Processing Plants .....
– Robotic Manipulators, NC Machines, …

Example of Control Variables

• Position • Pressure
• Speed • Turgidity
• Liquid level • Concentration
• Intensity
• Temperature • Etc.
5

A Feedback (Closed-Loop) Control System

Error signal Control signal

Setpoint +
or Controller Process Output
-

Summer Feedback
(Op-Amp)

A Conventional Control Block Diagram

Control objective.
• In many control systems, the objective is to design a controller such
that output can be controlled by giving the desired signals at the
input.

Example of control.
– Alignment of the front wheel of a vehicle follows that of the
driver’s steering wheel.
– Rudders of a boat follow its steering wheel.
– Concentration in liquids follows a given set-point.
– Temperature of furnace follows a given set-point. 6
 Control Systems Performance
• What are the factors affecting control systems?
– Choice of the Controllers (Algorithms)
– Tuning of Controllers
– Selection of Control Variables
– Selection of Performance Index
– Types of Sensors (Measurement Errors)
– Actuators
– Disturbances/Environment
– Noise
– Control Configuration
– Design Techniques
– Nonlinearity
7

• Several performance indices that can be used to

measure control system performance.

– Integral Squared Error (ISE).

– Integral Absolute Error (IAE).
– Integral Time Absolute Error (ITAE).
– Integral Time Squared Error (ITSE).
– Minimum Fuel, Time, etc.

8
Control system components

• A control system is made up of many sub-control

systems and components.

A military helicopter

A space shuttle

Analysis in Time Domain

• By observing the system time response we can measure its
performance.
• To measure the swiftness of system response: Check rise-
time and peak-time.
• To measure the closeness of system response to the desired
response: check overshoot, settling-time, rise-time,
steady-state error, ISE, etc.
• Other analysis can also be done.

10
 Example of Control Paradigms
Control Design Techniques and Control Paradigms

Modern
Classical

Artificial Intelligence
State Feedback
Root Locus State Estimation
Bode Plot Observers
Nichols Chart Optimal Control
Nyquist Plots Robust Control
H-Infinity
Internal Model Control
Adaptive Control
Neuro-Control
Fuzzy Control
Genetic Algorithms
Knowledge Based Systems
11

Example of Controllers
That Can Be Applied in Control Systems
Conventional Controllers Predictive Controllers
• Bang-Bang (on-off) • Smith Predictor
• Proportional • Generalized Predictive
Controllers
• PID Controller

Intelligent Controllers
Adaptive Controllers
• Knowledge Based
• Self-Tuning Controller
• Fuzzy Logic Controllers
• Self-Tuning PID
Controller • Neuro-Controllers
• Auto-Tuning PID • Adaptive Fuzzy
Controller
• Pole Placement 12
 Overview of Fuzzy Logic Control

2.1.2 Why Fuzzy Control?

• During the past decade, fuzzy logic control has emerged as one of
the most active and fruitful research areas in the application of
fuzzy set theory, fuzzy logic and fuzzy reasoning.

• The idea was first proposed by Mamdani and Assilian around 1972.

• Many industrial and consumer products using fuzzy logic

technology have been built and successfully sold worldwide.

• In contrast to conventional control techniques, fuzzy logic control is

best utilized in complex ill-defined processes that can be controlled
by a skilled human operator without much knowledge of their
underlying dynamics.
13

• The basic idea behind fuzzy logic control is to incorporate the

“expert experience” of a human operator in the design of a
controller in controlling a process whose input-output
relationship is described by a collection of fuzzy control rules
(e.g. IF-THEN rules) involving linguistic variables.

• This utilization of linguistic variables, fuzzy control rules, and

approximate reasoning provides a means to incorporate
human expert experience in designing the controller.

14
Benefits of Fuzzy Control
• Suitable for Complex Ill-defined Systems where
mathematical modeling is difficult or Plant is too
abstract e.g. Complex chemical plant, cement-kiln,
etc.

• Able to design along linguistic lines – usage of

rules based on experience

• Better performance than Conventional PID

Controllers

• Simple to design
15

Overview of Fuzzy Logic Control

2.2 Designing a Fuzzy

Logic Control System

 Where do we start?
 What are the Basics/Background
needed?
 What are the components of a Fuzzy
Controller?
 Where can it be applied?
16
 A fuzzy logic control system

Error Control Signal

Setpoint or Input Transducer Fuzzy Controller Process Output or

Reference + Control led Va riabl e

Def uzz
Fu zz
R u le s
-

Rate of Change
of Error Output Transducer
or Sensor

17

Components of the Fuzzy Logic Controller

2.2.1 Components of the fuzzy logic controller

• It generally comprises four principal components:
• fuzzifier
• knowledge base
• inference engine
• defuzzifier

• If the output from the defuzzifier is not a control action for

a process, then the system is a fuzzy logic decision support
system.

• The fuzzy controller itself is normally a two-input and a

single-output component. It is usually a MISO system. 18
 Components of the Fuzzy Logic
Controller
Fuzzy Controller

Knowledge
Base

Fuzzy Fuzzy

Inference
Fuzzifier Defuzzifier
Engine

Process Output Crisp Control

& State Signal
Process

19

Design Process
 Identify controller Inputs and Outputs as the Fuzzy
variables

 Break up Inputs and Outputs into several Fuzzy Sets

and label them according to the problem to be solved
and set up the fuzzy variables on the appropriate
universes of discourse <FUZZIFICATION>

 Configure/develop RULES to solve the problem

 Choose Inference Encoding procedure

 Choose a DEFUZZIFICATION Strategy

 Tune the adjustable parameters 20

Discuss how to design a fuzzy
controller for the following process?
• Hot shower system
• Automatic Fuzzy Car Wipers
• Rice Cooker
• Vacuum Cleaner
• Air conditioner
• Water Bath Temperature Control System

21

Components of the Fuzzy Logic Controller

2.2.2 Operations of the fuzzy controller

• A fuzzy logic controller usually has 2

inputs and 1 output for a SISO plant.

• The 2 inputs are used for accessing the

current conditions of the process such
that the necessary action can be taken by
providing the correct control signal (which
is its output).
22
• The inputs to the fuzzy controller is usually
the error (e) which measures the system
performance and the rate at which the error
changes (De). and the output is the change
of the control signal (Du).

Inference Mechanism
Sensor
e readings
Control

Defuzzification
signal
Fuzzi fication

Rule base
Du
De IFXis AAND Yis B
THEN Zis C

23

• A fuzzifier then transforms the crisp values of e and

De into corresponding fuzzy values (usually there are
several fuzzy values of e and De).

• From the rule base, the fuzzy values of e and De

determine which rules are to be fired through an
inferencing algorithm.

• Several values of Du will then be obtained and a

defuzzification mechanism will then transform these
values into one crisp value.

• The actual control signal is obtained by adding Du to

the past value of u which is send to the plant.
24
 Components of the Fuzzy Logic Controller

2.2.2 Fuzzification

What is
Fuzzification?

25

Fuzzification
• Involves the conversion of the input/output
signals into a number of fuzzy represented
values (fuzzy sets).
• Choose an appropriate Membership Function to
represent each Fuzzy Set.
• Label the Fuzzy Sets appropriately such that they
reflect the problem to be solved.
• Set up the Fuzzy Sets on appropriate Universes
of Discourse.
• Adjust / tune the widths and centerpoints of
membership functions judiciously.
26
Consider a Fuzzy Rice Cooker
• First, we need to identify the input and
output variables of the process, in this
case, the rice cooker.
• Study the control mechanism.

27

Input and Output Variables

of the Fuzzy Rice Cooker
What is the output of
the Fuzzy Controller?

Rate of Heating, DE

2 Best Candidates of
the Input Variables (in
order to make the
output decision) would
be:
Amount of Temp. of
Rice, W Rice, q

28
Fuzzification
Rate of • For each Variable, how
Heating, DE many fuzzy sets should each
be partitioned into?

• For simple problems, we can

Amount of start with 3, usually 5.
Rice, W
• What would be the correct
labels for each fuzzy set?

Temp. of • Use appropriate

Rice, q membership functions,
usually triangular.

29

Fuzzification

Membership
Value, mDE MEDIUM
LOW HIGH
1.0

Rate of
Heating, DE

30
Membership
Value, mW MEDIUM
LOW HIGH
1.0

Amount of
Rice, W

Membership
Value, mq MEDIUM
LOW HIGH
1.0

Temp. of
Rice, q
31

Example of Standard Labels used

for Quantization into 5 Fuzzy Sets
(usually for Control systems)

Negative Medium NM
Negative Small NS
Zero ZE
Positive Small PS
Positive Medium PM

32
 Consider an Air Conditioner

How do you design the fuzzification

component of the fuzzy Air-
conditioner system?

33

SOME ANALYSIS ON FUZZY SETS IN CONTROL

• Usually fuzzy sets are overlapped by about 25%.

• It can be observed that in fuzzy logic control the transition

from one fuzzy set to another provides a smooth transition
from one control action to another.

• If there is no overlapping in the fuzzy sets then the control

action would resemble bivalent control (transition from one
error region to another is rather abrupt).

• On the other hand if there is too much overlap in the fuzzy

sets, there would be a lot of fuzziness and this blurs the
distinction in the control action.
34
• For simplicity we can choose the same (triangular)membership
functions for each of the fuzzy subsets of all the fuzzy variables.

• The example below shows membership functions chosen for 2

fuzzy input variables and 1 output variable in controlling an
inverted pendulum.

mq
NM NS ZE PS PM
1

0
0 Pendulum Angle,q
mDq
NM NS ZE PS PM
1

0
0 Angular Velocity,Dq
mv
NM NS ZE PS PM
1

0
0 Motor Voltage, v

Example of Fuzzification for an Inverted Pendulum Control 35

Components of the Fuzzy Logic Controller

2.2.3 Knowledge Base

• The knowledge base of a fuzzy logic controller consists of a data base
and a rule base.

• The basic function of the data base is to provide the necessary

information for the proper functioning of the fuzzification module, the rule
base and the defuzzification module.

• This information includes

– the meaning of the linguistic values of the membership functions
of the process state and the control output variables.
– physical domains and their normalized counterparts together with
the normalization, denormalization and scaling factors.
– the type of the membership function of a fuzzy set.
36
– the quantization look-up tables defining the discretization policy.
 Fuzzy Control Rules
• The basic function of the rule base is to represent the expert
knowledge in a form of if-then rule structure.
• See C.C. Lee, "Fuzzy Logic Control: Part I", IEEE Transactions
on System, Man and Cybernetics, 1990.

• These four methods of deriving the rule base can be described

as follows:
(i) Expert experience and control.
engineering knowledge.
(ii) Based on operator’s control actions.
(iii) Based on fuzzy model of a process.
(iv) Based on learning.

37

• In this course we will be study how to derive the fuzzy rules

based on method (i).
(i) Expert experience and control engineering knowledge.
• This method is the least structured of the four methods and yet it
is one of the most widely used today.
• It is based on the derivation of rules from the experience based
knowledge of the process operator and/or control engineer.

(ii) Based on operator’s control actions.

• This method tries to model an operator’s skilled actions or
control behavior in terms of fuzzy implications using the input-
output data connected with his control actions.
• The idea behind this technique is that it is easier to model an
operator’s actions than to model a process, since the input
variables of the model are likely found by asking the operator
the kind of information he uses in his control actions.
38
(iii) Based on fuzzy model of a process.
• In the linguistic approach, the linguistic description of the
dynamic characteristics of a controlled process may be viewed
as a fuzzy model of the process.

• Based on the fuzzy model, we can generate a set of fuzzy

control rules for attaining optimal performance of a dynamic
system.

• The set of fuzzy control rules forms the rule base of the fuzzy
logic controller.

• Although this approach is somewhat more complicated, it yields

better performance and reliability and provides a more tractable
structure for dealing theoretically with the fuzzy logic controller.

39

(iv) Based on learning.

• Many fuzzy logic controllers have been built to emulate human

decision making behavior.

• Currently, many research efforts are focused on emulating

human learning, mainly on the ability to create fuzzy control
rules and to modify them based on experience.

• There are now many examples of fuzzy controllers which have

the capability to learn and to compose the rules involving neural
networks and genetic algorithm.

• One example is given in the research at CAIRO on Self-

Organizing Neuro-fuzzy Control Systems.

40
 Example of Fuzzy Control Rules

Example of rules for traffic control

IF traffic from the north of the city is HEAVY

AND the traffic from the west is LESS
THEN allow movement of traffic from the north LONGER.

IF traffic from the north of the city is AVERAGE

AND the traffic from the west is AVERAGE
THEN allow NORMAL movement of traffic for both sides.

41

Example of a Rule for a

Temperature Control System
If the error in the temperature is positive and large.
and the rate of change of error is almost zero.
then heater should be on positive and large.

This rule can be shorten as follows:

IF e is PL AND De is ZE THEN Du is PL.

or (PL, ZE; PL).

42
 Example 2.1
• For simplification a matrix can be used for
presenting the rules which is called fuzzy rule
matrix.
• Many control rules are written in the form below:
( Antecedents) (Consequent).

IF X AND Y THEN Z.

• Example of a fuzzy rule matrix for controlling a

temperature control process given as follows.

43

• For the rule (P, Z; P)

• For the rule (N, Z; N)
• For the rule (Z, N; N)
Antecedents
e P
N Z
De
Antecedents

N N N P

Z N Z P
P N P P
Steady State Rule

Consequents
44
 Rules?

•How do we
formulate the
rules??
45

Derivation of fuzzy rules

based on observations
• Though fuzzy rules can be formulated through
observation of the process, they are, however, could
not be easily derived if one has not much experience.

• A more systematic approach to formulate the fuzzy

control rules can be done following the approach
given by C.C. Lee.

• The fuzzy rules can be formulated based on the

conditions of error, E, and the rate of change of error,
DE, and the output is the change in the control signal,
Du.
46
• Mathematically, we can write:
for the error at the sampling instant k:
e(k )  r (k )  y(k )
de(t )
• For the derivative of the error in continuous time:
dt
• and in discrete time at sampling instant k:

De(k )  e(k )  e(k  1)

Inference Mechanism
Sensor
e readings
Control
signal

Defuzzification
Fuzzification

Rule base

Du
De
IF X is A AND Y is B
THEN Z is C

47

• The prototype of fuzzy control rules is tabulated in Table 1 and a

justification of fuzzy control rules is added in Table 2 (These are
for only 3 quantizations P, Z and N).
• As an example, based on the underdamped 2nd order system
response, the corresponding rule of Region 2 can be formulated as
rule R1 below which decreases the system overshoot and for
Region 3, Rule R2 has the effect of shortening the rise time.

• More specifically,
R1: if (E is negative and DE is positive )
then DU is positive,
R2: if (E is negative and DE is negative)
then DU is negative.

• Tables 3 and 4 provide the selection of the consequents for 7

quantizations of the fuzzy control variables (i.e. NB, NM, NS, ZE,
PS, PM, PB).

48
 Observation of system response
for deriving fuzzy control rules

e(k)=r(k)-y(k)
De(k )  e(k )  e(k  1)
e- e-
De+ De- e-
Setpoint De+

e+
De+

e+ e+ e- e+ e+ e- e- e+
De- De De- De- De+ De+ De- De-
+
a1
1 2 3 4 5 6 7 8 9 10 11 12
49

TABLE I
PROTOTYPE OF FUZZY CONTROL RULES WITH TERM SETS
(NEGATIVE, ZERO, POSITIVE)

Rule No. E DE DU Reference Point

1 P Z P a1, a2, a3
2 Z N N d1, d2, d3
3 N Z N c1, c2, c3
4 Z P P b1, b2, b3
5 Z Z Z set point

TABLE 2
PROTOTYPE OF FUZZY CONTROL RULES WITH TERM SETS
(NEGATIVE, ZERO, POSITIVE)

Rule No. E DE DU Reference Point

6 P N P 1 (rise time), 5
7 N N N 2 (overshoot), 6
8 N P N 3, 7
9 P P P 4, 8
10 P N Z 9
11 N P Z 10
50
 TABLE 3
PROTOTYPE OF FUZZY CONTROL RULES WITH TERM SETS
{NB, NM, NS, ZE, PS, PM, PB}

Rule No. E DE DU Reference Point

1 PB ZE PB a1
2 PM ZE PM a2
3 PS ZE PS a3
4 ZE NB NB b1
5 ZE NM NM b2
6 ZE NS NS b3
7 NB ZE NB c1
8 NM ZE NM c2
9 NS ZE NS c3
10 ZE PB PB d1
11 ZE PM PM d2
12 ZE PS PS d3
13 ZE ZE ZE set point

51

TABLE 4
PROTOTYPE OF FUZZY CONTROL RULES WITH TERM SETS
{NB, NM, NS, ZE, PS, PM, PM}

Rule No. E DE DU Reference Point

14 PB NS PM 1 (rise time)
15 PS NB NM 2 (overshoot)
16 NB PS NM 3
17 ND PB PM 3
18 PS NS ZE 9
19 NS PS ZE 10

52
 2.2.4 The Inference Mechanism

What is an Inference
Procedure in FLCS?
Fuzzy Controller

Knowledge
Base

Fuzzy Fuzzy

Inference
Fuzzifier Defuzzifier
Engine

Process Output Crisp Control

& State Signal
Process
53

Some Notes on the Inference Mechanism

• The Inference Mechanism provides the mechanism for invoking or
referring to the rule base such that the appropriate rules are fired.

• There are several methodologies but in Module 1 we have

discussed four methods (the compositional operators - Section
1.3.3).

• Two most common methods used in fuzzy logic control are the
max-min composition and the max-(algebraic) product composition.

• The inference or firing with this fuzzy relation is performed via the
operations between the fuzzified crisp input and the fuzzy relation
representing the meaning of the overall set of rules.

• As a result of the composition, one obtains the fuzzy set describing

the fuzzy value of the overall control output. 54
 Several Inference Procedures that
can be used in FLCS….
 Max-Min
 Max-Algebraic Product (or Max-Dot)
 Max-Drastic Product
 Max-bounded Product
 Max-bounded sum 2 most
 Max-algebraic sum commonly used
 Max-Max
 Min-Max
55

Example 2.2
• Consider a simple system where each rule comprises two antecedents and
one consequent. A fuzzy system with two non-interactive inputs x1 and x2
(antecedents) and a single output y (consequent) is described by a
collection of n linguistic if-then propositions:

• IF x1 is A1(k) and x2 is A2(k) THEN y(k) is B(k), k = 1,2,...n.

where A1(k) and A2(k) are fuzzy sets representing the kth antecedent pairs
and B(k) are the fuzzy sets representing the k th consequent.
• Based on the Mamdani implication method of inference, and for a set of
disjunctive rules, the aggregated output for the n rules will be given by

mB(k)(y) = Max Min[mA1(k)(input(i)), mA2(k)(input(j))]

• The equation above has a simple graphical interpretation, as seen in the

accompanying figure. 56
 Example of Fuzzy inferencing Using
Mamdani’s Max-min Compositional Operator
Rule 1

m m m
A11 A12 1 B1
1 1

min

0 x1 0 x2 y
input(i) input(j)

Rule 2
m m m
A21 1 A22 1 B2
1
min

0 x1 x2 y
input(i) input(j)

Defuzzification
m
1

y
y*

Some notes on the max-min inferencing method of Example 2.2

• The figure illustrates the graphical analysis of two rules, where the
symbols A11 and A12 refer to the first and second fuzzy antecedents
of the first rule, respectively, and the symbol B1 refers to the fuzzy
consequent of the first rule.

• The symbols A21 and A22 refer to the first and second fuzzy
antecedents, respectively, of the second rule, and the symbol B2
refers to the fuzzy consequent of the second rule.

• The minimum function arises because the antecedent pairs given in

the general rule structure for this system are connected by a logical
“and” connective.

• The minimum membership value for the antecedents propagates

through to the consequent and truncates the membership function
for the consequent of each rule. 58
 • This graphical inference is done for each rule.

• Then the truncated membership functions for each rule are

aggregated.

• The aggregation operation max results in an aggregated

membership function comprised of the outer envelope of the
individual truncated membership forms from each rule.

• Then, a crisp value for the aggregated output y* can be

obtained through the defuzzification technique which will be
discussed later in the next section.

59

Example 2.3

• In this example we show how another compositional operation

results which is based on the Larsen’s Max-product inference
technique.
• Based on the max-product implication technique, for a set of
disjunctive rules, the aggregated output for the n-th rule would be
given by

mB(k)(y) = Max [mA1(k)(input(i)) mA2(k)(input(j))]

• The resulting graphical equivalent of the equation above is shown

in the following figure.
• The effect of the max-product implication is shown by the
consequent membership functions as scaled triangles.
• This figure also shows the aggregated consequent resulting from a
disjunctive set of rules and a defuzzified value, y*, resulting from
the defuzzification method. 60
Example of fuzzy inferencing using
Larsen’s Max-product compositional operator

Rule 1
m m m
A11 A12 1 B1
1 1

0 x1 0 x2 y
input(i) input(j)

Rule 2
m m m
A21 1 A22 1 B2
1

0 x1 x2 y
input(i) input(j)

Defuzzification
m
1

y
y*

2.2.5 Defuzzification

Fuzzy Controller

Knowledge
Base

Fuzzy Fuzzy Crisp

Inference
Fuzzifier Defuzzifier
Engine
`

Process Output Crisp Control

& State Signal
Process

62
 Some Notes of
Defuzzification
• Defuzzification is a mapping from a space of fuzzy control
actions defined over an output universe of discourse into
a space of non-fuzzy (crisp) control action.
• This process is necessary because in many practical
applications crisp control action is required to actuate the
plant.
• The defuzzifier also performs an output denormalization if
a normalization is performed in the fuzzification module.
• There are many methods of defuzzification that have
been proposed in recent years.
• Unfortunately, there is no systematic procedure for
choosing a defuzzification strategy. 63

Four most common

defuzzification methods:

• Max membership method

• Center of gravity method
• Weight average method
• Mean-max membership method

64
(i) Max membership method
• This scheme is limited to peaked output
functions.
• This method is given by the algebraic
expression:
mz(z*)  mz(z) for all z Z
m
1

0 z
z*
65

(ii) Center of gravity method

• This procedure is the most prevalent and
physically appealing of all the defuzzification
methods.
• It is given by the algebraic expression

mz z* 
 m ( z ).z
Z

1  m ( z) Z

0
z
z* 66
(iii) Weight average method
• This method is only valid for symmetrical output
membership functions.
• The weight average method is formed by weighting
each membership function in the output by its
respective maximum membership value, z.
• It is given by the algebraic expression
m  m ( z ).z
z* 
z

 m ( z)
1
z

0
z1 z2 z 67

(iv) Mean-max membership method

• This method is closely related to the first method,
except that the locations of the maximum
membership can be non-unique.
ab
• This method is given by the expression z* 
2
m

0 a z* b z
68
 2.3 Design procedure of the fuzzy logic controller

• There is a general procedure that can be followed for designing a

fuzzy control system.

• Firstly, the designer have to identify the process input and output
variables that need to be considered. Thus, one must have a good
knowledge on the system to be controlled.

• Next, one should determine on the number of fuzzy partitions (or

fuzzy subsets) for the input and output linguistic variables.

• The number of fuzzy partitions of the input-output spaces should be

large enough to provide an adequate approximation and yet be
small enough to save memory space.

• This number has an essential effect on how fine a control can be

obtained.

69

• In the third step, the designer has to choose the membership

functions for the input and output fuzzy variables.

• There are different types of membership functions available as

discussed in the previous module, however, the most common
types are triangular, trapezoidal and bell-shaped functions.

• After choosing the membership functions, the designer need to

derive the fuzzy control rules based on one of the four methods
that have been discussed.

• In many practical applications method (i) i.e. deriving from

expert experience and control engineering knowledge have
been used.

• Next the inference engine need to be defined, but there is no

systematic methodology for realizing the design of an inference
engine. Most practitioners use empirical studies and results to
provide guidelines for their choices.
70
• Finally one have to choose the right choice of defuzzification
method for their particular application.
• Finally, once the fuzzy control system has been constructed, the
simulation of the system can be carried out.
• The performance of the system can be analyzed. If the results
are not as desired, changes are made either to the number of
the fuzzy partitions or the mapping of the membership functions
and then the system can be tested again.
• This trial-and-error approach will continue until satisfactory
results are obtained.
• Although this method is rather time-consuming and nontrivial, it
has been widely employed and has been used in many
successful industrial applications.

71

The design procedure of a fuzzy logic control system

Design Planning
-identify process input & output variables
-Identify Controllers inputs & outputs
- determine the number of fuzzy partitions
- choose types of membership functions
- derive fuzzy control rules-based
- define inference engine
- choose defuzzification method

Fuzzy Logic
Parameters Tuning Controller Operation
- mapping of membership
- Fuzzification
functions
- Fuzzy Inference
- fuzzy inference rules
- Defuzzification
- scaling factors

Simulation & Testing

N OK
o
Yes

End 72
 2.4 Summary of Module 2
• As fuzzy logic has been successfully applied to many control
problems, more than any other areas of applications, in this
module we review some control system basics.

• In this module also we discussed some of the reasons behind

the success of fuzzy logic control systems.

• The basic components of a fuzzy logic controller has been

discussed which include techniques in fuzzification,
knowledge base, inference mechanism and defuzzification.

• The design procedure of the fuzzy logic controller is also

discussed.

• Although we more specifically discussed fuzzy logic control,

the same techniques can be applied for the other areas as
well such as pattern recognition and diagnostic systems.

73