Chapter 3

Case Studies and Applications

of Fuzzy Logic

Dr Mohamad Hafis Izran B Ishak

Fakulti Kejuruteraan Elektrik
Universiti Teknologi Malaysia

Fuzzy Logic Control Systems

Chapter 3 Objectives
To understand several fuzzy logic control applications.

To understand how to apply fuzzy logic in practical applications.

To be able to understand the implementation of fuzzy logic

control in several applications.

To study several applications of fuzzy logic in consumer products

and industrial systems.

Discussions on the trends of fuzzy logic research and applications.

At the end of the course the student should understand how

fuzzy logic is applied in practical applications.

2
 Chapter 3 Contents
3.1 Fuzzy control of an inverted pendulum
3.2 Fuzzy control of a water bath
3.3 Fuzzy control of traffic lights
3.4 Intelligent Diagnosis of Power Transformers
3.5 Several issues in the application of fuzzy logic for
real-time control
3.6 Industrial/Commercial examples of fuzzy logic
control
3.6.1 Fuzzy washing machine
3.6.2 Canons fuzzy auto-focus camera
3.6.3 Minoltas fuzzy camera
3.6.4 Sendais fuzzy subway train
3.7 Summary of Chapter 3

Case Study #1

3.1 Fuzzy Control of an Inverted Pendulum

An inverted pendulum is a classic example of a
nonlinear and an unstable system
The control objective is to stabilize the inverted
pendulum in an upright manner
Several simulation packages have been developed.
In this course students are required to use a
simulation package developed by Togai Infralogic
Inc., U.S.A. as an assignment.
The simulation objectives are:
To understand how to design a fuzzy logic controller
To understand the components of a fuzzy logic
controller
To understand the operations of a fuzzy logic
controller

4
 Study the system

Fuzzy logic inverted pendulum control system

Vertical line + v

Fuzzy Logic Motor
Controller v

Initial steps to take when designing a fuzzy controller

Plan your design
What are its objectives?
What are the process inputs and outputs?
What are the inputs to the controller?
What are the controller outputs?

Observing the inverted pendulum system

Through observations the input and output
fuzzy variables can be identified. This
inverted pendulum has a fixed base.
Vertical Axis
Angular velocity,
Fuzzy Controller Variables
Bob Mass

Input: ~ Angle between pendulum shaft

and vertical line, Angle
~ Angular velocity of pendulum shaft , Shaft

Output: ~ Motor current or velocity, v Motor, v

(In this case, it is not necessary to take the
change of the control signal as the output)

6
 Quantize the fuzzy variables

The fuzzy variables have to be broken Negative Medium NM

into smaller modules which we call Negative Small NS

Zero ZE
quantization. Positive Small PS
Positive Medium PM
Quantize the input and output variables
into several modules which we call
fuzzy subsets and assign the appropriate
labels as given in this example.

You may quantize your variables (a).Gaussian

according to complexity of the problem

(in this case, we quantize each fuzzy (b). Triangular
variable into five fuzzy subsets).

Assign appropriate membership (c). Trapezoidal

functions to each fuzzy subset.

Three different shapes of membership functions

You may choose any kind of shape or

size of the membership functions.

Membership Functions
For simplicity lets choose triangular membership functions for
each of the fuzzy subsets of the three fuzzy variables.

m
NM NS ZE PS PM
1

0
0 Pendulum Angle,
m
NM NS ZE PS PM
1

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

0
0 Motor Velocity, v

Membership functions (fuzzy subsets) of the three fuzzy

variables. The fuzzy subsets are overlapped by about 25%.

8
 Designing the Inference Engine

The inference engine consists of the following:

Rule base
Encoding technique (compositional operator)

We need to develop rules to solve the stabilization problem.

This can be done by observing the states of the pendulum.
Rules are in the form of:

IF Conditions THEN Actions

The rules can be formulated through experience or through the use

of examples given by others (see Module 2).
In most control problem the encoding technique use is the max-
min composition.

Example of rules to control the pendulum

If pendulum angle is negative but medium = NM
and the angular velocity is about zero
then motor velocity should be position and medium
v = PM

IF =NM AND =ZE THEN v=PM

= PS

If pendulum angle is positive and small

and the angular velocity is about zero
then motor velocity should be negative small
v = NS

IF =PS AND =ZE THEN v=NS

= ZE
If pendulum angle is about zero
and the angular velocity is about zero
then motor velocity should be about zero
v = ZE

IF =ZE AND =ZE THEN v=ZE

10
 Rule Base

Basic fuzzy rules for controlling the pendulum.

IF X AND/OR Y (= Antecedents) THEN Z (= Consequent)

Antecedents

Pendulum Angle,
NM NS ZE PS PM
NM PM Consequent

NS PS ZE
ZE PM PS ZE NS NM
PS ZE NS
Steady-state rule
PM NM

11

Normally we would use matrix for simplicity to observe these

rules.
For a 5x5 antecedents there would be a total of 25 rules.
However, not all the banks in the matrix need to be filled up as
some rules do not necessarily need to be fired.
The basic fuzzy rules to control the pendulum are given as
follows:
Pendulum Angle,
NM NS ZE PS PM
Consequent
NM PM
NS PS ZE
ZE PM PS ZE NS NM
PS ZE NS
Steady-state rule
PM NM

12
 Choice of Encoding or Inference Techniques

As discussed we can choose either correlation minimum encoding or

correlation product encoding schemes.
Examples are given here.
IF e is PL AND e is ZE THEN u is PL
Input Fuzzy Variables Output Fuzzy Variable
e=PL at 0.5 e=ZE at 0.2 u=PL at 0.2
me me mu

e e u
Correlation-minimum-inference procedure

IF e is PL AND e is ZE THEN u is PL

Input Fuzzy Variables Output Fuzzy Variable

e=PL at 0.5 e=ZE at 0.2 PL PL
m u m e m e
ZE

me me mu

e e u
Correlation-product-inference procedure
13

Defuzzification Techniques
There are several defuzzification techniques.
Suppose the centroid defuzzification is used the following will be observed.

IF e=PS AND e=ZE THEN u = PS

m e m u
me
PS ZE PS

0 0 0
IF e=ZE AND e=ZE THEN u = ZE
me m e m u
ZE ZE
ZE

0 0 0

PS
m u
Value of e Value of e
at instant t at instant t ZE
m u

Fuzzy centroid, u

Example showing centroid defuzzification

with correlation-minimum encoding scheme

14
 Test run the fuzzy controller

When the control system has been configured, we can test the
system.
Run your controller.
If results not satisfactory then make minor modifications or
adjustments to the following:
(1). Scaling factors
(2). Overlap of membership functions
(3). Check if rules are correct
(4). Increase rules if necessary
(5). Increase quantization if necessary
Re-run the system.

15

FUZZY CONTROL OF AN INVERTED

ROTARY PENDULUM

Main objective
To design a Fuzzy Logic Controller to balance the
inverted pendulum at a specific orientation
within a limited range.

To control and stabilize the rotary inverted

pendulum using fuzzy logic control
through:
software simulation (Visual Basic 5.0) and
real-time control on hardware via PC-based
using DOS platform (Borland C++ 5.02 as
editor and iC-96 as compiler)
 SOFTWARE REQUIREMENTS

Visual Basic 5.0

Borland C++ 5.02
iC-96 Compiler V2.3
MCS-96 Relocator and Linker V2.4
iECM-96 V2.3
Fuzzy techniques

HARDWARE
REQUIREMENTS
The Micro-controller board
UC96-SD version 2.0
KRi Inverted pendulum
model PP-300
rotary inverted
pendulum structure
servo drive unit
power supply
 FUZZY LOGIC CONTROL SYSTEM
DESIGN METHODOLOGY
Start

Study the System

-determine objectives
-identify process and
controller's input and output

Fuzzification
-quantize the input and output
variables
-define the membership
function

Inference Mechanism
-derive fuzzy control rules-
based
-define fuzzy inference engine

Defuzzification
-choose defuzzification
method

Fuzzy Controller
Parameters Tuning
Operation
-mapping of
-Fuzzification
membership function
-Fuzzy Inference
-fuzzy inference rules
-Defuzzification

Simulation & testing

No
Performance
OK ?

Yes

End
 FUZZY LOGIC CONTROL
SYSTEM BLOCK DIAGRAM

Set-point err2 err

Fuzzy Fuzzy
(Vertical line) u
Logic u2 derr Logic
derr2 Motor
Controller Controller

2 1

Input: 1) Angle between pendulum shaft and vertical line,

2) Angular Velocity of pendulum shaft,
3) Angle between motor arm and horizontal line,
4) Angular Velocity of motor arm,

Output: 1) Motor PWM, u

DYNAMIC EQUATIONS OF
THE INVERTED PENDULUM

J o + m 1 L o 2 + 1 2 sin 2
1 m 1 1 L o cos 1 o
+
m 1 1 L 0 cos 1 J 1 + m 1 1 2 1
C o + 12 m 1 1 2 sin 2 1 1 - m 1 1 L o sin 1 + 1
2
m 1 1 2 sin 2 1 1 o
1 +
- 2 m 1 1 sin 2 1o 1
2
C1
0
- m 1 1 g 1 sin 1 0

o
0

1 0 0

0
- dg *


o
o
0 de - bg - bi df - bd

o
o
1 =

0
bc - ad bc - ad bc - ad o bc - ad
+
1 1 0

u~ y 0 0 1 0

0 0
1
1 ag - ce ai ah - cf 1 cg *

0 bc - ad bc - ad bc - ad
bc - ad

1
 REAL TIME FUZZY LOGIC
CONTROLLER DESCRIPTION

Singleton fuzzy output is chosen due to its

faster processing speed

n Bn = the weight of the

BnK n rule which is fired
Z* t 1
n
Kn = singleton output
t 1
Bn value for that
specific rule

INPUT MEMBERSHIP FUNCTIONS

Input membership functions for both controllers are

similar
Singleton controller does not have output
membership function

NM NS ZE PS PM NM NS ZE PS PM
1 1

0 err 0 derr
-5.4o -2.7o 0 2.7o 5.4o -5.4o -2.7o 0 2.7o 5.4o

First Input Membership Second Input Membership

Function Function
 FUZZY CONTROL RULES
err \ derr NM NS ZE PS PM
NM 855 837 804 346 0
NS 694 316 281 0 -290 First Fuzzy
ZE 641 271 0 -288 -600 Controller
PS 259 0 -284 -272 -713
PM 0 -324 -763 -796 -852

err \ derr NM NS ZE PS PM
NM -698 -539 -425 -250 -155
NS -74 -94 -72 -233 -477 Second Fuzzy
ZE 47 43 12 -41 -52 Controller
PS 200 192 254 517 675
PM 226 243 259 396 699

EXPERIMENTAL RESULT OF REAL TIME CONTROL

Pendulum Position Vs Number of Arm Position Vs Number of Sample

Sample
500
Arm Position

300
Pendulum

500
Position

300
100
100
-100 -100
-300
-300
-500
-500
1
92
183
274
365
456
547
638
729
820
911
1002

94

187

280

373

466

559

652

745

838

931

Number of Sample Number of Sample

Pendulum Velocity Vs Number of

Sample Arm Velocity Vs Number of Sample

500 500
Pendulum

300
A rm V elocity

300
Velocity

100
100
-100
-300 -100
-500 -300
-500
105

209

313

417

521

625

729

833

937
1

1 103 205 307 409 511 613 715 817 919 1021
Number of Sample Number of Sample
 EXPERIMENTAL RESULT AFTER DISTURBANCE IS
ADDED
Pendululm Position Vs Number Arm Position Vs Number of
of Sample Sample

500
Pendulum

500
Position

Arm Position
300 300
100 100
-100 -100
-300 -300
1
101
201
301
401
501
601
701
801
901
1001
-500
1 134 267 400 533 666 799 932
Number of Sample Number of Sample

Pendulum Velocity Vs Number Arm Velocity Vs Number of

of Sample Sample

Arm Velocity
Pendulum

500
Velocity

500
300 300
100 100
-100 -100
-300 -300
-500 -500
1
109
217
325
433
541
649
757
865
973

103

205

307

409

511

613

715

817

919

1021
Number of Sample Number of Sample

EXPERIMENTAL RESULTS WHEN SOME CONTROL

RULES ARE TAKEN OUT

Both Controllers have only (3x3) rules, instead of (5x5) rules

Pendulum Position Vs Arm Position Vs Number of

Number of Sample Sample
Pendulum

500 500
A rm P osition
Position

300 300
100
-100 100
-300 -100
-500 -300
115
229
343
457
571
685
799
913
1

-500
1 131 261 391 521 651 781 911
Number of Sample
Number of Sample
 ANALYSIS OF RESULTS

The research has

shown the
robustness of the
fuzzy logic
controller under
disturbances and
plant uncertainties

Next project- coming up

Swing up the inverted pendulum and balance

at a specific position
Using neuro-fuzzy controller for better
performance
 Case Study 2

Water-Bath Temperature
Control System

31

Objectives and Applications

To control precisely the

temperature of the
water/liquid materials in the
vessel
Industrial applications involve
experimental production of
drink products, chocolates,
sauces, etc.
Industries include Yeo Hiap
Seng, Nestle, F&N, Drinho, etc.

32
 Schematic diagram of the Water-
bath temperature control system

A/D and D/A

Interface

Control Signal

Sensor
Stirrer

AC ~
Coil
(Heater)

33

Block diagram of the water-bath

temperature control system

Disturbance

Control signal (Heater)

Error signal

Reference + Output
Fuzzy Logic Water Bath
(Desired Temp.) Temp.
Controller
-

Sensor

34
 Some Facts on the System

Math-model of system:

Practical system complexity

Non-linearity in sensors and relays
y (k 1) 0.998 y (k ) 0.222u (k ) Noise
Disturbance
Non-adiabatic
The performance of the system can
be measured from the error between Controller Limits: 0V(min)-5V(max)
the output and the reference signals (However, these characteristics are
not present in the above math-model
except for controller limits)
35

Q: Design a fuzzy controller for the

Water Bath Temperature Control
System
First identify the fuzzy
variables: 2 inputs and 1
output

36
 Configuration of the fuzzy controller

Error
e
Change in
FLC Control Signal
u
Derivative
of Error
e

Two input variables:

Error in temperature of the liquid, e(k) = r(k) - y(k)
Rate of change of error, e(k) = e(k) - e(k-1)
One output variable:
Change in the control signal, u(k)

37

Quantization of the Fuzzy Variables

For simplicity the 3 fuzzy variables can be broken into 3 fuzzy subsets
and the membership functions can be overlapped as follows:

e, e, and u N Z P
e
Linguistic Term Label
Negative N e
Zero Z N Z P
Positive P e

e
u
N Z P
The universe for the 3 variables can be
set accordingly through observation.

38
 Scaling the universes of discourse
We need to quantify each universe of discourse correctly within
the range of the respective variables.
Scaling factors can be used to scale the universes of discourse.
They act like gain control.

Ge
Error
e Change in
Gu Control Signal
FLC
Derivative u
of Error
e Ge

39

Determining the rule base

Every fuzzy logic system must have a rule base.
The rule base is used to infer the actions that need to be
taken based on the current conditions.
Example the centre rule which is the steady-state rule can be
written as follows:
IF Error in Temperature is about Zero
AND the Derivative of Error is about Zero
THEN the Change in Control Input is about Zero
(IF e=Z AND e=Z THEN u=Z)
A rule can be written in triple form such as this: Error, e
(Z, Z; Z)
N Z P
N
e Derivative
e u of Error, e Z z
A matrix of the rule base can be set up as shown.
P
The 1st row and the 1st column are the antecedents
and the boxes in the matrix are the consequents.

40
 Determining the rules

The choice of the consequents is based on observation and engineering

experience.
A simple way to fill up the rule base matrix is to use the information given in
Module 2.
Fill up the matrix such that the rule base is complete.

Error, e

N Z P

Derivative N
of Error, e
Z

41

Determining the encoding technique

In many control application the Example 3.1
Mamdanis max-min composition
technique is largely used.
N Z P
An error reading that is observed e
0.6
will fire the appropriate rule or rules
in the rule base. e(k) e
Example 3.1 shows how the N Z P
minimum encoding technique e 0.7
determines the output suppose the
rule (Z,Z;Z) is fired. e(k) e
Max composition is used to u N Z P
determine if more than 1 of the 0.6
same consequent resulted. This
occurs when several rules are fired. u
A defuzzification technique will be
Determined output u(k)
used to give a crisp output value.

42
 Defuzzification

c
Defuzzification is required to produce the actual
signal that the plant can use. Z
Thus, the fuzzy output value need to be
defuzzified.
The output of the fuzzy controller is usually the
change in the control signal.
The actual control signal to the plant is thus:
Max defuzzification

u(k+1) = u(k) + u(k)

c

In Example 3.1 suppose only one rule is fired (Z,

Z; Z) as shown, if the max defuzzification Z

technique is used, the output crisp value is:

c(z*) c(z) for all z Z
For centroid defuzzification, the value is given as
follows:
z*
C ( z ).z Centroid defuzzification
C ( z)
43

Example 3.2

Suppose the following rules are fired:

Rule1=(P,Z;P), Rule 2= (P,N;Z), Rule 3=(N,P;Z) and Rule 4=(Z,P;P).
In this case the max-min composition would be computed.
First the min computation is used to obtain the consequents from
each of rules fired which are: Rule 1=P, Rule 2=Z, Rule 3=Z and Rule
4=P.
This shows that there are 2 consequents for P from Rules 1 and 4
and 2 consequents for Z from Rules 2 and 3.
Now the max computation is used for obtaining only 1 P
consequent and 1 Z consequent.
Then any of the defuzzification technique can be used to get the crisp
output value.

In this example we show how the centroid defuzzification strategy is

used to solve this problem.

44
 P, Z; P Z, P; P

e N Z P N Z P
e
0.75
0.8
e e
e N Z P N P
e Z
0.3 0.4
e
e
u N Z P
u N Z P
0.3
0.4
u
u at P =0.3 u at P =0.4 u

Consider the first 2 rules fired:

Rule1=(P,Z;P) and Rule 4=(Z,P;P), this means that (e, e; u).
Suppose Rule 1: P=0.75, Z =0.3, thus, min (P, Z) --> P at 0.3

45

Suppose Rule 4: Z =0.8, P=0.4, thus, min (Z, P) --> P at 0.4

Since the consequent P is fired twice having 2 membership values, 0.3

and 0.4 thus, we need to use the max compositional operator to
obtain only 1 P value of the consequent.

Thus, max (P=0.3, P=0.4) will give P=0.4.

Similarly for Rules 2 and 3, the same computation will be done to

obtain only 1 N consequent (see next slide).

Consider the next 2 rules fired:

Rule2=(P,N;Z) and Rule 3=(N,P;Z)
Suppose Rule 2: P=0.5, N =0.6, thus, min (P, N) --> Z at 0.5
Suppose Rule 3: N =1.0, P=0.4, thus, min (N, P) --> Z at 0.4

46
 P, N; Z N, P; Z

N Z P e
e N Z P
1.0
0.5

e e
e N Z P N P
e Z
0.6
0.4
e e
u N Z P
u N Z P
0.5
0.4
u u
u at Z =0.5 u at Z =0.4

As the consequent N is fired twice having 2 membership values, 0.5

and 0.4 thus, we need to use the max compositional operator to obtain
only 1 N value of the consequent.
Thus, max (Z=0.5, Z=0.4) will give Z=0.5.
Suppose we use the centroid defuzzification strategy. The next slide
will show how this is done to get the crisp output.

47

For the example given, suppose the u N Z P

centroid defuzzification technique is
0.4
used to calculate the crisp output.
u
Graphically this can be shown as u at P =0.4
follows.
u N Z P
The first consequent P was obtained
at P =0.4 and the second 0.5

consequent Z at Z =0.5. u
u at Z =0.5
The centroid defuzzification
algorithm will calculate the crisp
value of u from the following u N Z P
equation: (k ).k
u
u
u
(k )
u
where k is the sample number across Crisp value of u
the universe for u.

48
 Exercise Based on Simulations

Study the output of the water bath fuzzy control system at

the particular sampling interval.

Note down the sampling interval, the setpoint (r), the

output (y), error (e), and del_error (e).

By studying the rule-base matrix, write down the rules

that are fired at this interval in triple form, eg. (P, Z; N),
etc..

Study the membership functions table of e and e and

indicate the appropriate alpha cut-set (firing angle) based
on the input values of e and e.

49

(continued)
Supposing the controller is designed using the
max-min inference algorithm, show in the graph
the firing angles of all the consequents based on
the rules fired. Plot the resultant waveform at this
sampling interval on a graph paper.

Using the centroid defuzzification strategy,

calculate the output u. (Approximately divide
the waveform into discrete samples of 0.5). Show
how u is calculated.

50
Membership Function Table of e, e and u
of Water Bath Fuzzy Controller

e N Z P
1.0

0.5

-50 -25 0 25 +50 e

e N Z P
1.0

0.5

-50 -25 0 25 +50 e

u
1.0
N Z P

0.5

-5 -2.5 0 +2.5 +5 u

51

Example of Exam Question

Design a fuzzy logic controller for a control system using

the following specifications:
Each antecedent (for E which is error and DE which is
change in error) and consequent (DU which is change in
control input) must have only 3 fuzzy sets: Negative (N),
Zero (Z) and Positive (P).
The membership functions for the two antecedents and
one consequent are already given in Table 1 which are in
normalized form.
The universes of discourse for each of these variables are
scaled as follows: E=5.0, DE=5.0 and DU= 1.0.
Use the max-min inference procedure and the centroid
defuzzification procedure.

52
 Answer the following questions:
(a). Using engineering common sense, write down the most
appropriate fuzzy control rules in matrix form for this system.
(4 marks)

(b). Based on your rules in 1(a) above, write down all the rules that
would fire and also compute the consequent firing angle for the
following cases:
(i). E= 30.0 and DE= 40.0
(ii). E= 10.0 and DE= 20.0
(iii). E= -5.0 and DE= -10.0

Your answer should be in triple form as follows [for example]. Also

approximate your answer to the nearest 0.1 accuracy for the
membership values [for example, 0.2, 0.3, 0.5, 0.7, 0.8, etc.].
Do not use values such as 0.22, 0.51, 0.65, etc. in your
calculations/computations. (9 marks)
53

(c). For the case of (b) (iii), sketch the resultant waveform of
the consequents.
(3 marks)

(d). For the case of (b) (iii) also and using Table 1, calculate
the actual output of the fuzzy controller using a discrete
sample of 1 for the universe.
(4 marks)

(e). Try part (c) and (d) again using the max-dot (alg.
product) inference procedure with a discrete sample of
2.

54
 Table 1
Antecedents
E,DE
N Z P
1.0
0.8

0.6

0.4

0.2

0
-10 -8 -6 -4 -2 0 2 4 6 8 10 E, DE

Consequent

DU Z
N P
1.0
0.8

0.6

0.4

0.2

0
-10 -8 -6 -4 -2 0 2 4 6 8 10 DU

55

Case Study #3

3.3 Fuzzy Traffic Lights Control

Objective: To control traffic lights

that can respond to density of vehicles
in an efficient manner.

At CAIRO, a traffic lights control

simulator has been developed which
can use:
a fuzzy logic controller
conventional preset timer

56
 Fuzzy Traffic Lights Control

Simulation Objectives
Understanding fuzzy logic application for traffic
lights control
Understanding the components of a fuzzy logic
controller
Able to set up rules for controlling traffic
conditions
Able to compare and understand the advantages
of fuzzy logic controller over fixed-time
controller of traffic lights

Control Objectives
To control traffic flow optimally at an isolated
junction
To minimise waiting time
To reduce fuel costs
To reduce waste of man-hours
57

Comparison between Conventional and Fuzzy traffic controllers

Conventional traffic lights controller

use a preset cycle time to change lights
combine preset cycle time with proximity sensors

Fuzzy traffic lights controller

mimic human intelligence for control of traffic conditions
example of a traffic lights control rule

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.

58
 Fuzzy Traffic Lights Control

Simulation of Traffic Lights Control

Control signal

Sensor Readings
Fuzzy Logic Relays/ Traffic
Traffic Flow Controller Switches Lights

Configuration of the Fuzzy Controller

Traffic Condition
(Arrival) Lights

R
FLC Switches A
G
(Queue)
Traffic Condition

59

Traffic
Lights

Sensors
State
Machine
Traffic
Counter
Light
- Queue -
Arrival
FLC Interface

Block diagram of the fuzzy

traffic lights controller

60
 Fuzzy Traffic Lights Control

CAIROs Fuzzy Traffic Lights Simulator

Design Assumptions
Traffic conditions are considered in 4 directions but
constraint to two: (1) Arrival and (2) Queue
Also North and South are assumed as one direction
whereas East and West as another
Only straight traffic flow is considered (no
turnings)
Fuzzy controller has 2 inputs (Quantity of Traffic
at Arrival and Queue) and 1 output (Extension of
Green Lights of the Arrival Traffic Lights (Green))
All fuzzy variables are quantized into 4 fuzzy
subsets
N
The shape of membership functions are trapezoidal
at the sides and triangular in the middle. W E
Flow density of cars can be controlled for each
direction.
S

61

Fuzzy Traffic Lights Control

Fuzzy Controller Design

Quantization of Fuzzy Variables

For simplicity, we quantize all the three fuzzy variables in a similar way,
i.e. into four fuzzy subsets:

Fuzzy Input Variables

Arrival Queue
LinguisticTerm Label LinguisticTerm Label
Too Many TM Large L
Many MY Many M
Few F Small S
Almost None AN Very Small VS

Fuzzy Output: Variable:

Extension of Green Lights

LinguisticTerm Label
Longer L
Medium M
Short S
Zero Z

62
 Membership Functions
Arrival ARRIVAL
For simplicity, triangular F MY TMY
AN
membership functions are
used for each variable.
In this case, the cardinals for No. of Cars
the universes of discourse are Queue QUEUE
integers.
VS S M L
The membership functions are
overlapped by about 25% to
allow smooth transition from 1 No. of Cars
fuzzy subset to another.
E.Time EXTENSION TIME (Green)

Z S M L

Extension Time (Secs)

63

Determining the rules

A rule base can be developed in the following

manner.
Two examples of rules:
IF there are too many cars (TMY) at the arrival
side
Rule#1 AND very small number of cars (VS) queueing
THEN extend the green light longer (L).
Arrival
IF there are few cars (F) at the arrival side
AN F MY TMY
Rule#2 AND very small number of cars (VS) queueing
THEN extend the green light short (S). VS

S
Fill up the rule base matrix as shown: Queue
In this application the centroid defuzzification is used. M

64
 Simulation procedure for Traffic Lights Simulation

Get into Windows Environment

Execute the Fuzzy Traffic Control Software (FTC)
Select fuzzy controller and configure your fuzzy controller
appropriately as have been discussed in the class
You need to select BOTH from the controller menu
You need to configure the density of cars for each direction. Try:
N=90, W=50, S=80, E=40
Then select <GO>
Fixed-time controller mode will operate first for 2 minutes
Observe the traffic conditions until the fixed-time controller operation
is completed

65

Write down the maximum number of cars you observed stopping at

the junction (any direction)
After the fixed-time cycle has finished, the fuzzy controller mode wil
automatically run for 2 minutes
Observe traffic conditions and write down the maximum number of
cars you observed stopping at the junction (any direction)
Highlight Graph and Plot:
(1). Flow density
(2). Wait Time
(3). Cost Function
Your fuzzy controller should show better results
If not, you need to reconfigure your fuzzy controller
Try simulating other conditions

66
 Example of Previous PSM Project

Application of Fuzzy Control to an AC Motor

67

68
 69

Fuzzy Control of an AC Motor

70
Intelligent Elevator
Control using Fuzzy
Logic

This has been a

Masters research
at CAIRO

Research Objectives

To apply a new fuzzy technique in

elevator group control
The fuzzy technique is based on the
Ordinal Structure Model of Fuzzy
Reasoning due to its simplicity in
meeting multiple constraints

72
 Hall Call
Group Controller States of Each Car
Car Position
Command Car Calls
Direction etc.

Car Controller 1 Car Controller 2 Car Controller n

Car Call
Actuator Actuator Actuator

Elevator 1 Elevator 2 Elevator n

Evaluation Criteria of
System Performance

minimize the waiting time of passengers at a

floor.
minimize the time passengers need to spend
in an elevator.
minimize crowding in elevators.
minimize the travelling distance of each
elevator.
Conventional System Fuzzy System
Usually has only a Can meet multiple
single control objective constraints or control
objective
Fixed evaluation
System developed
function
based on rules i.e.
Elevator stops to pick based on knowledge
up based on nearest and experience of
distance experts/human
Consists of multiple
evaluation function
Optimized operations

Examples of Rules in
Fuzzy Elevator System
IF (a new hall call in the high-zone is registered)
AND (the number of cars which are ascending to the
high-zone is large)
AND (the number of free cars is small)
THEN (select cars except free cars in the group as
candidates)
AND (select one assigned car from among the
candidates by using evaluation function.)

76
 Destination Oriented Elevator

press the push buttons for your desired

floor
read the elevator car assignment on the
display screen
walk to your elevator
< A
1 2 3
4 5 1
6
7 8 9 Keypad
Up /
Down * 0 #
Button

Destination Oriented Elevator

System by Schindler Elevator Corp.
 Advantages of the Ordinal Structure Model of
Fuzzy Reasoning compared to Conventional
Fuzzy Model
Conventional Fuzzy model
Difficult to develop the fuzzy rules when too many
conditions are observed..
Not practical to develop for a multi-input multi-output
system since need to develop based on multi-
dimensional space Difficult

Example of rule of a conventional fuzzy model:

IF Input1 is large AND Input2 is long AND Input3 is medium
THEN Output is big.

Ordinal Structure Model

Can be broken into several one dimensional space for

each input and output variable
Simple - regardless of the number of the inputs
Co-ordination of the rules are done with weights on the
rules

IF waiting time is long THEN priority is small.

IF traveling distance is medium THEN priority is medium.
IF no. of passengers is small THEN priority is big.
 For the n-inputs one-output system
Inferred value

n n
w i u i c iSi w ju jc jS j
i1 j1
y
n n
w i u iSi w ju jS j
i1 j1
where

Ri is the i-th fuzzy rule

Ai1, Ak2, Bi and Bk are fuzzy variables
n is the number of rules
wi is the weight of the rule Ri
ui is the membership degree of Ri
ci is the central position of Bi
Si is the area of the membership function of Bi

Elevator System Inputs

Short Medium Long Short Medium Long

1 1

0 0
Waiting Time Riding Time

Close Middle Far Small Medium Big

1 1

0 0
Traveling Distance
No. of Passengers
 Elevator System Inputs
Far Middle Close Middle Far
1

0
New Hall Call Area-weight (HC)
Far Middle Close Middle Far
1

0
New Hall Dest. Area-weight (HD)
Elevator System Output
Small Medium Big
1

0
0 0.5 1 Priority

Fuzzy Inference Rules

Short Big
IF Waiting Time is Medium THEN Priority is Medium
Long Small

Short Big
IF Riding Time is Medium THEN Priority is Medium
Long Small

Small Big
IF Loading is Medium THEN Priority is Medium
Big Small
 Fuzzy Inference Rules

IF Travelling Close Big

distance is Middle THEN Priority is Medium
Far Small

Close Big
IF Area_weight Middle THEN Priority is Medium
(HC) is Far Small

IF Area_weight Close Big

(HD) is Middle THEN Priority is Medium
Far Small

Configuration of the Fuzzy

Logic Controller

Waiting Time Assignment Elevator 1

Riding Time
FLC Elevator 2
Travelling Distance
Elevator 3
Number of Passengers

Area_Weight Area_Weight(
(Hall Call) Hall Dest.)
 Simulation

Intelligent Elevator Control System

3.4 Intelligent Fault Diagnosis of Power Transformers

 Summary

Major faults in transformers cause extensive damage,

interruption of electricity supply and results in large
revenue losses to power utility company.
Condition monitoring of transformers is an effective
technique to identify incipient or potential faults
inside the transformers.

Newspaper
Report
6th February 2000

90
 8th March 2000

91

New Transformer
Blast
New Sunday Times
23rd July 2000

TNB Distribution Sdn.

Bhd. Customer Services
GM (Selangor) initial
investigations revealed a
technical fault in one of
the RM1.5 million
transformers.
92
 Transformer Blast at Klang
due to improper maintenance
Estimated losses at RM4 million - TNB

93

Project Overview

Phase 1 Development
of a Database ADAPT
Software
Phase 2

Automatic
Interpretation AI
using Fuzzy Techniques
Logic
 Benefits - Immediate impacts
Use of Advanced Technology and local expertise
within Malaysia for solving complex industrial
problems
Increased efficiency and reduced operational
costs for power transformers maintenance
Early detection of abnormalities helps prevent
unscheduled outages, equipment damage, and
safety hazards.

Benefits - Future impacts

Savings of outflow of Ringgit with

less dependency on foreign
consultants
Increased expertise of local
consultants
 Dissolved Gas Analysis

Major power transformers are filled with a fluid that serves as a

dielectric media, an insulator, and as a heat transfer agent.
Normal
slow degradation of the mineral oil to yield certain gases.
Electrical fault
gases are generated at a much more rapid rate.
Different patterns of gases are generated due to different
intensities of energy dissipated by various faults.
The gases present in an oil sample make it possible to determine
the nature of fault by the gas types and their amount.

DGA
Key Gas Method
Roger Ratio Method

Purpose :
To identify fault type base on the
dissolve gases in oil
 Gases are produced by degradation of the oil as a result of elevated
temperatures which can be cause by :
lighting
severe overloading
switching transients
chemical decomposition of oil or insulation
overheated areas of the windings
bad connections which have a high contact resistance

Oil sample
from
Transformer

Gases generated from oil are :

Hydrogen (H2) Methane (CH4)
Ethane (C2H6) Ethylene (C2H4)
Acetylene ( C2H2) C.Monoxide ( CO)
Carbon Dioxide (CO2)
Different patterns of gases are generated due to different
intensities of energy dissipated by various faults

Diagnostic Method
Test Result

DGA

Key Gas Method Roger Ratio Method

-Thermal Fault -Thermal decomposition
-Corona -Partial Discharge
-Arcing -Arcing
-Cellulose Insulation Breakdown

H2 Corona The ranges of ratio are assigned to

O2 and N2 Non-fault related gases different codes which determine the fault
CO & CO2 Cellulose insulation breakdown type.
CH4 & C2H6 Low temperature oil breakdown
C2H4 High temperature oil breakdown Methane / Hydrogen
C2H2 Arcing Acetylene / Ethane
Ethylene / Ethane
 H
H yydd rroogg eenn S
S ta
ta tis
tis ti
ticcss

20 0 199
18 0
16 0

Test Value
14 0 150

12 0 123
10 0
95
80 82
60 69 65

40 45

M
M eth
e th aa n
n ee S
Statis
tatis ti
ticc ss
0 1 /0 1 /1 9 97 0 6/0 6/1 9 9 7 0 1/0 9/ 1 9 97 2 5/ 1 2/1 9 97 0 1/0 1/1 9 9 8 0 1 /0 6/ 1 9 98 11 /0 9/1 9 98 11 /1 1/1 9 98

S a m p li n g D a te
450
436
400
350
300

Test Value
250
200
150
100
50 54
37 33 34 41
0 22 21

E
E thy
thy le
le n
n ee S
S ta
ta ti
tistic
stic ss
01 /01 /199 7 06 /06 /19 97 01 /09 /19 97 25 /12 /19 97 0 1/0 1/1 998 0 1/0 6/1 998 1 1/0 9/1 998 1 1/1 1/1 998

Sa m p l in g D a te
70 0
651
60 0
50 0 500

Test Value
4 65
40 0
30 0 3 23 322

Report & Graphs for Analysis 20 0

10 0 1 45

0 12 14
0 1/01 /199 7 06/ 06/1 997 0 1/09 /199 7 25/1 2/19 97 0 1/01 /199 8 01/0 6/1 998 1 1/09 /199 8 11/1 1/1 998

S a m p l in g D a t e

Automatic
Interpretatio
n using
Fuzzy Logic

Transformer Condition

Good
Normal FUZZY
INTERPRETATION
Bad

Crisp to Fuzzy Fuzzy Inference Fuzzy to Crisp

Fuzzify Defuzzify

Real variable to linguistic variable

linguistic variable to real variable

IF CO=Hi And CO2=Hi Aggregation

then Condition A Composition
Software
output

Reports
&
Graphs
 3.5 Several issues in the application
of fuzzy logic for real-time control
Knowledge required:
Scope of work/project
Whether viable to use fuzzy logic control
Variables that can be measured
Type of actuators
Sensors to be used
PC operating environment
High/Low level programming languages
Hardware knowledge of microchips
Development systems of microchips
Knowledge regarding the process
Digital control theory
Electronics/ Digital electronics
Fuzzy logic control theory
Others

105

Simulation Exercises

Using the simulation packages developed by CAIRO, try to design your

fuzzy controller and understand fuzzy logic control system.
Observe different rules and membership values used and their impact
on the system.
There are 5 available packages:
1. Water bath temperature control
2. Traffic lights control
3. Coupled-tank liquid-level control
4. Vfuzzy Version 1.0
5. Elevator supervisory system

106
 Application using a personal computer

Interface Card
(Adclone, NS, etc.)

(Fuzzy Control Algorithm)

Sensor
Personal
Computer
Process

Actuator High-level
Languages
(C++, Pascal,
Visual Basic, etc.)

107

Application using a microprocessor/microcontroller

Fuzzy control
Personal development
Computer system or LLL

(Download)

Sensor Micro-
A/D
Process Controller
(DSP, Fuzzy Chip,
D/A
MC68HC11, etc.)
Actuator

108
 3.5.1 Applications of Fuzzy Micro-Controller (AL220)

Objective:
To use a fuzzy logic micro-controller AL220 in
practical applications.

Adaptive Logic AL220

Analog Micro-Controller
Programmable Analog IC (PAIC)
On-chip A/D and D/A converters
Four 8 bits Analog Inputs and Outputs
EEPROM / ROM versions
Minimum sampling rate up to 0.1 msec
Program directly from INSiGHT IIe Development System
 Servo Motor Control Using AL220 micro-
controller
Coupling Shaft

Flywheel Motor
Generator Drive

Tachometer Load Motor

Amplifier

Feedback
Control
y (s) u (s) Signal
Controller

PID Control System

PID Controller Servo System

Output
Setpoint k 0.96
+ k p i kd S
_ S (0.8s 1)

kp = 2.5 ki = 1.9 kd = 0.72

Fuzzy Logic Control System

Servo System
Output
Setpoint e Fuzzy Logic 0.96
+
_ Controller 0.8s+1
de
Fuzzy Inputs Variable

NL NS ZR PS PL
1

0
0 Error

NL NS ZR PS PL
1

0
0 Delta Error

Error
NL NS ZR PS PL

NL -5 -3 -2 -1 0

Delta NS -3 -2 -1 0 +1

Error ZR -2 -1 0 +1 +2

PS -1 0 +1 +2 +3

PL 0 +1 +2 +3 +5

Fuzzy rules for Servo System

 Fuzzy Logic Control Step Response

5000
4000
3000

Volt (mV)
2000
1000
0
100 200 300 400 500 600 700
Number of Sampling (100 msec)

PID Control Step Response

5000
4000
Volt (mV)

3000
2000
1000
0
100 200 300 400 500 600 700
Number of Sampling (100 msec)

Fuzzy Logic vs PID Control

on the Effect of Load Disturbances
5000
4500 Fuzzy Logic
4000
PID
3500
Load
3000 Disturbances
Volt (mV)

2500
2000
1500
1000
500
0
100 200 300 400 500 600 700 800
Number of Sampling (100 msec)
 3.6 Industrial/Commercial Examples
of Fuzzy Logic Systems

Due to the advancement of the microprocessor, many

commercial applications of fuzzy logic have been
successful since the 1980s.

In this part of the module, we discuss 3 examples of

industrial/commercial applications of fuzzy logic systems.

Fuzzy washing machine

The Fuzzy Rice Cooker
Canons fuzzy auto-focus camera
Minoltas fuzzy camera
Blood pressure meter
Sendais fuzzy subway train

117

Commercial Example #1

3.6.1 Fuzzy logic washing machine

In 1990 Matsushita (National) produced the first automatic controlled
fuzzy logic washing machine - Aisai Go (Beloved Wife) Day Fuzzy.

In 1990, the sales were so successful in Japan that it resulted in an

explosion of fuzzy logic home consumers products such as rice cookers,
camcorders, televisions, refrigerators, etc.

118
 Example of a fuzzy
logic washing machine
Samsung SW-5AI(S)

119

Example of the Samsungs

SW-5AI(S) Fuzzy logic
washing machine

Fuzzy course will

select the best
washing mode
automatically with
one touch of the
button based on
load, dirt, type of
fabric, etc.

120
Principle of the fuzzy washing machine
The fuzzy logic washing machine
operates based on the principle of
laundering :
When dirt has been removed, then
washing is stopped

Using fuzzy inference, the optimum

washing time is determined from a
wash sensor.

The wash sensor measures the

dirtiness (turbidity) of the water
through an optical sensor which is
installed near the drain valve.

121

A photo-transistor is used as a sensor to

measure dirtiness of the water.

Infrared Light Receptor

sensor (infrared
emitting Drained out water
diode)

122
 The wash sensor consists of an infrared light-emitting diode
(LED) and a photo-transistor.

The light beam generated by the infra-red LED, passing through

the wash water in the pipe, enters the phototransistor.

Wash sensor output

change over time.

123

The photo-transistor produces a voltage in proportion to the

intensity of the light.
If the clothes are dirty, the wash water being drained out will be
darker, thus less light will be passed.
The figures below show the light transmittance over time due to
the different types of dirt.
The change of the output of the wash sensor over time is shown in
the previous page.
When washing started, the dirt in the clothes is gradually washed
out and the wash water becomes dirty, causing the transmittance
of wash water to decrease as shown in (a).
The rate of decrease of the transmittance depends on the quality
of the dirt:
Fast for muddy dirt
Slow for oily dirt

124
 This is because muddy dirt is removed easily by the mechanical
force of the water flow produced by the rotation of the pulsator,
while oily dirt is not adequately removed until the detergent effect
takes place.
When most of the dirt in the clothes has been removed, the
transmittance of the wash water approaches a state of saturation.
The transmittance at saturation becomes lower when the clothes
are dirtier, and the transmittance becomes higher when the
clothes are less dirty.

125

Fuzzy inference in the washing machine

It is difficult to obtain the optimum relation between dirtiness and washing

time experimentally.
This is because there are many kinds of dirtiness of clothes which gives
different wash sensor output patterns, thus collection through experiments
are rather impossible.
Also the relation between the dirtiness and wash time is not linear.
Hence, fuzzy inference technique would be suitable for such application.

126
 Using fuzzy inference, linguistic rules can be used to solve the washing
problem and the nonlinear property of the relation is approximated by an
interpolation function of the fuzzy inference
In order to provide the inference 2 inputs are suitable:
saturation time, Ts
light transmittance, Vs
and 1 output
Washing time, Wt
Saturation time is chosen because it is related to the quality of the dirt and
light transmittance is chosen because it is related to the amount of dirt and
example of the membership functions are given as follows:

127

The washing machine can sense the following types of dirt:

Muddy dirt (faster to be washed away)
Oily dirt (need detergent and more time to be washed away)

When most of dirt has been washed away the washing machine
approaches a state of saturation:
Lower transmittance (heavy dirt)
Higher transmittance (light dirt)

Fuzzy Vs
rule base Low Middle High
Short T1 T2 T3
Ts
Long T4 T5 T6

128
 Examples of rules (2 rules are given here). A table of rules can be set
as shown.
If transmittance (Vs) is Low and Saturation time (Ts) is Long
Then Wash time (Wt) is Longer

If transmittance (Vs) is High and Saturation time (Ts) is Short

Then Wash time (Wt) is Shorter

For simplicity, the consequents of the fuzzy rules are expressed by

real numbers, T1 - T6 which are the washing times.

129

Advantages of the fuzzy logic

washing machine:
It would be very difficult to design
a rule base system without using
fuzzy logic concepts.
Thus, with fuzzy logic it reduces
amount of required memory.
Rules are acquired from skilled
launderers (experts).
Nonlinear relationship between
degree of dirtiness and wash time
can be overcome by the nonlinear
fuzzy logic controller.
Excessive or inadequate washing
can be avoided.
Many fuzzy logic washing machine
consists of one-touch button
system for the fuzzy wash.

130
 The Fuzzy Rice Cooker

131

General Overview
A Rice Cooker is an everyday used electrical
appliance which is an excellent example of a
temperature control system.

Its objective is to control the rate of heating to

have evenly cooked rice , avoid over burned rice
and most importantly minus the hassle of stove
rice cooker.

132
 Historical Overview
The rice cooker is a simple device began with a trivial
on-off switch in the 1960s, then evolved to power
control, electric timer, induction heating, and now a
neuro-fuzzy controller in which the heat application
is finely controlled, with the temperature at each
stage of the cooking process and period of time at
each stage controlled by information such as the
preferred rice stiffness, amount of cooked rice
(deduced from the temperature rise and the power
consumption).
133

There have been at least three generations in the

production of rice cooker. The first generation with
bi-metalic controls (electric foot warmer), 2nd
generation with sensors and micro CPU (electric rice
cooker and electric jar), and 3rd generation with
sensors, fuzzy, and neural networks (induction
heating rice cooker). The next generation will use
sensors as well as even more general kinds of
nonlinear information processing.

134
Fuzzy Rice Cookers in the Market

Sanyo Neuro-Fuzzy Deluxe Rice

Cooker (Model ECJ-5205SN, ECJ-
AC4E)

Panasonic National Neuro-Fuzzy

Rice Cooker (Model SRMK10 )

135

National Fuzzy Rice Cooker (SR-

MM10NW, SR-MM18NW)

Zojirushi Neuro Fuzzy Logic

Rice Cooker (Model NS-MYC- 18,
NS-JCC-18 )

136
 Components of a Fuzzy Rice
Cooker

137

The diagram shows the basic components

of a rice cooker.
The main component is the rice container
and is heated directly by a heating coil.
The heat sensor, which reads the
temperature of the rice container, sends
the reading as an input to the fuzzy
controller.
The fuzzy controller then control the
heating coil to obtain the desired heating
of rice.

138
 Input/Output Variables

Two Input Variables

1. Amount of Rice
2. Temperature of Rice Container

One Output Variable

Rate of Heating
(Change in the control signal)

139

Amount of rice in the container is not read by a

weight sensor since it is not economical.
If the same amount of power is supplied to the
heating coil, the temperature rise time and rising
rate will vary according to the amount of rice in the
rice container. With this basic concept, the amount
of rice can be estimated.
The temperature of the rice container is obtained
from a temperature sensor.
The rate of heating is controlled by sending the
correct control signal to the heating coil.

140
 Fuzzification
(Membership Functions of the I/O Variables)

Amount of Rice (Input)

LOW MEDIUM HIGH

1

0.8

0.6 LOW
MEDIUM
0.4 HIGH

0.2

Amount of Rice
in Container

141

Temperature of Rice Container (Input)

LOW MEDIUM HIGH
1

0.8

0.6 LOW
MEDIUM
0.4 HIGH

0.2

0
Temperature in
Container
Rate of Heating (Change in the control signal) (Output)
LOW MEDIUM HIGH
1

0.8

0.6 LOW
MEDIUM
0.4 HIGH

0.2
Rate of
0
Heating 142
 Design of Fuzzy Control Rules

IF amount of rice is HIGH

AND temperature of rice container is LOW
THEN rate of heating is HIGH.
IF amount of rice is LOW
AND temperature of rice container is HIGH
THEN rate of heating is LOW.
IF amount of rice is MEDIUM
AND temperature of rice container is MEDIUM
THEN rate of heating is MEDIUM.

143

Fuzzy Rule Matrix

Amount of Rice

L M H
Temperature
of L M H H
Rice
Container
M M M M
H L L M
144
 When is the Rice Cooked?

A sudden surge of temperature will occur at the

end of the cooking process if substantial amount
of water in the rice cooker has evaporated.
The rice is assumed to be cooked when the
sudden surge of temperature is detected.
It will then switch itself into the keep warm mode.

145

Fuzzy vs. Traditional Rice Cooker

A fuzzy rice cooker is so much faster than a
traditional rice cooker. It can reduce cooking
time to approximately 20 minutes.

A fuzzy rice cooker is not a cooker anymore, it is

a combination of cooker, steamer and warmer. A
traditional rice cooker could serve only as a
cooker.

A fuzzy rice cooker also offers a variety of

cooking possibilities which can not be achieved
by traditional rice cookers, such as the cooking
of porridge and sushi rice.
146
 Advantages of Fuzzy Rice Cooker

Fuzzy controlled rice cooker

offers a different rate of heating
at different stages of the cooking
process rather than just the
same heating rate for the whole
Traditional Rice Cooker cooking process.
This enable more even heating for smooth and
fluffy rice.
Fuzzy controller also makes continuous keep
warm feature possible.
147

Disadvantages of Fuzzy Rice Cooker

The capacity of a fuzzy rice cooker is usually

smaller than conventional rice cooker

Fuzzy models found in the market are usually

5 cups and 10 cups while conventional
models usually have models up to 33 cups.

The price of a fuzzy rice cooker is higher then

the conventional rice cookers.

148
 Conclusions
Although a fuzzy rice cooker is more
expensive then conventional models, the
features and benefits provided by it will
still worth the price.

This new trend of fuzzy rice cooker design

which will eventually take place in the
market and replace the conventional
models.

149

References
Fuzzy Logic Control - Module 2 & 3 (Prof.
Marzuki Bin Khalid)
http://www.sanyo.com/
http://www.howstuffworks.com/
http://www.amazon.com/
http://www.millionbuy.com/
http://www.appliancesource.com/

150
 FUZZY VACUUM CLEANERS

151

INTRODUCTION

Many products designed to make life and work easier emerged from the
(American) Industrial Revolution. Vacuum cleaner is one of the product.
Vacuum cleaner is a mechanical device using a draft of air to re move dust, loose
dirt, or other particulate matter from dry surfaces.
A vacuum cleaner that is able to do the cleaning based on different characteristics
of floor surfaces will be very desired for todays market.
This new requirement actually represents the first and most visible group of the
next generation of consumer products based on a new control-fuzzy logic.
Fuzzy logic is a relatively new technology that enables machines and products to
operate more efficiently and independently by processing information similar to
the way people do.
For many years, fuzzy logic has been an attractive idea to designers of industrial,
consumer and automotive products.
Fuzzy logic vacuum cleaner makes both easier holding and efficient cleaning
possible.
152
 THE EARLY VACUUM CLEANER
It was made in France at the beginning of the 20th century.
The trade mark is ATOM, produced by S. Mestitz & Fils, Raudnitz
Bokime.
The first vacuum cleaner had to be operated manually. Two
persons were needed for this: one to operate the bellows and the
other to move the mouthpiece over the floor. The dust was blown
into the air.
It consists of a round red metal cask, a sucker with a bar
attached to a kind of crankshaft, and a driving wheel. It had very
few accessories! THE FIRST
VACUUM CLEANER
This principle was the basis of a machine with a pair of bellows
that was operated by hand.
Vacuum cleaners became successful after the application of
small electric motors, which were used to drive a fan.

153

GENERAL FUNCTION
OF
A VACUUM CLEANER

It is especially useful on highly textured surfaces, such as carpets and

upholstery,that are difficult to clean by wiping or brushing.
Usually, an electrically powered fan is used to produce a zone in which the air
pressure is below atmospheric pressure, causing a draft of air to flow through
the material to be cleaned, carrying the small particles with it.
The draft passes through a filter bag which traps the particles, and the flow of
air is then discharged back into the atmosphere.
In some machines the electric motor and wiring are sealed so that wet surfaces
can be cleaned safely.
154
 EXAMPLE OF FUZZY VACUUM
CLEANER
This note describes a universal motor power control implemented
on a standard microprocessor running software using the fuzzy logic
concept. The different stages of development of the motor power control
for a vacuum cleaner are described with the ST5220 microprocessor and a
fuzzy logic development tool, the fuzzyTECH ST52 Explorer Edition.

Typical connection
scheme with ST52
Copyright 2001
STMicroelectronics
155

FUZZY PRODUCT IN MARKET

X204: Hand Vacuum Cleaner

Unique 600W Motor offers the most powerful
Suction available in a hand vacuum.
Includes two accessory brushes which connect
directly to the unit or can be used with the flexible hose .
With convenient shoulder strap.
Convenient dust cup design
X205: Rechargeable vacuum Cleaner
No replacement bags required.
Cordless for freedom of movement
Unrestricted cleaning without any loss of suction
power.
Compact, vertical storage system with convenient
accessory storage
With a small brush nozzle for sofas, stairs, etc.
With a long head for reaching difficult places
With a permanent, washable filter which is easy to
clean 156
 THE FUZZY VACUUM CLEANER
WINS 1998 WORLDSTAR
PACKAGING COMPETITION AWARD
WINNERS

A new design for export vacuum cleaners

was created to fit the various changes in
distribution and warehouse environments. The
brand name "Fuzzy" represents the union of
humans and technology. This package
comprehensively displays special characteristics and
POP functions of the
products on the package, which enables the
consumer to make a purchase decision without
actually seeing the product.

157

AI ASPECT : DUST SENSOR

A conceptual cross-sectional diagram

of a fuzzy vacuum cleaner in shown in
diagram on the right.

Construction of a vacuum cleaner

An enlarged view of the users

operation section is shown in diagram
on the right side.

Enlarged view of the handle

158
DUST SENSOR : CONTROL CIRCUIT

A block diagram of the control circuit is shown below. Dust on the

floor is sucked up by the force of the motor - driven inside the main unit. It
then passes through the dust sensor in the handle and is collected in a paper
bag in the main unit.

Control circuit

159

DUST SENSOR : OPERATION

The dust sensor is inside the pipe handle on which the controls are
located. A photo transistor is mounted opposite a light emitting diode. When
dust passes through, some infrared rays are cut off causing the amount that
reach the photo transistor to decrease. The varying component is extracted and
amplified, then a waveform is formed and converted to pulses. The amount of
dust is judged by counting these pulses. If the surface of the dust sensor
becomes dirty, the transmissivity of infrared rays will decrease, but by detecting
the differentiated change the measurement can be made insensitive to the dust.
In addition, the sensor has the configuration described below so that no
problems arise in practice.
The air path where the dust sensor is installed is partially restricted to change
the air flow so that less dust will adhere to the sensor
The light emitting diode and photo transistor are encased in hard transparent
acryl holders to protect them from scratching
The dust sensor is installed near an opening so that it can be cleaned

160
 DUST SENSOR : OPERATION
The time variation of dust amount that can be detected by this dust sensor
is shown in diagram below.

As cleaning proceeds the amount of dust decreases, but whereas this decrease is
fast in the case of a tatami mat or wooden floor, it is slow in the case of a wool
carpet. This is because it is easy for dust to be picked up off a tatami mat or
wooden floor but hard for it to be picked up off of a wool carpet. Thus, by
measuring the time variation of the rate at which dust is picked up., the floor
surface material can be judged.
161

DUST SENSOR : FUZZY INFERENCE

When cleaning is done, if the suction power is too strong the suction
nozzle will stick to the floor, making it difficult to handle. Conversely, if it is
too weak it will be hard to pick dust up. For this reason, there is an optimum
suction strength for a given floor material and amount of dust. In order to
optimize the tradeoff relationship between suction power and ease of handling,
providing a vacuum cleaner that is easy to use, the suction power id determined
using fuzzy inference. Fuzzy inference is performed as shown in the diagram
below. The suction power is determined from the cumulative amount of dust
detected in a unit of time and the floor surface condition.

Fuzzy Inference

162
 FUZZY INFERENCE : RULE BASE
Inference is performed with 15 rules such as the following.

If there is a great deal of dust and the floor surface has the properties of
a wool carpet, make the suction power very strong.

The fuzzy rule table is shown in the figure on the right. P1 to P15 in the figure
are levels of suction power.

Fuzzy Rule Base

163

FUZZY INFERENCE : MEMBERSHIP

FUNCTION
The membership functions of the fuzzy vacuum cleaner is shown in the figure below.

Membership functions
164
 FUZZY INFERENCE :INPUT OUTPUT
GRAPH
o Simplified fuzzy inference, with real
number type operation parts is used in
order to simplify the inference
operations, reduce the amount of
memory used and simplify the parameter
tuning
o Rule creation and parameter tuning are
done by cut and try based on data
from monitoring in ordinary households
and the developers cleaning know-how
o A diagram of input / output
characteristics, showing the results of
fuzzy inference after tuning is shown
beside

165

CHARACTERISTICS OF A FUZZY
VACUUM CLEANER

By using fuzzy inference, the optimum suction power for a

given amount of dust and floor surface material cab be determined,
eliminating the problems caused by a suction power which is too
strong so that the suction nozzle sticks to the floor, or, conversely, a
suction power that is too weak so that some dust is not picked up. In
addition, the completion of cleaning is indicated by a lamp on the
control section, reducing the incidence of incomplete cleaning in
places where it is difficult to see such as deep down in the wool hairs
and under beds.

166
FUZZY VACUUM CLEANER DIFFER
FROM CONVENTIONAL TECHIQUE

The objective of implementing fuzzy logic in conventional

vacuum cleaners corresponds with the necessity of vacuum cleaners
having to perform on different surface materials. Conventional
systems do not possess the capabilities of self adjustment according to
surface thus compromising performance. Fuzzy logic controlled
systems are capable of handling surfaces such as hard flat floor,carpets
with and without fur and wooden surfaces. Anybody can use fuzzy vacuum
cleaner easily, and moreover it picks dust up completely. It can be operated
easily without feeling resistance from the floor surface. It leaves the user with
a satisfied and secure feeling of having cleaned the floor completely. It is
quiet so that cleaning can be done even at night.

167

ADVANTAGES OF FUZZY LOGIC

An Alternative Design Methodology Which Is Simpler, And Faster :
i) Fuzzy Logic reduces the design development cycle
ii) Fuzzy Logic simplifies design complexity
iii) Fuzzy Logic improves time to market

A Better Alternative Solution To Non-Linear Control

i) Fuzzy Logic improves control performance
ii) Fuzzy Logic simplifies implementation
iii) Fuzzy Logic reduces hardware costs : reduce the sensor cost by using a
fuzzy model (Virtual sensor)
168
 ADVANTAGES OF
FUZZY VACUUM CLEANER

acoustic noise reduction

energy saving

automatic control of the air flow

automatic regulation of the aspiration force

monitoring bag status (empty, full)

169

REFERENCES
1) http://e-w.motorola.com/brdata/PDFDB/docs/AN1843.pdf

2) http://www.aptronix.com/fide/whatfuzzy.htm

3) http://www.aptronix.com/fide/howfuzzy.htm

4) http://staff.ncst.ernet.in/~sasi/fuzzy/main.html#method

5) http://www.aptronix.com/fide/howdofuzzy.htm#Fide%20helps%20you%20boost%20y
our

6) K.Hirota (Ed.), Industrial Applications of Fuzzy Technology, Springer Verlag, Tokyo,

1993.

170
 A FUZZY AIR-CONDITIONER

171

Introduction to the Fuzzy Air-

Conditioner System

Air conditioning systems are essential in most of

our daily lives.

Our expectations of such systems have been raised

to demand more than just temperature control, and
it is increasingly desirable to apply these systems in
varying situations and environments.

This fuzzy air-conditioner system serves to replace

the conventional systems for comfort and
simplicity.

172
 Introduction to the Fuzzy Air-
Conditioner System

The system is designed to control the temperature

as well as the humidity of the environment.
It is also known as the modified fuzzy air-
conditioner, due to its birth after the basic fuzzy air-
conditioner was introduced to the consumer
market.

Air
Fuzzy Air-cond ROOM

Temperature and humidity level detected by

sensors
173

The Fuzzy Air-Conditioner Design

Basically the design components consists of ;

2 Sensors Cooling Valve

2 Inputs 3 Outputs

Heating Valve Humidifying Valve

174
 Component Function in the Design

2 Sensors are required in the design, 1 sensor to

detect temperature variations and 1 sensor to
detect humidity/moisture level of the environment.

3 Control Elements: cooling valve, heating valve,

and humidifying valve, are required to adjust the
temperature and humidity of the air supply.

2 Inputs to the controller are measured

temperature and humidity.

3 Outputs generate control signals that is

feedback to the 3 valves for adjustments.

175

Structure of the Fuzzy Control System

Error
Fuzzy Controller
Output Air from
Cooling Valve
the Fuzzy Air-
Defuzzification

Air Supply +
Fuzzification

Rule Heating Valve Conditioner

- Base Humidifying Valve
Fan

temperature
Sensors
humidity

176
 Fuzzy Logic Control System

The input variables are transformed into fuzzy sets by

the fuzzification process. TEMPERATURE
cold
cool
warm
hot

HUMIDITY
low
medium
high

The fuzzy system outputs the optimal air circulation level by adjusting the
cooling valve, heating valve and humidifying valve in the air conditioner. The
combined results are sent to the fan to be propagated to the environment.The
amount of the valves openings depends on these output variables:
zero
low
medium
high
177

Membership Functions

TEMPERATURE
t
cold cool warm hot

Area weather
condition

h HUMIDITY
low med high

Area weather
condition
178
 Rule Base System

Rules for this controller can be formulated by adding rules

for humidity control to those already formulated for
temperature control in the basic model.

Additional rules must take the interference between

temperature and humidity into account.

For example, in the winter, when we use heat to raise

temperature, humidity is usually reduced. The air thus
becomes too dry. To address this condition, a rule statement
similar to the following is appropriate: (Note: Cooling valve is
abbreviated with CV, heating valve with HV and humidifying
valve with UV )

179

Rule Base System

IF temperature is cold AND humidity is high THEN CV is zero AND HV is high AND UV is low
IF temperature is cool AND humidity is high THEN CV is low AND HV is med AND UV is low
IF temperature is warm AND humidity is high THEN CV is med AND HV is low AND UV is low
IF temperature is hot AND humidity is high THEN CV is high AND HV is zero AND UV is low

IF temperature is cold AND humidity is med THEN CV is zero AND HV is high AND UV is med
IF temperature is cool AND humidity is med THEN CV is low AND HV is med AND UV is med
IF temperature is warm AND humidity is med THEN CV is med AND HV is low AND UV is med
IF temperature is hot AND humidity is med THEN CV is high AND HV is zero AND UV is med

IF temperature is cold AND humidity is low THEN CV is zero AND HV is high AND UV is high
IF temperature is cool AND humidity is low THEN CV is low AND HV is med AND UV is high
IF temperature is warm AND humidity is low THEN CV is med AND HV is low AND UV is high
IF temperature is hot AND humidity is low THEN CV is high AND HV is zero AND UV is high

180
 Rule Base System
temperature
cold cool warm hot
low zero low med high
humidity medium zero low med high
high zero low med high

Cooling valve (CV)

temperature temperature
cold cool warm hot cold cool warm hot
low high med low zero low low low low low
humidity medium high medium med
med low zero med med med
high high med low zero high high high high high

Heating valve (HV) Humidifying valve (UV)

181

OPERATION

RULES INTERFERENCE COMPOSITIONS DEFUZZIFICATION

182
 RULES

Take the interference between temperature & humidity

When heat is used to raise temperature, humidity is usually

reduced

When air becomes dry :

If temperature is LOW then open humidifying valve

SLIGHTLY
Act as predictor of humidity & also designed to prevent
overshoot in the output humidity curve
183

INTERFERENCE

Suppose the air conditioning system has measured

temperature & moisture levels & mapped them to the fuzzy
value of 0.7 & 0.1 respectively

The system need to infer the truth of each fuzzy rule

A myriad of difference inference technique exist but the

simplest method is called the MAX-MIN

184
 COMPOSITIONS

In MAX-MIN method, it uses the max fuzzy value of

inference conclusion as the final conclusion

Select a fuzzy value 0.7 since this was the highest

fuzzy value

185

DEFUZZIFICATION

Convert the fuzzy output 0.7 into one of the air

circulation level

Fuzzy output of 0.7 indicates that the circulation should

be set to HIGH

186
 DIFFERENCES

CONVENTIONAL
CONVENTIAL METHOD FUZZY FUZZY
METHODLOGIC
A less flexible control method An alternative design methodology which is
simpler & faster
non-linearity Reduces the design development cycle

Dead time Simplifies design complexity

External disturbances Improve time to market

Does not result in the satisfactory temperature A better alternative solution to non-linear
control control
Improves control performances

Simplifies implementation

Reduces hardware costs

187

ADVANTAGES

HANDY TOOL

SIMPLIER ACCUARATE RESULTS

USER FRIENDLY CUT COST

CUT TIME

188
 Temperature control in an automobile passenger environment is more
complex than that of a static room in a building.

It is critical to keep windows from being fogged, which is caused by a

temperature differential between inside and outside air in combination
with the interior humidity.

The strength of sunshine radiation and the automobile speed must also
be factored in.

Fuzzy controller which employs five sensors to obtain data for

temperature control and humidity control in an automobile.

It prevents rapid change of temperature in the car when doors or

windows are opened and then closed.

It even reacts to weather changes because interior humidity changes

caused by the weather can be detected by sensors.
189

Commercial Example #2

3.6.2 Canons fuzzy auto-focus camera

This camera was marketed in 1989.

It uses a 4-bit microcontroller with 500-byte of memory.
Fuzzy logic is used to determine the object of focus by evaluating
the field of view and controlling the autofocus mechanism to
focus on the object.
Earlier cameras used the object centred in the field of view as the
desired focus.
This sometimes led to error as in the case of two objects
presented off-center.
This problem had to be solved by the photographer, who focused
on one object, locked the auto-focus, and then re-oriented the
camera to get the desired shot.
This manual process is laborious and awkward to the
photographer.

190
Range of
the Canon
family of
cameras

191

To overcome the problem, fuzzy reasoning was introduced in cameras.

First, distances to 3 points in the field of view are measured. Using these
locations and the relationships between them, fuzzy logic decides the desired
focus point.
Using these locations and the relationships between them, fuzzy logic decides
where the desired focus lies and then focuses on that point.
In the Canon auto-focus camera the fuzzy rules were obtained by an analysis
of 288 photographs taken by 8 people.
Examples of some rules are given in the next slide.
One example of the rules is as follows:
If the object is near the Left
then the plausibility for object to be at Left is Very High.
Or simply,
If L is Near Then Pl is High
(Plausibility means the chance it will happen.)

192
 Example of 5 rules for the
Canon Autofocus camera

193

Example of how rules are fired

Compare figure (a) and (b) below:
We see that figure (a) has main subject on the left and figure (b) has main
subject at the center.
However in both cases, the relationship is L<C<R and both satisfy Rules (b)
and ( e) as given in the rule base in the previous page.
In this case, the decision depends on the values of L, C, and R and this
comparison is done with the help of membership functions.
The actual rule will be fired depending on the highest degree of the
membership function.
It is very difficult for binary logic rules to model this situation which may
require a large number of rules.
However, a few fuzzy rules can easily
deal with this problem.

194
 Conclusion
The performance of this method has been evaluated by using 288 pictures
taken by 8 persons.
The percentage of correctly focused pictures, with and without fuzzy rules are
given in the Table below.

Advantages of the fuzzy auto-focus camera:

Reduce the amount of required memory.
Rules are acquired from analysis of about 300 photographs.
Simplicity of usage as compared to conventional auto-focus technique
Results show an improvement of 23% in auto-focusing

Method Focusing rate

Three measured 96.5%

distances + fuzzy

Distance to the center

(conventional) 73.6%

195

Commercial Example #3

3.6.3 Minolta fuzzy auto-focus, auto-

exposure and auto-zoom camera

Minolta Camera Co. Ltd., Japan uses fuzzy logic to combine the 3
mechanisms of focusing, zooming and deciding exposure
automatically in its cameras since the early 1990s.

The figure in the next slide shows the mechanism of a typical Minolta
fuzzy camera with the above facilities.

A number of fuzzy modules make up the auto-focus, auto-exposure

and auto-zoom facilities in the camera.

196
 Auto-focus, auto-exposure and auto-zooming
mechanism of a Minolta Camera, Japan.

FS= Fuzzy
logic module

197

Operation Principle of the Minolta Camera

Auto-Focus Module
The fuzzy auto-focusing mode has 6 distance distributions to do the
fuzzy reasoning to locate the main subject, which are obtained by pre-
processing the outputs of 4 auto-focus sensors, lens information and
1 sensor that detects the camera position.

7 fuzzy rules, obtained from the analysis of approximately 1000

pictures, determine the location of the main subject to focus on.

By adding the fuzzy logic module for decision making leads to an

improvement of 15% in the focus hit rate.

198
 Auto-Exposure Module
To implement auto-exposure, fuzzy reasoning is used to determine
exposure value and the best combination of shutter speed and
aperture, depending on the type of scene being photographed.

The exposure value is determined by 3 fuzzy inference modules, using

brightness values obtained from 14 zones in the field of view and the
position of the main subject (which is determined by the above auto-
focus mechanism).

The first fuzzy module uses the difference in brightness between the
main subject and the background to give an output which is a
measure of the amount of back-lighting present.

199

The second fuzzy module decides whether the exposure is to be

focused only on the main subject or the entire scene.

The third fuzzy module uses the outputs of these two fuzzy modules;
weighs 3 measures of exposure i.e. at average, at center and at the
main subject, and then it outputs the final exposure value.

The optimal combination of the shutter speed and aperture is

determined by fuzzy inference using the type of scene and lens being
used (see example in the next slide).

The type of scene, for example snap, portrait, close-up, or natural

scenery, is determined by the distance to the main subject.
In a scenery shot the depth of field increases.

Using fuzzy inferencing techniques, fine control according to scenery,

lens charatcteristics, etc. can be achieved.
200
 Auto-Zooming Module
To implement auto-zooming, fuzzy reasoning is used to decide
the speed at which to zoom the lens.
When the main subject moves, the size of its image is held at
constant value by zooming appropriately to compensate for the
movement.
Fuzzy reasoning chooses the zooming speed by looking at the
ratio of current lens magnification to that 1 unit time ago and
its rate of change.
The rules change the speed of the lens, depending on how the
object moves.

201

Basic concepts of the inference rules for

auto-exposure of the Minolta camera

Antecedents

KEY
L= Large, S = small, B=Big Consequents

202
 Blood Pressure Meter

Designed by National, Matsushita Electric Co., Japan

By using fuzzy logic technology, measurements taken are
more reliable and accurate
Measure both systolic and diastolic blood pressure

203

Importance of Maintaining Normal Blood

Pressure

What is blood pressure?

Blood pressure is the pressure exerted by blood
pumped from the heart on the walls of the
blood vessels.
Systolic pressure is the pressure exerted when
the heart contracts and pumps blood into the
arteries, while diastolic pressure is the pressure
exerted when the heart expands.

Blood pressure can vary according to a variety of

factors including age, sex, and physical
condition.
It is lower during sleep and higher when 204

working.
 Abnormal blood pressure can indicate a number of
illnesses such as:
Cardiovascular problems
Hypertensions
Diabetes
Liver disorder
Blood pressure will be higher than usual when:
Excited or tense
Taking a bath
Exercising
Cold
Immediately after eating
Smoking tobacco, drinking coffee
Blood pressure will be lower than usual when
After drinking alcohol
After taking a bath

205

Fuzzy Logic measurement

in the Blood Pressure meter
Blood flow in the veins give pulse-
type waveforms. This allows
measurement of blood pressure
possible using pulse meters such
as these.
A number of factors affect blood
pressure measurements such as:
Condition of the blood vessels
Structure of arm

Thus making using conventional methods difficult.

National employs fuzzy logic technology which double
checks the strength and shape of the pulse wave and then
determine the blood pressure.
206
 Fuzzy logic inference engine
Pulse wave

Pulse wave Shape

strength

Pressure
Pressure Time

Pulse wave shape and its Pulse wave strength and

rate of change its rate of change

Fuzzy Inference

Determination of Actual Blood Pressure

207

Advantages of Fuzzy logic

Measurement reliability has been improved for

people with the following problems whose blood
pressure was difficult to measure:
People with weak pulse
People with irregular pulse
People with unstable pulse
Error in measurement is detected

208
 Industrial Example #4

3.6.4 The Sendai Fuzzy Logic Subway System

First proposed in 1978 by Hitachi Ltd.

Granted permission to operate in 1986 after 300,000 simulations and
3,000 empty runs
Improved stop position by 3X
Reduced power setting by 2X
Total power use reduced by 10%
Finally, Hitachi was granted contracts for Tokyo Subway in 1991

209

To measure the train performance, the following Performance Indices

can be used:

Traceability
Comfort
Safety
Running Time Margin
Stopping Accuracy

Membership functions
210
 211

The train has 2 fuzzy components:

(1) speed control and (2) stopping control.

Constant Speed Control (CSC)

In this mode, trains are started and run at predetermined target speed.

Train Automatic Stopping Control (TASC)

In this mode, trains are stopped in a predetermined target zone of a
station with high accuracy.

Each component has its own fuzzy inference and rule base.
Rules are developed based on the following procedures:
Step 1 : Provide a description of the typical operating
methods used by the operators
Step 2 : Define the performance indices of the system
Step 3 : Define a model for predicting
Step 4 : Convert expert human operating methods into control rules

212
 Example of the fuzzy rules

For the CSC

If the control command is Not Changed
And target speed is followed Good
Then Command should Not be Changed

If control command is changed to Zero

and coasting and riding comfort is Good
and the target speed is followed Very Good
Then Command should be changed to Zero

For the TASC

If control command is changed to Seventh step of braking
and safety against overrun is Very Bad
Then Command should be changed to Emergency Braking

213

Comparison of the train acceleration between the fuzzy

logic computer controller and a human controller.The
fuzzy controller shows a more optimal response.

214
 Comparison of the Stop Gap Measure of
the fuzzy controller and the PID controller

215

Comparison between the fuzzy system and manual operation

of the Sendai Subway operating from one station to another.
It can be observed that the fuzzy system is more optimal.
216
 Fuzzy Train Control by Hitachi Ltd.

217

Summary of the Sendai Fuzzy Train

Fuzzy inference has been used with success in this project.

A number of distinct advantages are present in the fuzzy
train system over conventional train system such as riding
comfort, stopping performance, power reduction, etc.
Improved overall performance.
Human reasoning can be accommodated rather easily.

218
 3.7 Summary of Module 3

A brief review of control system basics have been discussed for a

better understanding of fuzzy logic purpose in control
applications.
Specific case studies of fuzzy logic control applications have been
discussed in this module.
Several commercial examples have also been discussed.

219