1

A Table Lookup Scheme for Fuzzy Logic

Based Model Identification Applied to Time
Series Prediction
Prof. M Farooq
Department of Electrical & Computer Engg.
Royal Military College of Canada

Khatoon Shahida & Prof. Ibraheem

& Prof. Moinuddin
Department of Electrical Engg.
Faculty of Engg. Tech., J.M.I,
New Delhi-25,India
[email protected]

Kingston, Ontario K7K7B4

farooq@rmca

Abstract - Zadeh introduced Fuzzy sets in 1965 to

represent and manipulate data and information that possess
nonstatistical uncertainty. Since this date, fuzzy logic has
been applied to many fields such as industry, medicine,
economics and so on. The reason for this rapid growth in
the use of fuzzy logic worldwide is that fuzzy logic provides
an appropriate mechanism to describe the static and/or
dynamic behavior of complex physical systems that are
difficult to yield their conventional mathematical models.
We can consider a fuzzy set as a fuzzy model of human
concept. In this paper, we consider fuzzy modeling as an
approach to form a system model using a description
language based on fuzzy logic with fuzzy predicates. This
paper presents a general approach to modeling and
identification of dynamic systems based on fuzzy logic. A
table lookup scheme is presented to generate fuzzy rules
from numerical data. This method determines a mapping
from input space to output space based on the combined
fuzzy rule base using defuzzifying procedure. Application to
time series prediction problem is also presented.
Key words: Fuzzy Logic, Identification model, Time series,
fuzzy inference engine, Table-lookup scheme, COG method
of defuzzification

Introduction

Fuzzy modeling is the most important issue in fuzzy theory.

We can consider a fuzzy set as a fuzzy model of a human

concept[1]. In this paper, we consider fuzzy modeling to be

an approach to form a system model using a description
language based on fuzzy logic with predicates. We can
interpret the fuzzy modeling as a linguistic modeling
scheme by which we linguistically behavior using a natural
language [2]. The fuzzy modeling is a system description
with fuzzy quantities. Fuzzy quantities are expressed in
terms of fuzzy numbers or fuzzy sets associated with
linguistic labels. Therefore, the relation between input and
output variables can be viewed as a set of fuzzy logical rules
or fuzzy-set associations. Since functional variables are
stored in a distributed rule-based fashion, the value of the
function at any point in the input space is derived by
aggregating the consequences of fuzzy logical rules. It has
been shown that fuzzy systems are capable of approximating
any real continuous function to any desired degree of
accuracy [25-26].
Fuzzy modeling can be applied to various domains
of problems such as signal processing, pattern recognition,
and control [3-11]. The main problem of fuzzy control is to
design a fuzzy controller where we usually take an expertsystem-like approach [13]. That is, we derive fuzzy control
rules from a human operators experience and /or engineers
knowledge, which are mostly based on their qualitative
knowledge of an objective system. There are three other
approaches to generate the rules. Usually, the design
procedure is thus something like the following: first, we
build linguistic control rules; second, we adjust the
parameters of fuzzy sets by which the linguistic terms in the
control rules are quantitatively interpreted.

1449

2
Apart from fuzzy control, the studies of fuzzy
modeling are divided into two groups. The studies of the
first group deal with fuzzy modeling of a system itself or
fuzzy modeling for simulation. The studies of the second
group deal with fuzzy modeling of a plant for control. Just
as with the modern control theory, we can design a fuzzy
controller based on a fuzzy model of a plant if a fuzzy
model can be identified [11,12].
Let us consider the problem of modeling.
Modeling is classified from the view point of a description
language. However, there are some common problems to be
solved in modeling independently of both the description
language and data type. Thus we refer to the systems theory.
The most complicated problems arise when we take a black
box approach to modeling. In the black box approach, we
have to build a dynamical model using only input-output
data. This stage of modeling is usually referred to
identification.
From the afro-mentioned, system identification
involves finding a model that may be regarded as equivalent
to the objective system with respect to input-output data.
The identification for the fuzzy model has two aspects as
usual: structure identificationa and parameter identification.
This problem will be discussed in general in section 3.0.

Generation of Fuzzy Rules

process to be controlled and his general control engineering

sense.
There are some disadvantages in this method of design.
Firstly, sometimes an operator cannot tell linguistically what
actions he takes in such and such situation. Secondly, when
a process is complex, it is difficult to design a fuzzy
controller even from a control engineers point of view.
Another point to note is that this method is heuristic in
nature, and very difficult to be generalized.

2.2

Based on Modeling the Operators

control Actions

When an operators skill is important, it is very

useful to derive fuzzy control rules by modeling the
operators control actions [4]. For instance, consider when
we drive a car. We know car-driving techniques by our
hands and legs rather than by our brain [33-34]. If it is
possible to model an operators control actions in terms of
fuzzy implications using the input-output data concerned
with control actions, we can use the obtained model as a
fuzzy controller.
In this method, it is not difficult to find the input variables
of the model by asking an operator what kind of information
he uses in his control. Also we can find approximately the
number of fuzzy subspaces of the input space.

2.3

Based on the Fuzzy Model of a Process

The two previous methods for generating fuzzy

rules
work
well only in the case where an expert plays an
The most important topic to discuss about fuzzy
systems is the way to design and derive fuzzy rules. A important role in controlling the process. The first method
number of studies have been devoted to this problem [3-4, is somehow based on a rough idea about the characteristics
9-11, 25]. There are, in general, four methods of design and of a process. For example, the output increases as the input
derivation of fuzzy rules [4, 13,31]. Sometimes, it seems does, the process has a time lag, and so on. The second
that a combination of them would be necessary to construct method uses only the necessary variables available to
an effective method for the derivation of fuzzy rules. Until control a process, to which an operator can refer. In fact this
now there exists no systematic method for the general controller works better than the expert does, since it does
design of fuzzy rules. The following paragraphs will not make errors. However, if we cannot rely on an operator
and if we want to have better results than any operator gives,
explain these methods.
have to find another method of design.
2.1
Expert
Experience
and
control we
In the linguistic approach, the linguistic description of the
Engineering Knowledge
dynamic characteristics of a controlled process may be
Most fuzzy controllers have been designed so far viewed as a fuzzy model of the process. By fuzzy
by referring to human operators experience and/or control modeling, we mean to represent the characteristics of a
engineers knowledge [23]. In fact we can say that fuzzy process by a set of its fuzzy behaviors which are also
control was the first practical application of expert systems expressed by using fuzzy implications concerned with
The format of fuzzy implications clearly makes them inputs, state variables, and outputs.
suitable as a descriptive language to express an experts This method is somewhat more complicated, it yields better
thinking, which is essentially fuzzy in its nature. In our performance and reliability. However, this method to the
daily life most of the information on which our decisions are design of fuzzy rules has not as yet been fully developed.
based is linguistic rather than numerical in nature. Seen in Figure 1 shows an idea for a fuzzy controller consists of a
this perspective, fuzzy control rules provide a natural set of control rules and the fuzzy process model consists of
framework for the characterization of human behavior and set of process behaviors.
decisions analysis. Many experts have found that fuzzy
control rules provide a convenient way to express their
F u zz y
domain knowledge [32].
C o ntro
In many cases where an operator plays an important role in
C o ntro l
P ro cess
R u le s
process control, it is very useful to find his know-how on
control by interviews and to express it in terms of fuzzy
implications. It is also possible for a control engineer to list
a number of protocols based on his knowledge about a
Fig. 1 Fuzzy Controller Based on Modeling a Process

1450

3
a system which affect the output. In this case we select a
finite number out of a finite number and so there are some
systematic ways to solve this problem. In a conventional
black box approach in systems theory, this type of
identification is, however, not explicitly discussed, and
models are based on pre-assigned input-output variables.
The structure identification of type II, which is
concerned with input-output relations, is divided into the
subtypes II(a) and II(b). In II(a) , we have to determine the
number of fuzzy rules in a fuzzy model. By structure
identification in ordinary systems theory, what we mean is
2.4
Based on Learning:
Many FLCs have been built to emulate human to find the relations between the inputs and the outputs.
decision-making behavior, but few are focused on human Given a description language for modeling, it means the
learning [25,36], namely, the ability to create fuzzy control determination of the order of a model. Whereas in fuzzy
rules and to modify them based on experience. The Self model, the structure identification of this kind is stated in a
Organizing Controller (SOC) [37] has a hierarchical different way. A fuzzy model consists of a number of ifstructure which consists of two rule bases. The first one is then rules. The number of rules in a fuzzy model
the general rule base of an FLC. The second one is corresponds to the order in a conventional mode.
The identification II(b) implies determining how
constructed by meta-rules which exhibit human-like
learning ability to create and modify the general rule base the input space should be partitioned. There are two parts of
an if-then rule: the premise part and the consequent part.
based on the desired overall performance of the system.
So the rules have two structures: the premise structure and
the consequent structure. Type II(b) deals with premise
3
Identification in Fuzzy Modeling
structure. The premise space of the input variables of fuzzy
The identification is divided into two kinds: model is partitioned into several fuzzy subspaces (fuzzy
structure identification and parameter identification. sets); where the number of rules corresponds to the number
Structure identification can be divided into two types: type I of subspaces. This problem is combinatorial, therefore we
and type II, where each type is also divided into two need a heuristic method to find an optional partition together
subtypes, a and b. Figure 2 illustrates the classification of with some criterion, i.e. output error.
identification.
3.2
Parameter Identification:
In ordinary system identification, parameters are
Identification
the coefficients in a functional system model. In a fuzzy
model, the parameters are those in the membership
functions of the fuzzy sets. There is not a big difference
Parameter Identification
Structure Identification
between the two except in the number of the parameters,
there being many more in a fuzzy model. The structure
identification and the parameter identification cannot be
Structure Identification
Structure Identification
separately performed in principle. This fact makes the
identification very complicated.
However, in some
b: Input variables
A: Input Candidates
a: Number of rules
b: Input Space
approaches, the parameter identification can be separately
performed after the structure identification [11].
Figure 2
Classification of Identification
4
Fuzzy Model Identification Using
There are two ideas for designing a fuzzy controller
based on fuzzy model. The first one is a heuristic method
[35] in which we set fuzzy rules to be compatible with
system behavior by considering the control objective. The
second idea is the method of identification to determine the
structure and the parameters of fuzzy rules model so that the
system with a controller satisfies the control objective
[3,9,11].
We will investigate this method through
simulation in section 3.4.

3.1

Numerical Data

Structure Identification

Generally speaking, the structure identification of a

system has to solve two problems: one is to find input
variables (type I), the other one is to find input-output
relations (type II). Type I consists of I(a) and I(b). In I(a),
we find possible input candidates for the inputs to a system.
There are of course an infinite number of possible
candidates, which should be restricted to a certain number.
This type of identification can not be solved in general.
Because there is no systematic way to find the exact causes
of an unknown phenomenon. We have to take a heuristic
method based on experience an/or common sense
knowledge.
In the structure identification of type I(b), given the
possible in put candidates, we find a set of input variables to

In this section, we will deal with fuzzy model

identification based on a black box approach. Because even
if we can find the local mechanisms of a system, it is often
the case that we cannot build the whole system model by
arrogating the local mechanisms.
In the black box approach, the dynamical model is
built using the input-output data of the plant. Existing
identification schemes determine a model for a system based
on the input-output pairs resulting from existing the system
with an input signal and measuring the corresponding
outputs. For many industrial system, there is another
important information source that is human knowledge, who
are familiar with systems and can provide linguistic
descriptions about the behavior of the system in terms of
vague and fuzzy words. Although these linguistic

1451

4
descriptions are not precise, they provide important
information about the system. For many process control
problems, a human operator can determine a set of
successful control rules based only on the linguistic
description about the process. Existing identification
schemes ignore this important source of information and
cannot incorporate the linguistic descriptions directly into
the identifiers. Even if there are some identification scheme
based on combination of linguistic description and
numerical information, they still combine these two sources
of information in ad hoc way for a specific problem;
simulations are then performed to show that this approach
works well for the specific problem. This kind of approach
has two weak points: 1) it is quite problem dependent, i.e., a
method may work well for one problem but is not suited for
another problem; and, 2) there is no common framework for
modeling, which makes theoretical analyses for these
approaches very difficult. The aim of this section is to study
and investigate identifiers of nonlinear dynamic systems
which combine both linguistic descriptions and input-output
pairs in a inform fashion (a fuzzy rule base) into their
designs. The method which we will use for identification is
called A Table-Lookup Scheme [25]. This method
performs a one-pass operation on numerical input-output
pairs and linguistic fuzzy if-then rules. The key idea of this
method is to generate fuzzy rules from input-output pairs,
collect the generated rules and linguistic rules into a
common fuzzy rule base, and construct a final fuzzy logic
system based on the combined fuzzy rule base.

center of the region and has membership value unity; the

other two vertices lie at the centers of the two neighboring
respectively, and have membership values equal to zero. Of
course, other divisions and other shapes of membership
functions are possible.

Step 2: Generate Fuzzy Rules from Given Data

Pairs
First, determine the degrees of given x1(i) , x2(i)and
y in different regions. For example, x1(1) ,in Figure 1.3 has
degree 0.8 in B1, degree 0.2 in b2, and zero degrees in all
other regions. Similarly, x2(2) in Figure 3 has degree 1 in CE,
and zero degrees in all other regions. Second, assign a given
x1(i) , x2(i) or y(i)to the region with maximum degree.
For example, x1(1) in Figure 1.3 is considered to be B1, and
x2(2) in Figure 1.3 is considered to be CE.
[x1(1) ( 0.8 in B1, max), x2(1) (0.7
( x1(1) , x2 (1) ; y(1) )
(1)
in S1, max); y (0.9 in CE, max)]
Rule 1 : IF x1 is B1 and x2 is S1 , THEN y is CE;
[x1(2) ( 0.6 in B1, max), x2(2) (1 in
( x1(2) , x2 (2) ; y(2) )
(2)
S1, max); y (0.7 in CE, max)]
Rule 2 : IF x1 is B1 and x2 is CE , THEN y is B1;
The rules generated in this way are and rules,
that is, rules in which the conditions of the IF part must be
met simultaneously in order for the result of the THEN part
to occur.
(i)

(x1)
S2
1

4.1 Identification based on a Table-Lookup

scheme:
Suppose we are given a set of desired input-output
data pairs:
(x1(1), x2(1) ; y(1)), (x1(2), x2(2) ; y(2)), (x1(3), x2(3) ; y(3)) (1.1)

B1

-1

B2

X (k)
-2

(x1)
S2
1

B2

S1

CE

B1

-1

X(k)
4

S1 CE

B1

B2

B3

Step 1: Divide the Input and Output Spaces

into Fizzy Regions
Assume that the domain intervals of x1, x2 and y are
[-x1 ,+ x1 ], [- x2, + x2 ] and [- y, +y], respectively, where
domain interval of a variable means that most probably this
variable will lie in this interval. Divide each domain interval
into 2N+1 regions (N can be equal or unequal), denoted by
SN (Small N), , S1( Small 1), CE (Center), B1 (Big 1), ,
BN (Big N), and assign each region a fuzzy membership
function. Figure 1.3 shown an example where the domain
interval of x1 is divided into five regions (N=2), the domain
region of x2 is divided into seven regions (N =3), and the
domain interval into five regions (N=2). The shape of each
membership function is triangular; one vertex lies at the

CE

0
-3.5

where x1 and x2 are inputs, and y is the output. This simple

two-input one-output case is chosen in order to clarify the
basic ideas of this approach extensions to general multiinput multi-output cases are straightforward. The task is to
generate a set of fuzzy if-then rules from the desired inputoutput pairs of (1.1), and use these fuzzy if-then rules to
determine a fuzzy logic system
f : ( x1, x2) y. This approach consists of the following
five steps :-

S1

-3.5

-2

(x1)

1452

S3

S2

1
X(k)
0
-3.5 -2.6

-1.7 -0.8

1.3 2.2

3.1

Step 5: Determine a Mapping Based on the

combined Fuzzy Rule Base
Figure 5

The defuzzification strategy which is used in this

method determines the output y for given inputs (x1, x2) as
follows: First, for given inputs (x1, x2), we combine the
antecedents of the ith fuzzy rule using product operations to
determine the degree, of the output corresponding to (x1,
x2); that is,
(yi) = i (x1) i(x2)

974

Table-Lookup illustration of fuzzy rule base

Figure 4

5
4
3
2
1
0
-1
-2
-3
-4

893

B2

812

B1

Time series prediction is very important practical

problem. Applications of time series prediction can be
found in the areas of economic and business planning,
inventory and production control, weather forecasting,
signal processing, control, and lots of other fields.
Let x (k) ( k = 1,2,3,.) be a time series. The problem of
time- series prediction can be formulated as : given x( k - n
+ 1) , x( k - n + 2) , , x(k), determine x( k + l )
where
n and l are fixed positive integers. That is, determine a
mapping from [x( k - n + 1) , x( k - n + 2) , , x( k ) ] Rn
to [x( k+ l )] R
Now we apply the above method of identification to the
time series production problem. Assuming that x(1), x(2),
, x(k) are given, we can form k-n input-output pairs:
[ x( k - n ) , , x( k -1) ; x( k) ]
[ x( k n - 1 ) , , x( k -2) ; x( k- 1 ) ]

[ x( 1 ) , , x( n) ; x( n+1) ]
(1.5)
The plant to be identified is described by the second-order
difference equation
x( k +1) = g [ x( k ), x( k 1 ) ] + u( k )
(1.6)
where
g [x( k ) , x ( k-1)] = x( k ) x ( k-1 )[ x( k )+2.5]
1+ x2( k )+ x2( k-1 )
(1.7)
and
u( k ) = Random Number [-2, +1]
(1.8)

731

CE

Time-Series

650

S1

to

569

S2
S3
S2
S1
CE
B1
B2
B3

Application
Prediction

488

Figure 4 illustrates a table-lookup representation of

a fuzzy rule base. We fill the boxes of the base with fuzzy
rules according to the following strategy: a combined fuzzy
rule base is assigned rules from either those generated from
numerical data or linguistic rules; if there is more than one
rule in one box of the fuzzy rule base, use the rule that has
maximum degree. In this way, both numerical and
linguistic rule is an and rule, it fills only one box of the
fuzzy rule base, but if a linguistic rule is an or rule (that is,
a rule for which the THEN parts follows if any condition of
the IF part is satisfied), it fills all the boxes in the rows or
columns corresponding to the regions of the IF part.

5.

407

Step 4: Create a Combined fuzzy rule Base

Where y denotes the center value of the output region of

rule I, and n is the number of fuzzy rules in the combined
fuzzy rule base.

326

Since there are usually lots of data pairs, and each

data pair generates one rule, it is highly probable that there
will some conflicting rules, i.e., rules that have the same IF
part but a different THEN part. One way to resolve this
conflict is to assign a degree to each rule generated from
data pairs, and accept only the rule from a conflict group
that has maximum degree. In this way not only is the
conflict problem resolved, but also number of rules is
greatly reduced.
We use the following product strategy to assign a degree to
each rule: for the rule : IF x1 is A and x2 is B, THEN y is C,
the degree of this rule, denoted by D(Rule), is defined as
D (Rule) = A (x1) B (x2) C (y)
As examples, Rule 1 has degree
D (Rule 1)
= B1 (x1) S1 (x2) CE (y)
= 0.8 0.7 0.9 = 0.504
(see Figure 3) and rule 2 has a degree
D (Rule 2)
= B1 (x1) CE (x2) B1 (y)
= 0.6 1 0.7 = 0.42

245

Step 3: Assign a Degree to Each Rule

Then we use the center average defuzzification formula to

determine the output
=
( y i ) . yi
y
(yi)

164

Division of the input-output space into fuzzy

regions and the corresponding membership
functions

83

Figure 3

Total time series

we use (1.6) to generate the numerical data. In our

simulation used to generate the plant numerical data, we
start the generating operation from k=1 upto k = 10000,
where we assume the initial value as x(0) = 0 and x(1) = 0.1
. Fig 5 shows the outputs of the plant from k=0 to k= 1000.

1453

6
Steps 1-4 of the above method are used to generate a fuzzy
rule base based on the input-output pairs (1.5); then this
fuzzy rule base is used to forecast x ( k+l ) for l=1,2, .,
using the defuzzifying procedure of step, where the inputs to
predictor are x(k+ l- n), x(k+l-n+1), ,x(k+l-1). We
assume number of inputs equally to four (n=4), and number
of output equally to one (l=1); that is four point values in the
series were used to predict the value of the next point. We
simulated three cases (I) when the number of fuzzy sets of
membership function is equal to 5, (II) when the number of
fuzzy sets are 7 and (III) when 9.

(x)
S4 S3

1
0
0

-1

50

100

150

-2
-3

The membership function for any point in cases (I),(II),(III)

are shown in figures 6, 7 and 8 respectively for the
numerical predictor. The first 900 points of the series were
used as training data, and the final 100 points wee used as
test data.

(x1)
S2
1

S1

CE

B1

-1

B2

X (k)
-3.5

-2

Figure 6 The membership function of all input-output

variables in case (I)

(x1)
S3

S2

S1

CE

B1

B2

B3

1
X(k)
0
-3.5 -2.6 -1.7 -0.8 0
1.3 2.2 3.1 4
Figure 7
The membership function of all input-output
variables in case (II)

B2 B3 B4

X(k)
-3.5 -2.75 -2 -1.25 -0.5
Figure 8

5.1

x(k)
2

CE B1

4
3

S2 S1

4
1 1.75 2.5 3.25

The membership function of all input-output

variables in case (III)

Simulation Results

We simulated this identification procedure for the

time series of figure 5. We started with the fuzzy rule base
generated by the data x(1) to x(900), then made a prediction
of x(901) to x(1000).
Figures 9, 10 and 11 show the results for case (I),
(II), and (III) respectively. We show that prediction can be
greatly improved by dividing the domain interval into finer
regions. Comparing Figures 3.9, 3.10 and 3.11 we see
clearly that we obtain better and better results as the domain
intervals is more finely divided. Of course, the price paid for
doing this is larger rule base. For example: 208, 302 and 501
fuzzy rules are generated in case (I),(II) and (III)
respectively from the simulations.

6. Conclusion
A fuzzy model consists of a number of if-then rules
that describe the behavior of a dynamic system. At the
beginning we presented the general methods to generate
fuzzy model rules. Then, we discussed the identification
problem of a fuzzy model, where we explained the two
kinds of identification: structure identification and
parameter identification.
Based on the table-lookup representation of the
fuzzy rule base, we presented a general method to generate
fuzzy rules from numerical data. In this sense, we consider
this method as one of so-called black-box identification
approaches. We applied this method to time series
prediction problem. Simulation results show that the
generated fuzzy model system is capable of approximating
non linear continuous function. It also shows that prediction
can be greatly improved by dividing the domain interval
into finer regions. We performed three simulations: the first
one with five fuzzy sets of the membership function, the
second one with seven fuzzy sets of the membership
function and third one with nine fuzzy sets of the
membership function.
The most important advantage o f this method is its
simplicity, it just performs a simple one-pass operation on
the training data. The price paid for this simplicity is that
we have to determine the partitions of the domain intervals
and the membership functions in ad hoc manner.

1454

Figure 9

Outputs of the plant (solid line) and the identification model (dashed line) with 5 fuzzy sets.

Figure 10

Outputs of the plant (solid line) and the identification model (dashed line) with 7 fuzzy sets.

Figure 11

Outputs of the plant (solid line) and the identification model (dashed line) with 9 fuzzy sets.

1455

References

[13] C.C. Lee, Fuzzy Logic in Control Systems: Fuzzy

Logic Controller, Part I and II.
IEEE Trans. On Systems, Man and Cybernetics, Vol.20, No.
2, March/April. 1990, pp 404-435.

[1] L. A. Zadeh, Fuzzy Sets, Information and Control,

Vol. 8, 1965, pp.338-353.
[2] L. A. Zadeh, Outline of a New Approach to the Analysis
of Complex Systems and Decision Processes, IEEE Trans.
Systems, Man, Cybernetics, SMC-3, No. 1, January 1973,
pp28-44.
[3] Takagi and M. Sugeno, Fuzzy Identification of Systems
and its Applications to Modeling and Control, Trans. On
Systems, Man, Cybernetics, Vol. SMC-15, No. 1, Jan/Feb.
1985, pp 116-32.

[14] V.Novak, Fuzzy sets and their Applications, Adam

Higler Publishers, 2989.
[15] J. C. Bezdek, Fuzzy Models- What Are They and
Why? IEEE Trans. On Fuzzy Systems, Editorial, Vol.1,
No. 1, February 1993.
[16] B. Kosko and S. Isaka, Fuzzy Logic, Scientific
American, July, pp 76-81.

[4] T.Takagi and M. Sugeno, Derivation of Fuzzy Control

Rules from Human Operators Control Actions, IFAC
Fuzzy Information, Marseill, France, 1983, pp 55-65.

[17] J. C. Bezdek, A review of Probabilistic, Fuzzy and

Neural Models for Pattern Recognition, J. of Int. and Fuzzy
Systems, Vol. l (1), No. 1, February 1993, pp 1-25.

[5] Yagishita, O.Itoh and M. Sugeno, Application of Fuzzy

Reasoning to Water Purification Process, In Industrial
Applications of Fuzzy Control, M.Sugeno, E.D.,
Amsterdam; North Holland, 1985, pp19-40.

[18] A. Kaufmann, Introduction to Theory of Fuzzy

Subsets, Academic Press Publishers, Vol. 1, 1975.

[6] S. Yasunobu and S. Miyamoto, Automatic Train

Operation by Predictive fuzzy Control, In M. Sugeno, E.D.,
industrial Application of Fuzzy Control, Amsterdam, North
Holland, 1985, pp 1-8.
[7] S. Yasunobu and G. Hasegawa, Evaluation of an
Automatic Crane Operation System based on Predictive
Fuzzy Control, Control Theory and Advanced Technology,
2, 1986, pp. 419-432.
[8] H. Takahashi, K. Ikeura, and T. Yamamori, 5-Speed
Automatic Transmission Installed Fuzzy Reasoning, in
proceedings of the International Fuzzy Engineering
Symposium, 91 (IFES91), November 1991, pp 1136-1137.
[9] D. Singer and P. G. Singer, System Identification Based
on Linguistic Variables, Fuzzy Sets and Systems 47,
North-Holland, 1992, pp 141-149.
[10] Li-Xin Wang and J.M. Mendel, generating Fuzzy
rules By Learning from examples, IEEE Trans. On
Systems, Man and Cybernetics, Vol.22, No. 6, Nov/Dec.
1992, pp 1414-1427.

[19] Li-Xin Wang, Adaptive Fuzzy Systems and Control,

PTR Prentice-Hall, 1994.
[20] B. Kosko, Neural Networks and Fuzzy Systems,
Prentice-Hall, 1992.
[21] Ragot and M. Lamotte, Fuzzy Logic Control, Int. J.
Systems Science, Vol.24, No. 10,1993, pp 1825-1848.
[22] T. Yamakawa, A fuzzy Inference Engine in non linear
Analog mode and its Application to a fuzzy Logic Control,
IEEE Trans. On Neural Networks, Vol. 4, No. 3, May, 1993,
pp 496-521.
[23] P.P Bonissone and K.H. Chiang, Fuzzy Logic
Controllers: From Development to Deployment, IEEE
Trans. On Neural Networks, Vol. 5, No. 3, May, 1993, pp
610-618.
[24] H.P. Preub, Fuzzy Control in Process Automation,
Int. J. Systems Science. Vol.24, No. 10, 1993, pp 18491861.
[25] M. Sugeno, An Introduction Survey of Fuzzy
Control, Information Sciences 36, 1985, pp 59-83.

[11] M. Sugeno and T. Yasukawa, A Fuzzy Logic Based

Approach to Qualitative Modeling, IEEE Trans. On Fuzzy
Sets and Systems, Vol.1, No. 1, 1993, pp 7-31.

[26] E. H. Mamdani, Advances in the linguistic Synthesis

of Fuzzy Controllers, International J. Man-Machine Stud.,
8(6), 1976, 669-678.

[12] C.W. Xu and Y. Zailu, Fuzzy Model Identification

and self learning for Dynamic systems, IEEE Trans. On
Systems, Man and Cybernetics, Vol. SMC. 17, No. 4,
Jul/Aug. 1987, pp 683-689.

[27] M. Sugeno and M. Nishida, Fuzzy Control of Model

Car, Int. J. Fuzzy Sets and Systems 10, North-Holland,
1985, 103-113.

1456