Category: Systems and Software Engineering

6981

A Short Review of Fuzzy

Linear Equation Systems
Hale Gonce Kocken
Yildiz Technical University, Turkey
Inci Albayrak
Yildiz Technical University, Turkey

INTRODUCTION
Systems of linear equations play a major role in various
areas such as operational research, physics, statistics,
engineering and social sciences. In many applications,
the parameters of linear equation systems are not always exactly known and stable. This imprecision may
follow from the lack of exact information, changeable
economic conditions, etc.. A frequently used way of
expressing the imprecision is to use the fuzzy numbers
rather than the crisp numbers. It enables us to consider
tolerances for parameters of linear equation systems
(the entries of the coefficient matrix and right hand
side vector) in a more natural and direct way. Therefore,
fuzzy linear equation systems seem to be more realistic
and reliable than the crisp case.
This study has attempted to provide a short review
of notable papers on fuzzy linear equation systems with
a basic structural classification. Considering the vast
literature, we only aim to present the most cited and
leading papers for each classification. Our short review
will give a general framework about the progression
of the subject and its solution approaches.

BACKGROUND
Wide range of real world applications in many areas
including financial engineering, scientific management
and engineering technology are using linear equation
systems for modeling and solving their respective
problems. In many situations, the estimation of the
system parameters is imprecise because of the lack
of exact information, environmental and changeable
economic conditions, etc. The uncertainty of parameters involved in the process of actual mathematical

modeling is often represented by fuzzy numbers, so

it is important to develop mathematical models that
would appropriately treat fuzzy linear equation systems.
One of the most important applications of linear
equation systems to electrical engineering is to analyze
electronic circuits that cannot be described using the
rules for resistors in series or parallel. The goals are
to calculate the current flowing in each branch of the
circuit and to calculate the voltage at each node of
the circuit, which are known Loop Current and Nodal
Voltage Analysis, respectively. Owing to environmental
conditions, tolerance in the elements and noise, these
analyses can be modeled in the form of fuzzy linear
equation systems.
Another important application of linear equation
systems is input-output analysis developed by V. Leontief. Input-output analysis is an extremely effective
tool used in more than 70 countries over the world
for manufacturing processes optimization, economy
condition improvement and intersectors costs allocation
analysis. Considering dynamic nature of economics,
it can be easily stated that extending this analysis to
fuzzy version of linear equation system is significant.
There is a vast literature on the investigation of
solutions for fuzzy linear equation systems. Early
works in the literature are on to linear equation systems
whose coefficient matrix is crisp and the right hand
vector is fuzzy, that is known as Fuzzy Linear Equation
System (FLS). The crispness of the coefficient matrix
makes the modeling of real life problems restricted.
Linear systems, whose all the parameters i.e. both
coefficient matrix and right hand vector are fuzzy, are
named Fully Fuzzy Linear Equation System (FFLS).
The main intend of FFLS is to widen the scope of
FLS in scientific applications by removing the crispness assumption on the entries of coefficient matrix.

DOI: 10.4018/978-1-4666-5888-2.ch687
Copyright 2015, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

Category:
A
Short Review
Systems
of Fuzzy
and Software
Linear Equation
Engineering
Systems

Besides FLS and FFLS, there exist the dual forms of

both systems in the literature.
Generally, both FLS and FFLS are handled under
two main headings: square (n x n) and nonsquare (m
x n) forms. Most of the works in the literature deal
with square form. Fuzzy elements of these systems
can be taken as triangular, trapezoidal or generalized
fuzzy numbers in general or parametric form. While
triangular fuzzy numbers are widely used in earlier
works, trapezoidal fuzzy numbers are neglected for a
long time. Besides, there exist lots of works using the
parametric and level cut representation of fuzzy numbers. Another classification for FFLS can be made also
depending on whether FFLS has sign restrictions on its
parameters. Having sign restrictions for FFLS means
that all parameters of FFLS are assumed as positive.
Since the parameters are assumed as positive in the
most of the papers, further work is needed for FFLS
with arbitrary (no restrictions on sign) fuzzy numbers.
Solution methods for FLS and FFLS can be gathered
into two groups: direct and iterative methods. Direct
methods which are also known as computational or
classical methods in the literature are generally used
for square systems. Some of the well known direct
methods can be given as: the Gauss elimination
method, Cramers rule, QR Decomposition method, LU
Decomposition method, Quasi-Newton method. The
iterative methods, which are mathematical procedures
that generate a sequence of improving approximate
solutions for the related problem, are used for both
square and nonsquare linear equation systems. Some
of the well known iterative methods can be given as:
Gauss Seidel, Jacobi, Adomian Decomposition, Homotopy perturbation.
This paper is organized with a basic structural classification. Primary classification is made considering
whether the system is square or not. Secondary classification is based on the fuzziness of the coefficient
matrix that is FLS or FFLS. Final classification is
constructed with respect to the solution methods, i.e.
direct or iterative. Considering the limitation of the
chapter, each classification involves the most cited and
notable papers consist of scientific refereed journals.
The objective of our review is to provide structural
methodology on FLS and to reemphasize a number
of gaps in the literature.

6982

SQUARE SYSTEMS

Fuzzy Linear Equation Systems

Direct Methods
Friedman, Ming and Kandel (1998) firstly introduces
a general model for solving a n x n FLS,

Ax = b

where A is a crisp matrix and b is an arbitrary fuzzy

number vector. Using embedding approach, the original system is replaced by a 2n x 2n crisp linear equation system and then they give the conditions for the
existence of a unique fuzzy solution to a n x n FLS.
Besides Abbasbandy and Alavi (2005) proposes an
efficient method replacing the original system by two n
x n crisp linear equation systems, Abbasbandy, Ezzati
and Jafarian (2006) discusses the LU decomposition
method in case the coefficient matrix is symmetric
positive definite.
Starting from the work by Friedman, Ming and
Kandel (1998), Ma, Friedman and Kandel (2000)
analyses the solution of dual fuzzy systems of the form

A1 x = A2 x + b . They remark that this system is not

equivalent to the system ( A1 A2 ) x = b , since there

is no element v such that u + v = 0, for an arbitrary

fuzzy number u.
Allahviranloo and Salahshour (2011) proposes a
simple and practical method for the system

Ax = b ,

where x and b are fuzzy triangular vectors with

non-zero spreads and A is nonsingular. They derive
L
three types of solutions: maximal ( x ) and minimal
U
( x ) symmetric solutions and a fuzzy vector solution

( x ). The last type solution corresponds the convex

combination of the maximal and minimal spreads, and
enables the decision maker to analyse the system in
desire positions by attaining some values to .
Ezzati (2011) gives a new method for solving square
FLS by using the embedding method given in CongXin and Ming (1991) and replaces the original n x n
FLS by two n x n crisp linear equation systems. And
that study also investigates the perturbation analysis
in two n x n crisp linear equation systems instead of a
2n x 2n crisp linear equation system as the authors of
Wang, Chen and Wei (2009) have done.

A Short
Category:
ReviewSystems
of Fuzzyand
Linear
Software
Equation
Engineering
Systems

Iterative Methods

Fully Fuzzy Linear Equation Systems

Allahviranloo (2004) which is based on Friedman,

Ming and Kandel (1998) applies Jacobi and Gauss
Sidel iterative methods for approximate of the unique
solution of FLS, since solving FLS as analytically is
difficult. Also he discusses convergence theorems
and states that when the unique solution of system
is a strong or weak fuzzy number, the approximate
solution of iterative method also would be a strong
or weak fuzzy number, respectively.
Allahviranloo (2005a) transforms the Gauss
Sidel iterative method in Allahviranloo (2004) to
the successive over relaxation (SOR) method. The
proposed method is followed by its convergence
theorem and the algorithm is compared with the
GaussSidel method by solving a numerical example.
Also Allahviranloo (2005b) applies the Adomian
decomposition method and shows that this method
is equivalent to the Jacobi iterative method. Allahviranloo, Ahmady, Ahmady and Shams Alketaby
(2006) proposes the Block Jacobi two-stage method
with GaussSidel inner iterations for approximation
of the unique solution of a FLS.
Dehghan and Hashemi (2006a) firstly investigates
a FLS using the embedding approach and then several
well-known numerical algorithms for solving crisp
linear equation systems such as Richardson, Extrapolated Richardson, Jacobi, JOR, GaussSeidel,
Extrapolated GaussSeidel (EGS), SOR, Accelerated
overrelaxation (AOR), Extrapolated SOR (ESOR),
Symmetric SOR (SSOR), Unsymmetric SOR (USSOR), Extrapolated modified Aitken (EMA) and
Modified SOR (MSOR) are extended for solving
FLS. This paper is remarkable in respect to introduce a large number of traditional algorithms for
the fuzzy case.
Saberi Najafi, Edalatpanah and Refahi Sheikhani
(2013) presents an efficient numerical algorithm
based on homotopy perturbation method. Apart from
the routine numerical methods, this method does not
involve discretization of the variables and hence is
free from rounding off errors and does not require
large computer memory or time. The authors make
comparisons between Adomians decomposition
method and the proposed method.

Direct Methods
Buckley and Qu (1991a) constructs solutions to the
fuzzy matrix equation

A x = b when the entries of

A and b are triangular fuzzy numbers. They present

six type solutions, and show that five are identical.

Another works of the same authors about FFLS are
Buckley and Qu (1990) and Buckley and Qu (1991b).
These papers are generally accepted as pioneer works
on FFLS.
Muzzioli and Reynaerts (2006) clarifies the link
between interval linear systems and fuzzy linear systems
and provides a generalization of the vector solution
proposed in Buckley and Qu (1991a) to the dual fuzzy

A1 x + b1 = A 2 x + b2 with A1 and A 2 square

fuzzy matrices and b1 , b2 fuzzy number vectors. They
system

give the conditions under which the system has a vec-

A x = b
x + b = A x + b , with A = A A and
and A
1
1
2
2
1
2

have
the
same
vector
solutions,
where
b = b2 b1
tor solution and show that the linear systems

denotes the standard difference.

Allahviranloo, Mikaeilvand, Kiani and Shabestari
(2008) proposes a novel method for finding the nonzero fuzzy solutions of FFLS with positive coefficient
matrix. They replace the original n x n FFLS by a 2n x
4n parametric linear equation system and solve the 2n
x 2n system by omitting 2n columns of 2n x 4n matrix.
In Allahviranloo and Mikaeilvand (2011), non-zero
fuzzy number definition is given as fuzzy number
which 0 is not inner point of its support. Based on this
definition, an algorithm is proposed using embedding
approach for FFLS with arbitrary coefficients to find its
non-zero fuzzy number solutions. Using this algorithm,
one cannot solve the systems that have zero solutions,
i.e., it is not applicable to all FFLS.
Kumar, Babbar and Bansal (2011) gives an approach
based on linear programming to solve a FFLS with
arbitrary coefficients and arbitrary solution vector, that
is, there is no sign restriction on parameters of the
system. Similarly, Otadi and Mosleh (2012a) develops
a linear programming approach with equality constraints to find a non-negative fuzzy number matrix

= b . The restriction of this

x which satisfies Ax

paper is the positivity assumption of variables.

6983

Category:
A
Short Review
Systems
of Fuzzy
and Software
Linear Equation
Engineering
Systems

In Babbar, Kumar and Bansal (2013), three new

numerical methods are discussed to solve a FFLS with
triangular fuzzy numbers. The first and second methods
remove the sign restriction on the coefficient matrix
and the solution vector, respectively. And the third
method removes joint restrictions on the coefficients
as well as the solution vector; however, confining the
solutions to exclude near zero fuzzy numbers.
Moloudzadeh, Allahviranloo and Darabi (2013)
presents a simple and practical method that works
only for triangular fuzzy numbers to solve a FFLS with
arbitrary fuzzy parameters and variables by using the
cross-product of fuzzy numbers. The limitation of the
proposed method is to generate a fuzzy solution only
providing the theorem 3.2 in that paper.

Iterative Methods
Dehghan and Hashemi (2006b) extends the Adomian
decomposition method for solving FFLS to find a

= b . They
positive fuzzy vector x that satisfies Ax
employ approximate arithmetic operators on LR fuzzy
numbers, use the Adomian decomposition method and
compare this method with the Jacobi iterative technique.
Dehghan, Hashemi and Ghatee (2007) focuses on
the positive solution of FFLS with some well-known
iterative methods such as Jacobi, GaussSeidel, SOR,
AOR, SSOR, USSOR and EMA. In addition, the methods of Newton, quasi-Newton and conjugate gradient
are applied to FFLS.
Otadi and Mosleh (2011) is one of the many different attempts to derive learning algorithms of fuzzy
neural networks with fuzzy input, fuzzy output and
fuzzy weights. The effectiveness of the derived learning
algorithm is demonstrated by computer simulation of
numerical examples.
NONSQUARE SYSTEMS
Fuzzy Linear Equation Systems
Direct Methods
Asady, Abbasbandy and Alavi (2005) develops a
method for solving m x n FLS for m n using the
same approach in Friedman, Ming and Kandel (1998).

6984

After the original system is replaced by a crisp linear

equation system, then it is solved by least square
method.
Allahviranloo and Afshar Kermani (2006) investigates m x n consistent FLS using a numerical method.
They first replace the original m x n fuzzy linear system
by a 2m x 2n crisp linear equation system, and then
approximate solutions are obtained by using pseudo
inverse, orthogonal matrix and least-square method.
Zheng and Wang (2006) also studies on m x n
consistent FLS. By the embedding method, the original
system is converted to a 2m x 2n crisp linear equation
system. And then the existence and expression of the
strong fuzzy solution to the system are discussed by
using the generalized inverses of the coefficient matrix.
These authors study the inconsistent form in Wang
and Zheng (2006). The least squares solution and the
minimum norm least squares solution to the system
are obtained by the same method, and they provide
a sufficient condition for the least squares solution
being a fuzzy vector. Gong and Guo (2011), which is
an extension of Wang and Zheng (2006), presents a
model to solve a class of inconsistent nonsquare FLS
by the embedding method.
Abbasbandy, Otadi and Mosleh (2008) proposes a
numerical method based on pseudo-inverse calculation
for finding minimal solution of m x n dual FLS. Two
necessary and sufficient conditions for the minimal
solution existence are given.
Otadi and Mosleh (2012b) develops a method to
solve an arbitrary inconsistent FLS by using the embedding approach and investigates perturbation analysis
in two m x n crisp linear equation systems instead of
2m x 2n crisp linear equation system as the authors of
Ezzati (2011) and Wang, Chen and Wei (2009).

Iterative Methods
Abbasbandy, Jafarian and Ezzati (2005) and Abbasbandy and Jafarian (2006) apply the conjugate
gradient method and the steepest descent method for
approximation of the unique solution of FLS, respectively. They also give the convergence theorem of the
proposed algorithms and illustrate them by solving
some numerical examples.
Ghanbari and Mahdavi-Amiri (2010) firstly introduces the use of a ranking function to define a compromised solution for nonsquare FLS and then proposes a

5 more pages are available in the full version of this document. Because your
institution is experiencing a trial period, the remainder of this document is
reserved. Please contact your Electronic Resources Librarian and share your
interest in gaining access to this collection. For further details contact
eresources@igi-global.com. This document may also be purchased by
visiting www.igi-global.com/ondemand. With access to thousands of
searchable documents in the fields of information science technology and
management, enables the user to purchase individual articles specific to their
needs.