ARTICLE IN PRESS

Journal of Computational and Applied Mathematics ( ) –

Contents lists available at ScienceDirect

Journal of Computational and Applied

Mathematics
journal homepage: www.elsevier.com/locate/cam

Fuzzy symmetric solutions of fuzzy linear systems

T. Allahviranloo ∗ , S. Salahshour
Department of Mathematics, Science and Research Branch, Islamic Azad University, Tehran, Iran

article info abstract

Article history: In this paper, we propose a simple and practical method to solve a fuzzy linear system
Received 7 January 2010 AX̃ = b̃, where X̃ and b̃ are fuzzy triangular vectors with non-zero spreads and matrix A
is nonsingular with real coefficients. The aim of this paper is twofold. First, we obtain the
Keywords: crisp solution of a fuzzy linear system. To this end, we solve the 1-cut of a fuzzy linear
Fuzzy linear system system. Second, we allocate some unknown symmetric spreads to any rows of fuzzy linear
Symmetric spread
system in 1-cut position. Thus, a fuzzy linear system in 1-cut will be transformed to a
Fuzzy vector solution
United Solution Set (USS)
system of interval equations. The symmetric spreads of each element of a fuzzy vector
Tolerable Solution Set (TSS) solution are derived by solving such an interval system. Moreover, based on the obtained
Controllable Solution Set (CSS) symmetric spreads we derive three types of solutions. However, one of the mentioned
spreads has pessimistic/optimistic attitude that is determined with a decision maker. It
seems that such a solution is a connection solution between Tolerable Solution Set (TSS)
and Controllable Solution Set (CSS). Also, we derive the maximal solution and the minimal
solution of an original fuzzy linear system which are placed in TSS and CSS, respectively.
Finally, some numerical examples are given to illustrate the efficiency and ability of the
proposed method.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Linear systems of equations, with uncertainty on the parameters, play a major role in several applications in various areas
such as economics, finance, engineering and physics. Hence, it is important to develop mathematical models and numerical
procedures that would appropriately treat fuzzy linear systems and solve them.
The n × n fuzzy linear system has been studied by several authors. Buckley and Qu in [1] construct solutions to the fuzzy
matrix equation e AeX =e b when the elements in e A and e
b are triangular fuzzy numbers. They presented six solutions of which
they shown that five are identical. Also further work by them is given in [2,3].
In [4], Friedman et al. proposed a general method for solving such fuzzy linear systems by using an embedding approach.
Ma et al. [5], starting from the work in [4], analyze the duality of fuzzy systems. They remark that the system A1 x = A2 x + b
is not equivalent to the system (A1 − A2 )x = b, since for an arbitrary fuzzy number u there exists no element v such that
u + v = 0.
Consequently, Muzzilio and Reynaerts [6] proposed a generalization of the vector solution of Buckley and Qu [1] to the
fuzzy system A1 x + b1 = A2 x + b2 : they gave the conditions under which the system has a vector solution and they showed
that the linear systems e AeX =e X + be1 = Ae2e
b and Ae1e X + be2 , with e
A = Ae1 − Ae2 and e
b = be2 − be1 have the same vector solutions,
where ‘−’ denotes the standard difference.
In computational methods, Allahviranloo [7–9] used the iterative Jacobi and Gauss–Siedel method, the Adomian method
and the successive over relaxation method, respectively. Also, in [10,11], Abbasbandy et al. solve a fuzzy system of equations
using the LU decomposition method and the steepest decent method, respectively. For more research papers see [12–17].

∗ Corresponding author. Tel.: +98 9123508816; fax: +98 2144804172.

E-mail addresses: [email protected] (T. Allahviranloo), [email protected] (S. Salahshour).

0377-0427/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.cam.2010.02.042

Please cite this article in press as: T. Allahviranloo, S. Salahshour, Fuzzy symmetric solutions of fuzzy linear systems, Journal of Computational and
Applied Mathematics (2010), doi:10.1016/j.cam.2010.02.042
 ARTICLE IN PRESS
2 T. Allahviranloo, S. Salahshour / Journal of Computational and Applied Mathematics ( ) –

The aim of this paper is twofold. First, we solve the fuzzy linear system in crisp case or equivalently, 1-cut of fuzzy linear
system. Second, by some fuzzification, we try to obtain the fuzzy vector solutions by investigating symmetric spreads of
each elements of fuzzy vector solution.
The outline of the paper is as follows: In Section 2 we review some important basic concepts. In Section 3 we introduce
our proposed method and focuses on how we could derive the fuzzy vector solution or equivalently, how we could derive
the symmetric spreads of each component of fuzzy vector solution corresponding to TSS or CSS. In Section 4, some numerical
examples are given to show the efficiency of proposed method and conclusion is drawn in Section 5.

2. Preliminaries

A nonempty subset A of R is called convex if and only if (1 − k)x + ky ∈ A for every x, y ∈ A and k ∈ [0, 1]. By Pk (R), we
denote the family of all nonempty compact convex subsets of R.
There are various definitions for the concept of fuzzy numbers (see [18–20]).

Definition 2.1. A fuzzy number is a function such as u : R → [0, 1] satisfying the following properties:
(i) u is normal, i.e. ∃x0 ∈ R with u(x0 ) = 1,
(ii) u is a convex fuzzy set i.e. u(λx + (1 − λ)y) ≥ min{u(x), u(y)}∀x, y ∈ R, λ ∈ [0, 1],
(iii) u is upper semi-continuous on R,
(iv) {x ∈ R : u(x) > 0} is compact, where A denotes the closure of A.
The set of all fuzzy real numbers is denoted by E. Obviously R ⊂ E. Here R ⊂ E is understood as R = {χx :
x is usual real number}. For 0 < r ≤ 1, it is denoted that [u]r = {x ∈ R; u(x) ≥ r } and [u]0 = {x ∈ R; u(x) > 0}. Then it
is well-known that for any r ∈ [0, 1], [u]r is a bounded closed interval. For e u,ev ∈ E, and λ ∈ R, where sum e u +e v and the
product λ.e u are defined by [u + v]r = [u]r + [v]r , [λ.u]r = λ[u]r , ∀r ∈ [0, 1], where [u]r + [v]r = {x + y : x ∈ [u]r , y ∈ [v]r }
means the conventional addition of two intervals (subsets) of R and λ[u]r = {λx : x ∈ [u]r } means the conventional product
between a scalar and a subset of R (see e.g. [18]).
Another definition for a fuzzy number is as follows:

Definition 2.2. An arbitrary fuzzy number in the parametric form is represented by an ordered pair of functions (u(r ), u(r )),
0 ≤ r ≤ 1, which satisfy the following requirements:
1. u(r ) is a bounded left-continuous non-decreasing function over [0, 1].
2. u(r ) is a bounded left-continuous non-increasing function over [0, 1].
3. u(r ) ≤ u(r ), 0 ≤ r ≤ 1.
A crisp number α is simply represented by u(r ) = u(r ) = α , 0 ≤ r ≤ 1. We recall that for a < b < c, a, b, c ∈ R, the
triangular fuzzy number e u = (a, b, c ) determined by a, b, c is given such that u(r ) = a +(b − a)r and u(r ) = c −(c − b)r are
the endpoints of the r-level sets, for all r ∈ [0, 1]. Here u(r ) = u(r ) = b is denoted by [u]1 . For arbitrary u = (u(r ), u(r )),
v = (v(r ), v(r )) we define addition and multiplication by k as
1. (u + v)(r ) = (u(r ) + v(r )),
2. (u + v)(r ) = (u(r ) + v(r )),
3. (ku)(r ) = ku(r ), (ku)(r ) = ku(r ), k ≥ 0,
4. (ku)(r ) = ku(r ), (ku)(r ) = ku(r ), k < 0.

Definition 2.3. For arbitrary fuzzy quantity u = (u, u) and v = (v, v) the quantity
Z 1 Z 1
 12
D(u, v) = (u(r ) − v(r )) dr +
2
(u(r ) − v(r )) dr
2
, (1)
0 0

is the distance between u and v .

Definition 2.4. The n × n linear system of equations



 a11 xe1 + a12 xe2 + · · · + a1n xen = be1
a21 xe1 + a22 xe2 + · · · + a2n xen = be2

.. (2)


 .

an1 xe1 + an2 xe2 + · · · + ann xen = ben

where the elements, aij of the coefficients matrix A, 1 ≤ i, j ≤ n, are real values and the elements, e
bi , of the vector e
b and the
unknown vector solution X = (e
e x1 , . . . , xen ) are fuzzy vectors is called fuzzy linear systems.
t

Please cite this article in press as: T. Allahviranloo, S. Salahshour, Fuzzy symmetric solutions of fuzzy linear systems, Journal of Computational and
Applied Mathematics (2010), doi:10.1016/j.cam.2010.02.042
 ARTICLE IN PRESS
T. Allahviranloo, S. Salahshour / Journal of Computational and Applied Mathematics ( ) – 3

X = (e
Definition 2.5. A fuzzy vector e x1 , . . . , xen )t given by e
xi = [xi (r ), xi (r )] , 1 ≤ i ≤ n, 1 ≤ r ≤ n is called the solution of
(2) if
n
X n
X n
X n
X
aij xj = aij xj = bi , aij xj = aij xj = bi .
j=1 j =1 j =1 j =1

If for a particular i, aij > 0, j = 1, . . . , n we simply get

n
X n
X
aij xj = bi , aij xj = bi .
j=1 j =1

Clearly, Definition 2.5 is for exact solution and in this paper we will investigate the approximate fuzzy symmetric solution.

Definition 2.6 ([21]). The united solution set (USS), the tolerable solution set (TSS) and controllable solution set (CSS) for
the system (2) are respectively as following:
X∃∃ = x0 ∈ Rn : Ax0 ∩ b0 6= φ


X∀∃ = x0 ∈ Rn : Ax0 ⊆ b


X∃∀ = x0 ∈ Rn : Ax0 ⊇ b .


X = (e
Definition 2.7. A fuzzy vector e x1 , . . . , xen )t given by e
xi = [xi (r ), xi (r )] , 1 ≤ i ≤ n, 1 ≤ r ≤ n is called the minimal
Y = (e
symmetric solution of (2) which is placed in CSS if for any arbitrary symmetric solution e y1 , . . . , yen )t which is place in
Y (1) = e
CSS that is e X (1) we have
X , i.e., (e xi ), i.e., (σeyi ≥ σexi ), ∀i = 1, . . . , n


Y ⊇e
e yi ⊇ e
where σeyi and σexi are symmetric spreads of e
yi and e
xi , respectively.

X = (e
Definition 2.8. A fuzzy vector e x1 , . . . , xen )t given by e
xi = [xi (r ), xi (r )] , 1 ≤ i ≤ n, 1 ≤ r ≤ n is called the maximal
Z = (e
symmetric solution of (2) which is placed in TSS if for any arbitrary symmetric solution e z1 , . . . , zen )t which is place in
Z (1) = e
TSS that is e X (1) we have
Z , i.e., (e zi ), i.e., (σexi ≥ σezi ), ∀i = 1, . . . , n


X ⊇e
e xi ⊇ e
where σexi and σezi are symmetric spreads of e
xi and e
zi , respectively.

3. The method

In this section, we will propose a new simple method to solve the fuzzy linear system (2) which give us a fuzzy vector
solution. In the beginning, to obtain the solution of fuzzy linear system (2), we solve the 1-cut of system (2) and derive the
crisp solution. To this end, system (2) is converted to crisp system:
n
X
aij xj = bi (1), i = 1, 2, . . . , n (3)
j=1

where, bi (1) ∈ R and xj is unknown real variable, j = 1, . . . , n. Since A is nonsingular, we could obtain the crisp solution
of system (2), equivalently, the solution of system (3). However, using such crisp solution we fuzzify the crisp system (3)
by allocating some unknown symmetric spread to each row. To this end, the crisp system (3) is converted to the following
system of interval equations:

a11 [x1 − α1 (r ), x1 + α1 (r )] + · · · + a1n [xn − α1 (r ), xn + α1 (r )] = [b1 (r ), b1 (r )]

.. (4)
.
an1 [x1 − αn (r ), x1 + αn (r )] + · · · + ann [xn − αn (r ), xn + αn (r )] = [bn (r ), bn (r )]

where, αi (r ), i = 1, . . . , n are unknown spreads and xj , j = 1, . . . , n are known solutions of crisp system (3).
Now, we solve the interval system (4) to obtain αi (r ), i = 1, . . . , n.
So, without loss of generality and for simplicity to express the theory, is assumed that the coefficients matrix A are
positive. Then, ith equation of interval system (4) is

ai1 [x1 − αi (r ), x1 + αi (r )] + · · · + ain [xn − αi (r ), xn + αi (r )] = [bi (r ), bi (r )]. (5)

Please cite this article in press as: T. Allahviranloo, S. Salahshour, Fuzzy symmetric solutions of fuzzy linear systems, Journal of Computational and
Applied Mathematics (2010), doi:10.1016/j.cam.2010.02.042
 ARTICLE IN PRESS
4 T. Allahviranloo, S. Salahshour / Journal of Computational and Applied Mathematics ( ) –

Eq. (5) is rewritten in compact form:

n
X
aij xj − αi (r ) = bi (r ), i = 1, . . . , n


(6)
j =1

and
n
X
aij xj + αi (r ) = bi (r ), i = 1, . . . , n.


(7)
j =1

So, after some manipulations, Eqs. (6) and (7) are transformed respectively to:

αi (r ) = f1 x1 , . . . , xn , ai1 , . . . , ain , bi (r ) , i = 1, . . . , n


(8)
and
αi (r ) = f2 x1 , . . . , xn , ai1 , . . . , ain , bi (r ) , i = 1, . . . , n.


(9)
From now to end, we replace αi1 (r ) with αi (r ) in Eq. (8) and replace αi2 (r ) with αi (r ) in Eq. (9). Therefore, we have
αi1 (r ) = f1 x1 , . . . , xn , ai1 , . . . , ain , bi (r ) , i = 1, . . . , n

and
αi2 (r ) = f2 x1 , . . . , xn , ai1 , . . . , ain , bi (r ) , i = 1, . . . , n.

However, αi1 (r ) is function of x1 , . . . , xn , ai1 , . . . , ain , bi (r ) and αi2 (r ) is function of x1 , . . . , xn , ai1 , . . . , ain , bi (r ) such that
αi1 (r ) and αi2 (r ) are obtained spreads of ith equation in system (4). Perhaps, αi1 (r ) and αi2 (r ) did not satisfied to the rest
of interval equations (4). Therefore, we should determine the reasonable spreads according to decision makers. To this end,
three type of spreads are proposed as following:
αL (r ) = min {αi1 (r ), αi2 (r )} , i = 1, . . . , n, 0 ≤ r ≤ 1, (10)
r

αU (r ) = max {αi1 (r ), αi2 (r )} , i = 1, . . . , n, 0 ≤ r ≤ 1, (11)

r

αλ (r ) = λαU (r ) + (1 − λ)αL (r ), i = 1, . . . , n, 0 ≤ r ≤ 1, λ ∈ [0, 1]. (12)

Hence, by such computations, the fuzzy vector solution of system (2) under proposed spreads (10)–(12) will be as following
for i = 1, . . . , n, 0 ≤ r , λ ≤ 1:

t
XL = (g
e x1 )(r ), . . . , (g
xn )(r ) , where (f
xi )(r ) = [xi − αL (r ), xi + αL (r )] , (13)

t
XU = (g
e x1 )(r ), . . . , (g
xn )(r ) , where (f
xi )(r ) = [xi − αU (r ), xi + αU (r )] , (14)

t
Xλ = (g
e xn )(r ) , where (f
x1 )(r ), . . . , (g xi )(r ) = [xi − αλ (r ), xi + αλ (r )] . (15)
Now, we show that our method always give us a fuzzy vector solution provided that the right-hand side of system (2) be a
triangular fuzzy vector with non-zero left and right spreads.

Theorem 3.1. Let the right-hand side of system of linear equation (2), is e b(r ) = (b̃1 (r ), . . . , b̃n (r ))t where, b̃i (r ) = [bi (1) −
σi + σi r , bi (1) + βi − βi r ] i = 1, . . . , n, and let αL (r ), αU (r ), and αλ (r ) are defined by (10)–(12), then αL (r ), αU (r ), and αλ (r ),
are positive for all 0 ≤ r , λ ≤ 1, such that
 

 

σi (1 − r ) βi (1 − r )

 

αL (r ) = min n
, n , (16)
i=1,...,n  P P 


 |a ij | | a ij | 


j =1 j =1
 

 

σi (1 − r ) βi (1 − r )

 

αU (r ) = max n
, n , (17)
i=1,...,n  P P 


 |aij | |aij | 


j =1 j =1
   

 
 
 

σi (1 − r ) βi (1 − r ) σi (1 − r ) βi (1 − r )

 
 
 

αλ (r ) = λ max n
, n + (1 − λ) min n
, n . (18)
i=1,...,n  P P  i=1,...,n  P P 


 | a ij | |a ij | 




 |a ij | |a ij | 


j =1 j=1 j =1 j =1

Please cite this article in press as: T. Allahviranloo, S. Salahshour, Fuzzy symmetric solutions of fuzzy linear systems, Journal of Computational and
Applied Mathematics (2010), doi:10.1016/j.cam.2010.02.042
 ARTICLE IN PRESS
T. Allahviranloo, S. Salahshour / Journal of Computational and Applied Mathematics ( ) – 5

Proof. We prove only for (16) and the others are similar. So, it is sufficient to prove that
σi (1 − r ) β i (1 − r )
αi1 (r ) = n
, αi2 (r ) = n
, i = 1, . . . , n, 0 ≤ r ≤ 1.
P P
|aij | |aij |
j =1 j=1

Let us consider the ith row of interval equations (4), then by applying Eqs. (6)–(7), we have
X X n
X n
X
aij (xj − αi1 (r )) + aij (xj + αi1 (r )) = aij xj − |aij |αi1 (r ) = bi (r ) = bi (1) − σi + σi r
j∈J1 j∈J2 j=1 j =1

X X n
X n
X
aij (xj + αi2 (r )) + aij (xj − αi2 (r )) = aij xj + |aij |αi2 (r ) = bi (r ) = bi (1) + βi − βi r
j∈J1 j∈J2 j=1 j =1

where J1 and J2 are include positive and negative components of coefficient matrix A, respectively. Since, bi (r ) = bi (1) =
Pn
j =1 aij xj , then for fixed i, is obtained as follow:

σi (1 − r ) β i (1 − r )
αi1 (r ) = n
, αi2 (r ) = n
.
P P
|aij | |aij |
j =1 j=1

Also, it is obvious that σi , βi and the denominator are positive numbers. Therefore, αL (r ) > 0. 

Theorem 3.2. Our proposed solution of fuzzy linear system (2) is always gives fuzzy vector.
Proof. Using Theorem 3.1, proof is obvious. 
Notice that, the proposed symmetric spreads of solution of system (2) in crisp manner should be zero, since, in 1-cut,
system (2) is converted to the crisp problem and has a crisp solution. To prove this property we have the following:

Theorem 3.3. Let fuzzy linear system (2) has a crisp solution X = (x1 , . . . , xn )t where xi ∈ R, i = 1, . . . , n and αi1 (r ) and
αi2 (r ), 0 ≤ r ≤ 1, are obtained from (6) and (7), respectively. then
(a) αi1 (1) = 0, i = 1, . . . , n
(b) αi2 (1) = 0, i = 1, . . . , n
(c) e
xi (1) = χ{xi } , i = 1, . . . , n.

Proof. By applying (6) and (7), we have for i = 1, . . . , n and 0 ≤ r ≤ 1:

n
X n
X
aij (xj − αi1 (r )) = bi , aij (xj + αi2 (r )) = bi .
j=1 j =1

Set r = 1, since, e bi is triangular fuzzy number we have bi (1) = bi (1). Also, xj , j = 1, . . . , n is jth element of crisp solution
X = (x1 , . . . , xn )t . Thus, crisp system (3) will be as:
n
X
bi (1) = bi (1) = bi (1),
aij xj = e i = 1, . . . , n.
j=1

Therefore:
n
X n
X n
X n
X
aij (xj − αi1 (1)) = aij xj − aij αi1 (1) = bi (1) − aij αi1 (1) = bi (1).
j=1 j =1 j =1 j =1

So, we have
n
X
aij αi1 (1) = 0, i = 1, . . . , n. (19)
j=1

Since, A is nonsingular and we assumed that the coefficients of matrix A are positive, is concluded that αi1 (1) = 0 for all
i = 1, . . . , n. Also similar discussion is valid to obtain αi2 (1) = 0. However, using (10)–(12) lead to drive, respectively:
xi (r ) = [xi − αL (r ), xi + αL (r )] = χ{xi } , i = 1, . . . , n, 0 ≤ r ≤ 1,
e
xi (r ) = [xi − αU (r ), xi + αU (r )] = χ{xi } , i = 1, . . . , n, 0 ≤ r ≤ 1,
e
xi (r ) = [xi − αλ (r ), xi + αλ (r )] = χ{xi } , i = 1, . . . , n, 0 ≤ r ≤ 1, 0 ≤ λ ≤ 1. 
e
Please cite this article in press as: T. Allahviranloo, S. Salahshour, Fuzzy symmetric solutions of fuzzy linear systems, Journal of Computational and
Applied Mathematics (2010), doi:10.1016/j.cam.2010.02.042
 ARTICLE IN PRESS
6 T. Allahviranloo, S. Salahshour / Journal of Computational and Applied Mathematics ( ) –

XL , e
Theorem 3.4. Let us consider spreads (10)–(12) and corresponding solutions e XU , then we get:
XL ∈ TSS
1. e
XU ∈ CSS.
2. e

XL , e
Theorem 3.5. Let us consider spreads (10)–(11) and corresponding solutions e XU and e
Xλ , then we get:
1. if {λk }nk=0 be an increasing sequence of real numbers with starting point λ0 = 0 such that λk −→ 1, when k −→ ∞, then
Xλ is a connection solution with start form TSS and end in CSS, i.e.,
e
XU ∈ CSS , λk −→ 1, λ0 = 0.



XL ∈ TSS −→ e
e k −→ ∞,

2. if {λk }nk=0 be a decreasing sequence of real numbers with starting point λ0 = 1 such that λk −→ 0, when k −→ ∞, then e
Xλ
is a connection solution with start form CSS and end in TSS, i.e.,
XL ∈ TSS , λk −→ 0, λ0 = 1.



XU ∈ CSS −→ e
e k −→ ∞,

Proof. Using definitions of TSS and CSS, the proofs are obvious. 

In addition, we can find the maximal and minimal solution of fuzzy linear system (2) which are placed in TSS and CSS
when the cores of compared solutions in each cases are equal.

Theorem 3.6. e
XL is maximal symmetric solution in TSS.

Theorem 3.7. e
XU is minimal symmetric solution in CSS.

Moreover, we could express our proposed method by algorithm as following:

Algorithm. • Let system (2) has crisp solution, or equivalently, A is nonsingular. Set the crisp solution as X = (x1 , . . . , xn )t ,
where xi ∈ R, i = 1, . . . , n.
• By applying crisp solution (solution of 1-cut), system (2) is transformed to the system of interval equations (3).
• The spread of all elements of fuzzy vector solution will be obtained by solving system (4), whereas, spreads are named
as αi1 (r ) and αi2 (r ), respectively, i = 1, . . . , n, 0 ≤ r ≤ 1.
• The symmetric spreads can be assessed using (10)–(12).
• The fuzzy vector solutions are derived by (13)–(15).

4. Numerical examples

In this section, we solve some examples to illustrate the utility of proposed method. To this end, we shall obtain maximal
symmetric solution which is placed in TSS and minimal symmetric solution which is placed in CSS. Moreover, we obtain a
connection solution which has optimistic/pessimistic attitude. Notice that, we use (1) as error between approximate solution
and exact solution.

Example 4.1 ([9]). Consider 2 × 2 fuzzy linear system:

xe1 − xe2 = (r , 2 − r ),

xe1 + 3xe2 = (4 + r , 7 − 2r ).

Then, 1-cut of system is

x1 − x2 = 1, x1 + 3x2 = 5.
Therefore, the crisp solution is X = (x1 , x2 )t where, x1 = 2 and x2 = 1. Now the system of interval equations (4) is as follow:
[2 − α1 (r ), 2 + α1 (r )] − [1 − α1 (r ), 1 + α1 (r )] = [r , 2 − r ]
[2 − α2 (r ), 2 + α2 (r )] + 3[1 − α2 (r ), 1 + α2 (r )] = [4 + r , 7 − 2r ].
Hence, the following results are obtained for all r ∈ [0, 1] as:
1−r 1−r
α11 (r ) = α12 (r ) = α22 (r ) = , α21 (r ) =
2 4
and applying (10)–(12) we get for all 0 ≤ r , λ ≤ 1
1−r 1−r (1 − r )(1 + λ)
αL (r ) = , αU ( r ) = , αλ =
4 2 4
Please cite this article in press as: T. Allahviranloo, S. Salahshour, Fuzzy symmetric solutions of fuzzy linear systems, Journal of Computational and
Applied Mathematics (2010), doi:10.1016/j.cam.2010.02.042
 ARTICLE IN PRESS
T. Allahviranloo, S. Salahshour / Journal of Computational and Applied Mathematics ( ) – 7

Fig. 1. Error of approximate solution e

Xλ in Example 4.1.

and the fuzzy vector solutions are obtained as following for all 0 ≤ r , λ ≤ 1:
   t
1−r 1−r 1−r 1−r
XL ( r ) =
e 2− ,2 + , 1− ,1 + ,
4 4 4 4
   t
1−r 1−r 1−r 1−r
XU (r ) =
e 2− ,2 + , 1− ,1 + ,
2 2 2 2
(1 − r )(1 + λ) (1 − r )(1 + λ) (1 − r )(1 + λ) (1 − r )(1 + λ) t
   
Xλ ( r ) =
e 2− ,2 + , 1− ,1 + .
4 4 4 4
Also, the exact solution [9] is:
xe1 (r ) = [1.375 + 0.625r , 2.875 − 0.875r] ,
xe2 (r ) = [0.875 + 0.125r , 1.375 − 0.375r ].
In addition, we plot error of approximate solution Xλ = (xλ,1 , xλ,2 )t in compare of exact solution e
X = (x1 , x2 )t (see Fig. 1).

Example 4.2 ([22]). Let us consider the 5 × 5 fuzzy system

6xe1 + xe2 + 3xe3 − xe4 + 6xe5 = (1 + r , 3 − r )


5xe1 + 9xe2 + xe3 + 2xe4 + 3xe5 = (6 + r , 8 − r )


2xe1 + 3xe2 + 9xe3 + 2xe4 + 3xe5 = (5 + r , 7 − r )
x1 + xe2 + 3xe3 + 8xe4 + 3xe5 = (3 + r , 5 − r )
−e



xe1 + 2xe2 + 2xe3 + xe4 + 9xe5 = (2 + r , 4 − r ).

The 1-cut solution of system is X (r ) = (0.0053, 0.6519, 0.3659, 0.2525, 0.0785)t .

However, we have the following:
1−r 1−r 1−r 1−r
α11 = α12 = , α21 = α22 = , α31 = α32 = , α41 = α42 = ,
17 20 19 16
1−r
α51 = α52 = .
15
Hence, symmetric spreads are derived as following for all 0 ≤ r , λ ≤ 1:
1−r 1−r (1 − r )(1 + λ)
αL (r ) = , αU (r ) = , αλ ( r ) = .
20 15 60
XLt (r ) will be:
Therefore, e
 
1−r 1−r
x1 (r ) = 0.0053 −
e , 0.0053 + ,
20 20
 
1−r 1−r
xe2 (r ) = 0.6519 − , 0.6519 + ,
20 20
Please cite this article in press as: T. Allahviranloo, S. Salahshour, Fuzzy symmetric solutions of fuzzy linear systems, Journal of Computational and
Applied Mathematics (2010), doi:10.1016/j.cam.2010.02.042
 ARTICLE IN PRESS
8 T. Allahviranloo, S. Salahshour / Journal of Computational and Applied Mathematics ( ) –
 
1−r 1−r
xe3 (r ) = 0.3659 − , 0.3659 + ,
20 20
 
1−r 1−r
xe4 (r ) = 0.2525 − , 0.2525 + ,
20 20
 
1−r 1−r
xe5 (r ) = 0.0785 − , 0.0785 + .
20 20

XUt (r ) will be:

Also, e
 
1−r 1−r
x1 (r ) = 0.0053 −
e , 0.0053 + ,
15 15
 
1−r 1−r
xe2 (r ) = 0.6519 − , 0.6519 + ,
15 15
 
1−r 1−r
xe3 (r ) = 0.3659 − , 0.3659 + ,
15 15
 
1−r 1−r
xe4 (r ) = 0.2525 − , 0.2525 + ,
15 15
 
1−r 1−r
xe5 (r ) = 0.0785 − , 0.0785 + ,
15 15

Xλt (r ) will be:

and e
(1 − r )(1 + λ) (1 − r )(1 + λ)
 
xe1 (r ) = 0.0053 − , 0.0053 + ,
60 60
(1 − r )(1 + λ) (1 − r )(1 + λ)
 
xe2 (r ) = 0.6519 − , 0.6519 + ,
60 60
(1 − r )(1 + λ) (1 − r )(1 + λ)
 
xe3 (r ) = 0.3659 − , 0.3659 + ,
60 60
(1 − r )(1 + λ) (1 − r )(1 + λ)
 
xe4 (r ) = 0.2525 − , 0.2525 + ,
60 60
(1 − r )(1 + λ) (1 − r )(1 + λ)
 
xe5 (r ) = 0.0785 − , 0.0785 + .
60 60

X (r ) = (e
Also, exact solution [22], is: e x1 (r ), . . . , xe5 (r ))t , 0 ≤ r ≤ 1 where,
xe1 (r ) = [−0.04091 + 0.04622r , 0.051526 − 0.04622r] ,
xe2 (r ) = [0.61305 + 0.03885r , 0.69074 − 0.03885r] ,
xe3 (r ) = [0.31945 + 0.04647r , 0.41236 − 0.04647r] ,
xe4 (r ) = [0.18541 + 0.67110r , 0.31396 − 0.67110r] ,
xe5 (r ) = [−0.00105 + 0.07957r , 0.15807 − 0.07957r] .
Xλ = (xλ,1 , . . . , xλ,5 )t in compare of exact solution e
In addition, we plot error of approximate solution e X = (x1 , . . . , x5 )t (see
Fig. 2).

5. Conclusion

In this paper, we proposed a simple and practical method to solve fuzzy linear systems. Initially, we obtain the 1-cut
solution of fuzzy linear system. This solution is exists, since matrix A is assumed nonsingular. Then, we fuzzify the 1-cut
solution with unknown symmetric spreads. To obtain these unknown spreads, we should resolve each equations of interval
system (4). Consequently, we proposed three type of symmetric solution where one of them has pessimistic/optimistic
attitude which will be determined by decision maker. It seems that this is new point of view in theory of fuzzy linear system
to obtain fuzzy solution such that decision maker can analyze the system in desire positions. This procedure ensure that our
method always give us fuzzy vector solution. Also by proposed method we obtain maximal and minimal symmetric solutions
which are placed in CSS and TSS, respectively. Clearly, symmetric solution has important role in theory of approximation
of fuzzy numbers. For future work, we try to extend our method to solve a fuzzy linear system when, at least one of the
elements on the right-hand side is a trapezoidal fuzzy number.

Please cite this article in press as: T. Allahviranloo, S. Salahshour, Fuzzy symmetric solutions of fuzzy linear systems, Journal of Computational and
Applied Mathematics (2010), doi:10.1016/j.cam.2010.02.042
 ARTICLE IN PRESS
T. Allahviranloo, S. Salahshour / Journal of Computational and Applied Mathematics ( ) – 9

Fig. 2. Error of approximate solution e

Xλ in Example 4.2.

References

[1] J.J. Buckley, Y. Qu, Solving system of linear fuzzy equations, Fuzzy Sets and Systems 43 (1991) 33–43.
[2] J.J. Buckley, Y. Qu, Solving linear and quadratic fuzzy equations, Fuzzy Sets and Systems 38 (1990) 43–59.
[3] J.J. Buckley, Y. Qu, Solving fuzzy equations: a new concept, Fuzzy Sets and Systems 39 (1991) 291–301.
[4] M. Friedman, Ma. Ming, A. Kandel, Fuzzy linear system, Fuzzy Sets and Systems 96 (1998) 209–261.
[5] Ma. Ming, M. Friedman, A. Kandel, Duality in fuzzy linear systems, Fuzzy Sets and Systems 109 (2000) 55–58.
[6] S. Muzzilio, H. Reynaerts, Fuzzy linear system of the form A1 x + b1 = A2 x + b2 , Fuzzy Sets and Systems 157 (2006) 939–951.
[7] T. Allahviranloo, Numerical methods for fuzzy system of linear equationa, Appl. Math. Comput. 155 (2004) 493–502.
[8] T. Allahviranloo, The Adomian decomposition method for fuzzy system of linear equations, Appl. Math. Comput. 163 (2005) 553–563.
[9] T. Allahviranloo, Successive over relaxation iterative method for fuzzy system of linear equations, Appl. Math. Comput. 162 (2005) 189–196.
[10] S. Abbasbandy, R. Ezzati, A. Jafarian, LU decomposition method for solving fuzzy system of equations, Appl. Math. Comput. 172 (2006) 633–643.
[11] S. Abbasbandy, A. Jafarian, Steepest decent method for system of fuzzy linear equations, Appl. Math. Comput. 175 (2006) 823–833.
[12] T. Allahviranloo, A comment on fuzzy linear systems, Fuzzy Sets and Systems 140 (2003) 559.
[13] T. Allahviranloo, M. Afshar Kermani, Solution of a fuzzy system of linear equation, Appl. Math. Comput. 175 (2006) 519–531.
[14] M. Dehghan, B. Hashemi, Iterative solution of fuzzy linear systems, Appl. Math. Comput. 175 (2006) 645–674.
[15] K. Wang, B. Zhend, Inconsistent fuzzy linear systems, Appl. Math. Comput. 181 (2006) 973–981.
[16] X. Wanga, Z. Zhong, M. Ha, Iteration algorithms for solving a system of fuzzy linear equations, Fuzzy Sets and Systems 119 (2001) 121–128.
[17] B. Zheng, K. Wang, General fuzzy systems, Appl. Math. Comput. 181 (2006) 1276–1286.
[18] D. Dubois, H. Prade, Fuzzy Sets and Systems: Theory and Applications, Academic Press, 1980.
[19] R. Goetschel, W. Voxman, Elementary calculus, Fuzzy Sets and Systems 18 (1986) 31–43.
[20] H.J. Zimmermann, Fuzzy Sets Theory and Applications, Kluwer, Dorrecht, 1985.
[21] R. Baker Kearfott, Regorous Global Search: Continuous Problems, Kluwer Academic Publishers, The Netherlands, 1996.
[22] T. Allahviranloo, E. Ahmady, N. Ahmady, Kh. Shams Alketaby, Block Jacobi two stage method with Gauss–Siedel inner iterations for fuzzy system of
linear equations, Appl. Math. Comput. 175 (2006) 1217–1228.

Please cite this article in press as: T. Allahviranloo, S. Salahshour, Fuzzy symmetric solutions of fuzzy linear systems, Journal of Computational and
Applied Mathematics (2010), doi:10.1016/j.cam.2010.02.042