International Journal of Algorithms, Computing and Mathematics

Volume 3, Number 4, November 2010

©Eashwar Publications

An Algorithmic Approach to Multi Objective Fuzzy Linear

Programming Problem

R. Sophia Porchelvi
Department of Mathematics, A. D. M. College for Women (Autonomous),
Nagappattinam-611 001, Tamil Nadu, India
[email protected]

A.Nagoorgani
PG & Research Department of Mathematics, Jamal Mohamed College (Autonomous),
Tiruchirappalli-620020, Tamil Nadu, India
[email protected]

R.Irene Hepzibah
Department of Mathematics, A.V.C.College of Engineering,
Mayiladuthurai - 609305, Tamil Nadu, India
[email protected]

Abstract
In this Paper, a ranking procedure based on fuzzy numbers distance is applied to
Multi -objective Linear Programming Problem with fuzzy coefficients. Here the
problem is transformed into a Crisp problem and then solved by Preemptive
optimization method. Numerical Examples are provided to illustrate its feasibility.

Keywords: Distance, Ranking, Triangular and Trapezoidal fuzzy numbers, Multi-objective

Linear Programming, Preemptive Optimization.

1. Introduction

Multi-objective optimization is the process of simultaneously optimizing two or more

conflicting objectives subject to certain constraints. In many real world problems, there are
situations where multiple objectives may be more appropriate rather than considering single
objective. However, in such cases emphasis is on efficient solutions, which are optimal in a
certain multi-objective sense. Bellman and Zadeh proposed the concepts of decision making
in fuzzy environments [1]. Qiu - Peng Gu, and Bing - Yuan Cao solved Fuzzy Linear
programming problems based on Fuzzy numbers distance [5]. Zimmermann [8] presented a
fuzzy approach to solve multi-objective linear programming problems. Tong Shaocheng[7]
focused on the fuzzy linear programming problems with interval numbers. In this paper, an
attempt is made to solve Multi-objective Linear Programming Problem (MOFLPP) under
constraints with Fuzzy coefficients. Here, the MOFLPP is transformed into a Multi-objective
Linear Programming Problem (MOLPP) and it can be solved accordingly. The paper is
organized as follows: Certain definitions and Notations are provided in the next section. The
Ranking algorithm is given in section 3. Section 4 presents the method for solving a

61
International Journal of Algorithms, Computing and Mathematics

MOFLPP. In section 5, numerical examples are discussed. The last section draws some
concluding remarks.

2. Definitions and Notations

2.1 Interval Number

Let R be the set of real numbers. Then closed interval [a, b] is said to be an interval number,
where a, b R, a< b.

2.2 Distance between Interval Numbers

Let a = [a1, a2], b = [b1, b2] be two interval numbers. Then the distance between a, b, denoted
by d (a, b), is defined by
1
2
a1 a2 b1 b2
d ( a, b) x(a2 a1 ) x(b2 b1 ) dx
1 2 2
2

2.3 Fuzzy Number and Membership Function

The fuzzy set A in the set of real numbers is called a fuzzy number. If its membership
function is
a x
L 2 , a x a2 ,
a2 a1 1
1, a2 x a3 ,
A
( x) ……… (2.1)
x a3
R ,a x a4 ,
a4 a3 3
0, x < a1 , x > a 4 ,
where L and R are strictly decreasing functions in (0,1), and satisfy
L(x) = R(x) = 1 (x 0); L(x) = R(x) = 0 (x 1).
When L(x) = R(x) = 1- x, fuzzy number defined in (2.1) is a Trapezoidal fuzzy number,
denoted by A = (a1,a2,a3,a4); when L(x) = R(x) = 1- x and a2 = a3, fuzzy number defined in
(2.1) is a Triangular fuzzy number, denoted by A = (a1,a2,a3).

2.4 - Level cut

For any [0, 1], - level cut of fuzzy number à is an interval number
1 a2 x
à = [aL( ), aR( )]. Let LA denote the inverse function of à = L about the
a2 a1
a2 x 1 x a3
variant ; RA denote the inverse function of à = R about the
a2 a1 a4 a3
x a3
variant . So, aL( ) = a2 - (a2 - a1) LA 1
; aR( ) = a3 - (a4 – a3) RA 1
.
a4 a3

62
An Algorithmic Approach to Multi Objective Fuzzy Linear Programming Problem

2.5 Distance between Fuzzy Numbers

Let Ã, B be two fuzzy numbers. The distance between the fuzzy numbers à and B is defined
1
by D( A, B) d ( A , B )d , ……… (2.2)
0

3. Ranking Algorithm

Qiu – peng Gu and Bing – Yuan Cao [1] developed a ranking procedure by defining a
distance between fuzzy numbers (see [1]). Based on this ranking procedure, a ranking
algorithm is developed for Triangular and Trapezoidal Fuzzy numbers. Moreover, it is
applied to MOFLPP under constraints with fuzzy coefficients.

Algorithm:
~
Step 1: Consider the fuzzy numbers à & B which are either triangular or trapezoidal.
Step 2: Find supremum M = sup (s (A) s (B)), where s (A) = support set of à and s
~
(B) = support set of B .
~
Step 3: If à and B are triangular fuzzy numbers, then go to step 6.
~
Step 4: Otherwise, Take à = (a1, a2, a3, a4) & B = (b1, b2, b3, b4) as trapezoidal fuzzy
numbers.
a1 a2 a3 a4 b b b b
Step 5: Calculate D( A, M ) M & D( B, M ) M 1 2 3 4
4 4
and go to step 8.
~
Step 6: Take à = (a1, a2, a3) and B = (b1, b2, b3).
a1 2a2 a3 b 2b2 b3
Step 7: Calculate D( A, M ) M & D( B, M ) M 1
4 4
~ ~
Step 8: If D (Ã, M) < D ( B , M) then à is greater than B
~ ~
If D (Ã, M) = D ( B , M) then à is equal to B
~ ~
If D (Ã, M) > D ( B , M) then à is less than B
Step 9: Stop.

4. Method of Solving Multi-objective Fuzzy Linear Programming Problem

In this paper, we discuss a multi-objective model in the constraint conditions with fuzzy
coefficients. Although a multi-objective model may have several objective criteria, they are
rarely of equal importance. Hence, they can be considered in priority order with respect to the
problem undertaken. Moreover, the objectives may be of all minimization type or all
maximization type or mixed with both minimization and maximization types. In this paper,
we discuss a model whose standard form is
Maximize Z1= c1 x
Minimize Z2 = c2 x, ………
subject to AX b, x 0,
where ci = (ci1 ,ci2 ,.......,cin ) is an n-dimensional crisp row vector, Ã = aij is an m x n fuzzy
T
number matrix, b b1 , b2 ,....., bm is an m-dimensional fuzzy line vector and
X ( x1 , x2 ,......, xn )T is an n-dimensional decision variable vector.

63
International Journal of Algorithms, Computing and Mathematics

We now consider a bi-objective Fuzzy Linear Programming Problem with constraints having
fuzzy coefficients is given by

Maximize Z1 = c11x1 + c12 x 2 + ....... + c1n x n

Minimize Z2 = c21x1 + c22 x 2 + ....... + c2n x n
subject to ã i1x1 + ã i2 x 2 +.......+ ã in x n bi ,
x1 , x 2 ,........,x n 0 , i = 1, 2, ...... , m.
where fuzzy numbers are triangular fuzzy ones, that is, ã i1 = ã i11,ã i12 ,ã i13 ,

ã i2 = ã i21,ã i22 ,ã i23 ,....... ,ã in = ã in1,ã in2 ,ã in3 ; bi = bi1,bi2 ,bi3 .

By Zadeh’s Extension principle, the sum of any triangular fuzzy numbers is still a triangular
one. The above formulation can be written as
Maximize Z1 = c11x1 + c12 x 2 + ....... + c1n x n
Minimize Z2 = c21x1 + c22 x 2 + ....... + c2n x n
subject to (a i11x1 +a i21x 2 +......+a in1x n , a i12 x1 +a i22 x 2 +......+a in2 x n ,
a i13x1 +a i23x 2 +......+a in3x n ) (bi1,bi2 ,bi3 ),
x1, x 2 ,......,x n 0 , i = 1,2, .....,m.

By the ranking algorithm, the above MOFLPP is transformed into a MOLPP as follows:
Maximize Z1 = c11x1 + c12 x 2 + ....... + c1n x n
Minimize Z2 = c21x1 + c22 x 2 + ....... + c2n x n
subject to (a i11x1 +a i21x 2 +......+a in1x n ) + 2(a i12 x1 +a i22 x 2 +......+a in2 x n )
+ (a i13 x1 +a i23x 2 +......+a in3x n ) bi1 + 2bi2 +bi3 ,
x1, x 2 ,......,x n 0 , i = 1,2, .....,m. ........ I

By the same ranking algorithm, a MOFLPP using trapezoidal fuzzy numbers can similarly be
transformed into a MOLPP as follows:
Maximize Z1 = c11x1 + c12 x 2 + ....... + c1n x n
Minimize Z2 = c21x1 + c22 x 2 + ....... + c2n x n
subject to (a i11x1 +a i21x 2 +......+a in1x n ) + 2(a i12 x1 +a i22 x 2 +......+a in2 x n )
+ 2(a i13x1 +a i23x 2 +......+a in3x n ) + (a i14 x1 +a i24 x 2 +......+a in4 x n )
bi1 +2bi2 +2bi3 +bi4 , x1 , x 2 ,......,x n 0 , i = 1,2,....., m. ........ II

In this method, the most important objective subject to the fuzzy constraints is considered
first. Using I or II, this can be converted into a single objective problem subject to constraints
with transformed crisp number coefficients and hence solved accordingly. Then the next
objective is optimized subject to a requirement that the first achieve its optimal value which
is nothing but the preemptive optimization.

Similarly, multi-objective problems with more than two objectives can also be solved using
the above procedure subject to a requirement that the previous objectives achieve its optimal
value. Here, in the very first stage itself the problem is transformed into a crisp problem and
afterwards there will be no more fuzziness in the constraints as well as in the problem.

64
An Algorithmic Approach to Multi Objective Fuzzy Linear Programming Problem

5. Numerical Examples

Consider the following MOFLPP

Maximize z1 = x1 + x 2
Minimize z 2 = 2x1 + 3x 2
subject to ã11x1 + ã12 x 2 b1
ã 21 x1 + ã 22 x 2 b2 , x1 ,x 2 0,
where ã11 =(2,3,4), ã12 =(1.5, 2, 3), b1 =(4,5,7 ), ã 21 =(-1,0,1), ã 22 =(0.5, 1, 1,5), b 2 = (1.5, 2, 3).
We assume that the maximization objective has a higher priority. Therefore, we consider the
problem first with the maximization objective.
Maximize z1 = x1 + x 2
subject to (2, 3, 4) x1 +(1.5, 2, 3) x 2 (4,5,7 )
(-1, 0, 1)x1 +(0.5,1,1.5) x 2 (1.5, 2, 3), x1 , x 2 0.

By I, this can be transformed into a Crisp LPP as

Maximize z1 = x1 + x 2
subject to (2x1 + 1.5x 2 ) + 2(3x1 + 2x 2 ) + (4x1 + 3x 2 ) 4 + 10 + 7,
(-x1 + 0.5x 2 ) + 2(0x1 + x 2 ) + (x1 + 1.5x 2 ) 1.5 + 4 + 3,
That is, 12x1 + 8.5x 2 21, 4x 2 8.5, x1 , x 2 0.
The solution is x1 = 0 , x 2 = 2.4706 and Max z1 = 2.4706.

Next, we proceed to solve the problem with minimization objective and an additional
constraint x1 + x2 2.4706, obtained from previous step.

Minimize z 2 = 2x1 + 3x 2
subject to 12x1 +8.5x 2 21, 4x 2 8.5, x1 + x 2 2.4706, x1 ,x 2 0.
Now, the optimal solution is x1 = 0 , x 2 = 2.4706 and Min z2 = 7.4118.

The feasibility of the solution will be explained as follows:

When all the fuzzy numbers are in the smallest possible values, the MOLPP becomes
Maximize z1 = x1 + x 2 , Minimize z2 = 2x1 + 3x 2
subject to 2x1 +1.5x 2 4, -x1 + 0.5x 2 1.5, x1 , x 2 0.
The optimal solution to the above problem is x1 = 0 , x 2 = 8/3, Max z1 = 8/3 and Min z2 = 8.

When all the fuzzy numbers are in their peak values, the MOLPP takes the form
Maximize z1 = x1 + x 2 , Minimize z2 = 2x1 + 3x 2
subject to 3x1 + 2x 2 5, x 2 2, x1 , x 2 0.
The optimal solution to the above problem is x1 = 1/3, x 2 = 2, Max z1 = 7/3 and Min z2 = 20/3.

When all the fuzzy numbers are in the highest possible values, the MOLPP becomes
Maximize z1 = x1 + x 2 , Minimize z2 = 2x1 + 3x 2
subject to 4x1 + 3x 2 7, x1 + 1.5x 2 3, x1 , x 2 0.
The optimal solution to the above problem is x1 = 1/2, x 2 = 5/3, Max z1 = 13/6 and Min z 2 = 6.

65
International Journal of Algorithms, Computing and Mathematics

Moreover, x1 [0,1/2], x 2 [5/3,8/3], Max z1 [7/3,8/3] and Min z2 [6,8], so the solution is
feasible and the same problem can also be solved with trapezoidal fuzzy numbers by using II.

Consider the following Bi-objective FLPP model with both the objectives of minimization
type.
Minimize z1 = 10x1 + 20x 2
Minimize z 2 = 20x1 + 30x 2
subject to ã11x1 + ã12 x 2 b1
ã 21 x1 + ã 22 x 2 b2 , x1 ,x 2 0,
where ã11 =(1,1,2), ã12 =(0,1,2), b1 =(4,5,7 ), ã 21 =(1,3,5), ã 22 =(3,5,5), b 2 = (6,7,8).

The optimal solution obtained by using the above procedure is given by x1 = 4.2, x2 = 0, Minimum
Value = 84. The feasibility of the solution is verified for the smallest, peak and highest values of the
fuzzy numbers. After analyzing, we get x1 [7/2,5], x2 [0,2/3] and the minimum lies in the interval
[70,100]. So, the solution is feasible and the same problem can also be solved with trapezoidal fuzzy
numbers.

6. Conclusion

In this paper, Multi-objective Fuzzy Linear Programming Problem under constraints with
fuzzy coefficients is considered. A specific ranking method based on distance between fuzzy
numbers is used for developing the ranking algorithm. By the Ranking algorithm, MOFLPP
using triangular fuzzy numbers is transformed into MOLPP and then solved by Preemptive
optimization method. Again it remains to research MOFLPP with general fuzzy numbers and
MOFLPP with fuzziness in objective functions.

References

[1] Bell man, R. and L.Zadeh (1970), “Decision Making in a fuzzy environment”, Management
Science, vol.17, 141-164.
[2] Chen S.H., Hsieh C.H., (1999), “Graded Mean Integration representation of generalized fuzzy
number”, Journal of Chinese Fuzzy systems, 5(2), 1-7.
[3] Qiu – Peng Gu., Bing – Yuan Cao, (2005), “Approach to linear programming with fuzzy
coefficients based on Fuzzy numbers distance”. IEEE Transactions, 447 – 450.
[4] Ronald L. Rardin (2003), “Optimization in Operations Research”, Pearson Education Pvt. Ltd.,
New Delhi.
[5]Tanaka.H,Asai.K., 1984. “Fuzzy linear programming problems with fuzzy numbers”, Fuzzy sets
and Systems, 13, 1-10.
[6]Tanaka.H,Tchinashi and Asai.K, 1984. “A Formulation of fuzzy linear programming problem
based on comparison of fuzzy numbers”, Control Cybernetics, 13(3), 184-194.
[7]Tong Shaocheng, 1994. “Interval Number and Fuzzy number linear programmings”, Fuzzy sets
and Systems, 66, 301-306.
[8]Zimmermann.H.J., 1976. “Discretion and optimization of fuzzy systems”, International Journal
General Systems, 2, 209-215.
[9] Zimmermann, H. (1991), “Fuzzy set theory and its applications”, Second edition, Kluwer
Academic Publishers, Germany.

66