Distinct Methods for Solving Fully Fuzzy Linear Programming Problems with Pentagonal Fuzzy Numbers
D. Stephen Dinagar
1and M. Mohamed Jeyavuthin
2*1
Associate Professor,
PG & Research Department of Mathematics, T.B.M.L. College, Porayar, Tamil Nadu, INDIA.
2
Research Scholar,
PG & Research Department of Mathematics, T.B.M.L. College, Porayar, Tamil Nadu, INDIA.
email: [email protected], [email protected]
(Received on: May 8, 2019) ABSTRACT
The focus of this paper is to find fuzzy optimal solution of fully fuzzy linear programming problems with pentagonal fuzzy numbers. New approaches for solving fully fuzzy linear programming problems with pentagonal fuzzy number have been proposed, based on distinct ranking functions. The proposed methods are very easy to understand.
Keywords: Linear programming problem, Fully fuzzy linear programming, Pentagonal fuzzy number, Robust ranking.
1. INTRODUCTION
The concept linear programming problem is to find out the best solution to the real- world problems where the available information is not exact or not precise. In that situation linear programming model helps lot.
First, the concept Fuzzy linear programming was proposed by Tanaka et al.
20. It plays a vital role in Fuzzy modeling, which can formulate the uncertainty. Nasseri
17has proposed a new method for solving the Fuzzy linear programming problems in which he has used the fuzzy ranking method for converting the fuzzy objective function into crisp objective function.
Fuzzy linear programming was studied by many researchers
23,20,7,6,5. Liou and Wang
15discussed ranking fuzzy numbers with interval values. Verdegay
22have proposed three
methods for solving three models of fuzzy integer linear programming. Abbasbandy and
Asady
1discussed a sign distance method for ranking fuzzy numbers. In 1998, Cheng
9used a centroid based distance method. Kauffmann and Gupta
11proposed Fuzzy Arithmetic and many researchers has studied in the relevant field
18,16,14,8,4.
In 1984, Kolman and Hill
12introduced fully fuzzy linear programming problem. A new method for solving fully fuzzy linear programming problems using ranking function is proposed by Allahviranloo et al.
2. Kumar et al.
13proposed a new method for solving fully fuzzy linear programming problems with inequality constraints. Apurba Panda and Madhumangal Pal
3discussed a detailed study on pentagonal fuzzy number. Dhurai and Karpagam
10proposed a new method for fully fuzzy linear programming problems with hexagonal fuzzy numbers by using ranking function. In
19Stephen Dinagar and Mohamed Jeyavuthin discussed the concept of solving fully fuzzy integer linear programming problems under robust ranking technique.
The structure of this paper is, section 1 is introduction. In Section 2, some basic concepts have been reviewed. In Section 3, with the help of proposed ranking function algorithms of fully fuzzy linear programming problem have been presented. Section 4 is conclusion.
2 PRELIMINARIES
Definition 1 (Fuzzy set)
A Fuzzy set 𝐴̃ is defined by 𝐴̃ = {(𝑥, 𝜇
𝐴̃(𝑥)): 𝑥 ∈ 𝑋, 𝜇
𝐴̃(𝑥) ∈ [0,1] }. In the pair (𝑥, 𝜇
𝐴̃(𝑥)), the first element 𝑥 belong to the classical set 𝑋, the second element 𝜇
𝐴̃(𝑥) belong to the interval [0, 1], called Membership function.
Definition 2 (Support of Fuzzy set)
The support of fuzzy set 𝐴̃ is the set of all 𝑥 in 𝑋 such that 𝜇
𝐴̃(𝑥) > 0. That is 𝑆𝑢𝑝𝑝 (𝜇
𝐴̃) = {𝑥/ 𝜇
𝐴̃(𝑥) > 0}.
Definition 3 (𝜶-cut)
The𝛼-cut of fuzzy set 𝐴̃ is a set consisting of those elements of the universe 𝑋 whose membership values exceed the threshold level 𝛼.
That is 𝐴̃
𝛼= {𝑥 / 𝜇
𝐴̃(𝑥) ≥ 𝛼}.
Definition 4 (Convex Fuzzy set)
A fuzzy set 𝐴̃ is convex if 𝜇
𝐴̃(𝜆𝑥
1+ (1 − 𝜆𝑥
2)) ≥ min(𝜇
𝐴̃(𝑥
1), 𝜇
𝐴̃(𝑥
1)) , 𝑥
1, 𝑥
2∈ 𝑋 and 𝜆 ∈ [0,1]. Alternatively, a fuzzy set is convex, if all 𝛼- level sets are convex.
Definition 5 (Fuzzy Number)
A fuzzy number 𝐴̃ is a subset of real line R, with the membership function 𝜇
𝐴̃(𝑥) holds the following conditions:
(i) 𝜇
𝐴̃(𝑥) is piecewise continuous in its domain
(ii) 𝐴̃ is normal. That is, there is a 𝑥
0∈ 𝐴̃ such that 𝜇
𝐴̃(𝑥
0) = 1.
(iii) 𝐴̃ is convex. That is,𝜇
𝐴̃(𝜆𝑥
1+ (1 − 𝜆𝑥
2)) ≥ min(𝜇
𝐴̃(𝑥
1), 𝜇
𝐴̃(𝑥
1)) , 𝑥
1, 𝑥
2∈ 𝑋
Definition 6 (Pentagonal Fuzzy Number)
A fuzzy number 𝐴̃
𝑃is pentagonal fuzzy number denoted by 𝐴̃
𝑃= (𝑎
1, 𝑎
2, 𝑎
3, 𝑎
4, 𝑎
5), where 𝑎
1, 𝑎
2, 𝑎
3, 𝑎
4, 𝑎
5are real numbers and its membership function 𝜇
𝐴̃𝑃(𝑥) is given by
𝜇
𝐴̃𝑃(𝑥) =
{
0, 𝑥 < 𝑎
11
2 [ 𝑥 − 𝑎
1𝑎
2− 𝑎
1], 𝑎
1≤ 𝑥 ≤ 𝑎
21
2 + 1
2 [ 𝑥 − 𝑎
2𝑎
3− 𝑎
2], 𝑎
2≤ 𝑥 ≤ 𝑎
31, 𝑥 = 𝑎
31
2 + 1
2 [ 𝑎
4− 𝑥
𝑎
4− 𝑎
3], 𝑎
3≤ 𝑥 ≤ 𝑎
41
2 [ 𝑎
5− 𝑥
𝑎
5− 𝑎
4], 𝑎
4≤ 𝑥 ≤ 𝑎
50, 𝑥 > 𝑎
5Definition 7
A pentagonal fuzzy number can be defined as 𝐴̃
𝑃= (𝑀
1(𝑥),𝐽
1(𝑥),𝐽
2(𝑥),𝑀
2(𝑥)) for 𝑥 ∈ [0,1]
where,
(i) 𝑀
1(𝑥)is strictly increasing continuous function on [0,0.5]
(ii) 𝐽
1(𝑥)is strictly increasing continuous function on [0.5,1]
(iii) 𝐽
2(𝑥)is strictly decreasing continuous function on [1,0.5]
(iv) 𝑀
2(𝑥) is strictly decreasing continuous function on [0.5,0]
Remark 8
The pentagonal fuzzy number 𝐴̃
𝑃becomes triangular fuzzy number if 𝑎
3− 𝑎
2= 𝑎
4− 𝑎
3.
Figure:1 Graph of a Pentagonal Fuzzy Number
3 FULLY FUZZY LINEAR PROGRAMMING PROBLEM (FFLPP) Consider the fully fuzzy linear programming problem
Maximize (or Minimize) 𝑍̃ = 𝐶̃𝑥̃, Subject to 𝐴̃𝑥̃ ≤, =, ≥ 𝑏̃ , 𝑥̃ ≥ 0
where 𝑥̃ is a non-negative pentagonal fuzzy number and 𝐶̃ = [𝑐̃
𝑗]
1×𝑛, 𝐴̃ = [𝑎̃
𝑖𝑗]
𝑚×𝑛𝑥̃ = [𝑥̃
𝑗]
𝑛×1, 𝑏̃ = [𝑏̃
𝑗]
𝑚×1, 1 ≤ 𝑗 ≤ 𝑛 and 1 ≤ 𝑖 ≤ 𝑚.
3.1 Method I
In this method we are finding fuzzy optimal solution for fully fuzzy linear programming problem with pentagonal fuzzy number with the aid of average ranking function.
3.2 Algorithm
Step 1. Arrange the fully fuzzy linear programming problem in the following manner, Max(or)Min 𝑧̃ = 𝑐̃
𝑗𝑥̃
𝑗; Subject to𝐴̃
𝑖𝑗𝑥̃
𝑗≤ 𝑏̃
𝑗; 𝑥̃
𝑗≥ 0.
Step 2. Convert the problem into fuzzy variable linear programming problem by using the ranking function 𝔑(𝐴̃
𝑃) =
𝑎1+𝑎2+𝑎53+𝑎4+𝑎5.
Step 3. Applying the simplex / Big-M procedure. to solve the fuzzy variable linear programming problem. Let the solution be 𝑥̃
𝑗. Hence the solution of the fully fuzzy linear programming problem 𝑥̃
𝑗∗.
Step 4.Convert all the inequations to the constraints into equations by introducing the slack and/or surplus variables in the constrains.
Step 5.To compute the net evaluations using the relation 𝑍̃ = 𝐶
𝐵𝑌
𝑗− 𝐶
𝑗; 𝑗 = 1,2, … , 𝑛.
If, all 𝑍̃ ≥ 0for maximization problem,𝑍̃ ≤ 0 for minimization problem, then optimal solution is reached.
Step 6.Determine the basic variable 𝑥̃
𝑘, it will be replaced by the non-basic variable, where
𝑘 = arg min {𝔑(𝐵̃̃
𝑖)} , 𝑖 = 1,2, … , 𝑚 in maximization and 𝑘 = arg max {𝔑(𝐵̃̃
𝑖)}, 𝑖 = 1,2, … , 𝑚 for minimization problem.
Step 7. perform the pivot operation and return to step 5. Then repeat the procedure until a fuzzy optimal solution is obtained.
3.3 Method II
In this method we are finding fuzzy optimal solution for fully fuzzy linear programming problem with pentagonal fuzzy number. This method is used to convert the inequality constraints in to equality constraints.
3.4 Algorithm:
Step 1. To Check type of all the constraints, convert the inequality constraints into equality constraints. ∑
𝑛𝑗=1𝑎̃
𝑖𝑗× 𝑥̃
𝑗= 𝑏̃
𝑖𝑜𝑟 ∑
𝑛𝑗=1𝑎̃
𝑖𝑗× 𝑥̃
𝑗≠ 𝑏̃
𝑖∀ 𝑖 = 1,2, … , 𝑚
If ∑
𝑛𝑗=1𝑎̃
𝑖𝑗× 𝑥̃
𝑗≠ 𝑏̃
𝑖for some 𝑖, then the following cases arise:
Case (i) If ∑
𝑛𝑗=1𝑎̃
𝑖𝑗× 𝑥̃
𝑗≤ 𝑏̃
𝑖for some 𝑖, then convert such type of inequality constraints into
equality constraints by introducing non-negative variable 𝑆̃
𝑖to the left side of the constraints.
∑
𝑛𝑗=1𝑎̃
𝑖𝑗× 𝑥̃
𝑗+ 𝑆̃
𝑖= 𝑏̃
𝑖for some 𝑖, where 𝑆̃
𝑖is a non-negative pentagonal fuzzy number.
Case (ii) If ∑
𝑛𝑗=1𝑎̃
𝑖𝑗× 𝑥̃
𝑗≥ 𝑏̃
𝑖for some 𝑖, then convert such type of inequality constraints into equality constraints by introducing non-negative variable 𝑆̃
𝑖to the right side of the constraints. ∑
𝑛𝑗=1𝑎̃
𝑖𝑗× 𝑥̃
𝑗= 𝑏̃
𝑖+ 𝑆̃
𝑖for some 𝑖, where 𝑆̃
𝑖is a non-negative pentagonal fuzzy number.
Step 2. Let Maximize (or Minimize) 𝐶̃
𝑇× 𝑋̃ Subject to 𝐴̃𝑋̃ ≤, =, ≥ 𝑏̃ where 𝑋̃ is a non- negative pentagonal fuzzy number, be converted into Maximize (or Minimize) 𝐶̃
𝑇× 𝑋̃ Subject to 𝐴̃𝑋̃ = 𝑏̃ where 𝑋̃ is a non-negative pentagonal fuzzy number, and 𝐶̃ = [𝑐̃
𝑗]
1×𝑛, 𝐴̃ = [𝑎̃
𝑖𝑗]
𝑚×𝑛, 𝑋̃ = [𝑥̃
𝑗]
𝑛×1, 𝑏̃ = [𝑏̃
𝑖]
𝑚×1, 1 ≤ 𝑗 ≤ 𝑛 and 1 ≤ 𝑖 ≤ 𝑚.
Step 3. Substituting 𝐶̃ = [𝑐̃
𝑗]
1×𝑛, 𝐴̃ = [𝑎̃
𝑖𝑗]
𝑚×𝑛, 𝑋̃ = [𝑥̃
𝑗]
𝑛×1, 𝑏̃ = [𝑏̃
𝑖]
𝑚×1into the fully fuzzy linear programming problem, obtained in step 2, then the fully fuzzy linear programming problem may be written as, Maximize (or Minimize) ∑
𝑛𝑗=1(𝐶̃
𝑇× 𝑋̃) subject to
∑
𝑛𝑗=1𝑎̃
𝑖𝑗× 𝑥̃
𝑗= 𝑏̃
𝑖∀ 𝑖 = 1,2, … , 𝑚, 𝑥̃
𝑗is a non-negative pentagonal fuzzy number.
Step 4. If all the parameters 𝑐̃
𝑗, 𝑥̃
𝑗, 𝑎̃
𝑖𝑗and 𝑏̃
𝑖are represented by pentagonal fuzzy numbers (𝑝
𝑗, 𝑞
𝑗, 𝑟
𝑗, 𝑠
𝑗, 𝑡
𝑗), (𝑥
𝑗, 𝑦
𝑗, 𝑧
𝑗, 𝑢
𝑗, 𝑣
𝑗), (𝑎
𝑖𝑗, 𝑏
𝑖𝑗, 𝑐
𝑖𝑗, 𝑑
𝑖𝑗, 𝑒
𝑖𝑗) and (𝑏
𝑖, 𝑔
𝑖, ℎ
𝑖, 𝑘
𝑖, 𝑙
𝑖) respectively then the fully fuzzy linear programming problem written as,
Maximize (or Minimize) ∑
𝑛𝑗=1(𝑝
𝑗, 𝑞
𝑗, 𝑟
𝑗, 𝑠
𝑗, 𝑡
𝑗) × (𝑥
𝑗, 𝑦
𝑗, 𝑧
𝑗, 𝑢
𝑗, 𝑣
𝑗) subject to
∑
𝑛𝑗=1(𝑎
𝑖𝑗, 𝑏
𝑖𝑗, 𝑐
𝑖𝑗, 𝑑
𝑖𝑗, 𝑒
𝑖𝑗) × (𝑥
𝑗, 𝑦
𝑗, 𝑧
𝑗, 𝑢
𝑗, 𝑣
𝑗) = (𝑏
𝑖, 𝑔
𝑖, ℎ
𝑖, 𝑘
𝑖, 𝑙
𝑖) ∀ 𝑖 = 1,2, … , 𝑚, (𝑥
𝑗, 𝑦
𝑗, 𝑧
𝑗, 𝑢
𝑗, 𝑣
𝑗) is a non-negative pentagonal fuzzy number.
Step 5. Assuming (𝑎
𝑖𝑗, 𝑏
𝑖𝑗, 𝑐
𝑖𝑗, 𝑑
𝑖𝑗, 𝑒
𝑖𝑗) × (𝑥
𝑗, 𝑦
𝑗, 𝑧
𝑗, 𝑢
𝑗, 𝑣
𝑗) = (𝑎
𝑖𝑗′, 𝑏
𝑖𝑗′, 𝑐
𝑖𝑗′, 𝑑
𝑖𝑗′, 𝑒
𝑖𝑗′) the fully fuzzy linear programming problem obtained in the previous step may be written as,
Maximize (or Minimize) ℛ (∑
𝑛𝑗=1(𝑝
𝑗, 𝑞
𝑗, 𝑟
𝑗, 𝑠
𝑗, 𝑡
𝑗) × (𝑥
𝑗, 𝑦
𝑗, 𝑧
𝑗, 𝑢
𝑗, 𝑣
𝑗)), where ℛ =
𝑎1+𝑎2+2𝑎3+𝑎4+𝑎56
subject to ∑
𝑛𝑗=1(𝑎
𝑖𝑗′, 𝑏
𝑖𝑗′, 𝑐
𝑖𝑗′, 𝑑
𝑖𝑗′, 𝑒
𝑖𝑗′) = (𝑏
𝑖, 𝑔
𝑖, ℎ
𝑖, 𝑘
𝑖, 𝑙
𝑖) ∀ 𝑖 = 1,2, … , 𝑚, (𝑥
𝑗, 𝑦
𝑗, 𝑧
𝑗, 𝑢
𝑗, 𝑣
𝑗) is a non-negative pentagonal fuzzy number.
Step 6. Find the optimal solution 𝑥
𝑗, 𝑦
𝑗, 𝑧
𝑗, 𝑢
𝑗and 𝑣
𝑗by solving the crisp linear programming problem obtained in step 5.
3.5 Method III
In this method we are finding fuzzy optimal solution for fully fuzzy linear programming problem with pentagonal fuzzy number with the aid of Robust ranking function.
where the Robust Ranking Technique satisfies the properties such as compensation,
linearity, additivity. It provides results which are consist human intuition. If 𝐴̃
𝑃is a pentagonal fuzzy number then the Robust ranking is defined by
R(𝐴̃
𝑃) = ∫(𝐴̃
𝑃𝐿, 𝐴̃
𝑃𝑈)(0.5)
1
0
𝑑𝛼
where (𝐴̃
𝑃𝐿, 𝐴̃
𝑃𝑈) is the 𝛼 level cut of the fuzzy number 𝐴̃
𝑃.
Figure 2
Pentagon is divided into four parts which includes two triangles and two trapezoids.
By using robust ranking technique for triangle fuzzy number and trapezoidal fuzzy number the ranking function is
R
1(𝐴̃
𝑃) = ∫{(𝑎
2− 𝑎
1)𝛼 + 𝑎
1, 𝑎
5− (𝑎
5− 𝑎
4)𝛼}
1
0
(0.5) 𝑑𝛼
3.6 Algorithm
Step 1. Formulate the chosen problem into the following fuzzy linear programming problem as
Maximize 𝑧̃ = ∑
𝑛𝑗=𝑖𝑐̃
𝑗𝑥̃
𝑗Subject to, ∑
𝑛𝑗=𝑖𝑎̃
𝑖𝑗𝑥̃
𝑗≤, =, ≥ 𝑏̃
𝑖, i=1,2,3,… ,m, 𝑥̃
𝑗≥ 0, 𝑗 = 1,2,3, … , 𝑛
Step 2.Using the Ranking functions, the fully fuzzy integer linear programming problem (FFILPP) transformed into fuzzy variable integer linear programming problem (FVILPP).
Step 3. Optimal solution.
4. CONCLUSION
The proposed three methods provide solution to fully fuzzy linear programming problems with pentagonal fuzzy number. Some distinct ranking methods have been employed in each method for solving the optimal solutions. This notion can be extended to some other optimization problems, especially integer programming problems in future.
Figure 2 𝜇𝐴̃𝑃(𝑥)
C 0.75
1
𝐽2(𝑥) 𝐽1(𝑥)
𝑀2(𝑥) 𝑀1(𝑥)