2015 4th Iranian Joint Congress on Fuzzy and Intelligent Systems (CFIS)
DUAL SIMPLEX METHOD FOR SOLVING
FULLY FUZZY LINEAR PROGRAMMING
PROBLEMS
Akbar Hashemi Borzabadi Hossein Alemy
School of Mathematics and Computer Science School of Mathematics and Computer Science
Damghan University Damghan University
Damghan , Iran Damghan , Iran
borzabadi@[Link] Alemi.457935@[Link]
Abstract— This study proposes a novel technique for solving Ͳ ݔ൏ ݑ
linear programming problems in a fully fuzzy environment. A ݔۓെݑ
ۖ ݑ ݔ൏ ݒ
modified version of the well-known dual simplex method is used ݒെݑ
for solving fuzzy linear programming problems. The use of a ߤ ் ሺݔሻ ൌ ݓ െݔ
۔ ݒ ݔ ݓ
ranking function together with the Gaussian elimination process ۖ ݓെ ݒ
helps in solving linear programming problems in a fully uncertain ݔͲ ە ݓ
environment. The proposed algorithm is flexible, easy and Now, if T1 = (u1, v1, w1) and T2 = (u2, v2, w2) are two triangular
reasonable. fuzzy numbers ,then the basic arithmetic operations on the
triangular numbers are as given below:
Keywords— Fuzzy linear programming problems (FLP). Fuzzy
simplex method .Ranking and selection. 1. Image T = (u, v, w), -T = (-w, -v, -u)
2. Addition
I. INTRODUCTION T1 + T2 = (u1, v1, w1) + (u2, v2, w2) = (u1 + u2, v1 + v2, w1 + w2)
A Linear Programming Problem in a fully fuzzy 3. Substraction
environment can be defined as T1 + ( – T2) = (u1, v1, w1) + ( – w2, – v2, – u2) =
Min ݖǁ ൌ ܿǁ ௧ ݔ (u1 – w2, v1 – v2, w1 – u2)
s.t ܣሚݔ ൌ ܾ෨ 4. Multiplication k.T = (ku, kv, kw), k > 0
ݔ Ͳ k.T = (kw, kv, ku), k < 0
Here, ݖǁ , ܿǁ ௧ = (ܿǁଵ ,……, ܿǁ ), ݔ= (ݔଵ ,……, ݔ ), ܣሚ = ൣܽ ൧ൈ , T1.T2 = (u1, v1, w1).(u2, v2, w2) = (u1.u2, v1.v2, w1.w2)
5. Division ܶଵ ǣ ܶଶ = (ݑଵ , ݒଵ , ݓଵ ) : (ݑଶ , ݒଶ , ݓଶ )= (ݑଵ , ݒଵ , ݓଵ ) ൈ
ܾ෨ ൌ ሺܾ෨ଵ ǡ ǥ ǥ ǡ ܾ෨ ሻ௧ are the fuzzy objective function, fuzzy ଵ ଵ ଵ ௨ ௩ ௪
objective function coefficients, fuzzy decision variables, fuzzy ( ǡ ǡ ሻ ൌ ( భ ǡ భ ǡ భሻ
௪మ ௩మ ௨మ ௪మ ௩మ ௨మ
technical coefficients and fuzzy resource constraints of the A ranking function is a mapping from a fuzzy set to the set of
linear programming problem,respectively. real numbers ܴ . Define the ranks on ܴ as
In this study, ݖǁ denotes triangular fuzzy number, whereas 1. T1 t RT2 iff R(T1) t R(T2)
ܿǁ ௧ =(ܿǁଵ ,……, ܿǁ ) , ݔ= (ݔଵ ,……, ݔ ) , ܣሚ = ൣܽ ൧ൈ , 2. T1 > RT2 iff R(T1) > R(T2)
ܾ෨ ൌ ሺܾ෨ଵ ǡ ǥ ǥ ǡ ܾ෨ ሻ represent fuzzy vectors consisting of
௧ 3. T1 = RT2 iff R(T1) = R(T2)
triangular fuzzy numbers with weighted averaged estimators as 4. If T1 t RT2 and T3 t RT4 then R(T1 + T3) t R(T2 + T4)
given below: For ranking triangular fuzzy numbers,this study employs the
ݖǁ ൌ ሺݖଵ ǡ ݖଶ ǡ ݖଷ ሻǡ ranking function as :
ܿǁ ൌ (ci j ) , i = 1, 2, ..., n, j = 1, 2, 3, R(T) = u + w – V
௪ି௨
ݔ ൌ ሺݔ ሻ, i = 1, 2, ..., n, j = 1, 2, 3 where, ߪ ൌ is the variance between u and w .
෨ܾ ൌ ሺܾ ሻ, i = 1, 2,..., m, j = 1, 2, 3,
II. ARITHMETIC ON TRIANGULAR FUZZY NUMBERS AND
RANKING FUNCTION
A triangular fuzzy number is represented by a triplet
T=(u, v, w) and is defined as [1, 2]:
978-1-4673-8545-9/15/$31.00 ©2015 IEEE
III. DUAL SIMPLEX METHOD (1.75,2,2.25) ݔଵ - (0.75,1,1.25) ݔଶ + (2.75,3,3.25)
Consider the following tableau representing a basic ݔଷ (3.75,4,4.25)
solution at some iteration. Suppose that the tableau is dual ݔଵ ; ݔଶ ; ݔଷ 0
feasible (that is, ݖǁ െ ܿǁ ൏ Ͳ for a minimization problem) and is
Table 1: The initial tableau of the primal problem
not primal [Link] a FFLP problem in the feasible
canonical form: Table 2: The optimal tableau.
Min ݖǁ ൌ ܿǁ ௧ ݔ
s.t ܣሚݔ ൌ ܾ෨
ݔ Ͳ
Step 1. Find a basis B of the primal such that ݖǁ െ ܿǁ Ͳfor all
j.
Step 2. Find the rank values R(ܾ෨ ሻ, i = 1, 2,. . . , n . For
minization problems, select the pivot row r with the most
negative ranking value. Let it be R(ܾ෨ ሻ .
Since R(ܾത෨ ሻ> 0 and R(ݖǁ െ ܿǁ ሻ< 0 for all j, the optimal primal
Step 3. Find the rank values R(ݕ ሻ, j=1, 2,.., n. If R(ݕ ሻ Ͳ
and dual solutions are at hand. In particular,
for all j; stop; the dual is unbounded and the primal is infeasible.
ݖǁ ൌ ሺݖଵ ǡ ݖଶ ǡ ݖଷ ሻ= (2.47, 5.6, 8.79)
Otherwise find the rank values R(ݖǁ െ ܿǁ ሻ, j=1, 2,.., n.
ݔଵ = (ݔଵଵ ;ݔଵଶ ;ݔଵଷ )= (1.60, 2.2, 2.76)
Step 4. Find the least ranking ratio from ݔଶ = (ݔଶଵ ;ݔଶଶ ;ݔଶଷ )= (-0.09, 0.4, 1)
ୖሺ௭ೕିǁೕ ሻ
, j = 1, 2, 3, . .. for R(ݕ ሻ< 0 . Let it be related to the ݔଷ = (ݔଷଵ ;ݔଷଶ ;ݔଷଷ )= (0, 0, 0).
ୖሺ௬ೝೕ ሻ
basic column variable ݔ .
Step 5. Next create a vector (1, 1, 1) at the position of pivotal V. CONCLUSION
element R(ݕ ሻand (0, 0, 0) above and below it by using
Gaussian elimination method. Furthermore, the basic variable In this paper, a dual simplex method is used for solving Fully
Fuzzy Linear Programming Problems. The problem is solved
ݔ leaves the basis and ݔ enters the basis.
directly without converting it to its crisp equivalent.
Step 6. Repeat Steps 2-5 until there is no negative ranking value
of the objective function coefficient R(ܾ෨ ሻ.
References
IV. NUMERICAL EXAMPLE [1] Bector, C.R., Chandra, S., Fuzzy Mathematical Programming and Fuzzy
Matrix Games. Studies in Fuzziness and Soft Computing. Springer, Berlin
Consider the following problem. Heidelberg 2005.
Minimize ݖǁ = (1.75, 2, 2.25) ݔଵ + (2.75, 3, 3.25) ݔଶ + (3.75, [2] Kaufmann, A., Gupta, M.M., Fuzzy Mathematical Models in Engineering
4, 4.25) ݔଷ and Management Science. Elsevier Science, Amsterdam, 1988.
s.t. [3] Ullah Khan, I., Ahmad, T., Maan, N., A Simplified Novel Technique for
(0.75,1,1.25) ݔଵ + (1.75,2,2.25)ݔଶ + (0.75,1,1.25) ݔଷ Solving Fully Fuzzy Linear Programming Problems, J Optim Theory
Appl 159:536-546, 2013.
(2.75,3,3.25)
