1

Ford-Fulkerson Method via Linear Ranking Function

for Solving Fuzzy Maximum Flow Problems

Kalle R. Omer1, Abdulqader Othman Hamad Ameen*

1
Corresponding author e-mail: [email protected]

2*
Department of Mathematics, College of Science and Health, University of Koya

Abstract

In this paper, the fuzzy linear programming formulation of fuzzy maximal flow problems
are proposed and on the basis of the proposed formulation a method is proposed –to find
the fuzzy optimal solution of fuzzy maximal flow problems. In the Ford-Fulkerson
method all the parameters are represented by triangular fuzzy numbers. By using the
Ford-Fulkerson method the fuzzy optimal solution of fuzzy maximal flow problems can
be easily obtained. To illustrate the Ford-Fulkerson Method a numerical example is
solved and the obtained results are discussed.

Keywords: Fuzzy maximal flow problem; Ranking function; Triangular fuzzy number.

1.Introduction

A network consists of a set of nodes linked by arcs. Network flow problems have
many applications in the most diverse fields, such as: telecommunications,
transportations, computation, manufacturing etc. The maximal flow problem is an
important problem in the network flow problems, it consists of sending the biggest
 2

amount of flow between two nodes (source S and sink N), holding the capacity
restrictions of each arc[1].

The best example of flow network is sending water from one reservoir to other using
different paths, the sending reservoir is source S and receiving reservoir is destination D.
Each path is made of pipe lines of different diameters (capacity), here assumption is that
the source generates infinite amount of water. Now the problem is how to send the
maximum amount of water from source S to destination D per unit of time. Here, there is
no intermediate storage, so flow conservation principle will follow; that is, at any
intermediate node flow in is equal to flow out. In addition, flow in a pipe is either less or
equal to its capacity (capacity constraints)[2].

In the real life situations there always exist uncertainty about the parameters (e.g.:
costs, capacities and demands) of maximal flow problems. To deal with such type of
problems, the parameters of maximal flow problems are represented by fuzzy numbers
and maximal flow problems with fuzzy parameters are known as fuzzy maximal flow
problems

2 Preliminaries of Fuzzy Concepts

In this section some basic definitions and arithmetic operations are reviewed.

2.1 Basic Definitions

Definition 2.1.1Fuzzy Numbers

Fuzzy numbers are a those with a continuity of grades of membership[3] characterized by

a function which assigns to each number a grade of membership within a range in the
closed unit interval [4]. There are two common kinds of fuzzy numbers, namely the
trapezoidal fuzzy numbers (TpFNs), and triangular fuzzy numbers(TrFNs) , [5] [6], [7].

The Trapezoidal Fuzzy Number (TpFN)

 3

~ ~
A T P FN is A=¿) where [a L , aU ] is the modal set of A , and [a L−α , aU + β ] its
support part[5],[6], [7],[4].

The Triangular Fuzzy Number (TrFN):

~
Let a=a L =aU then the T p FN is reduced to T r FN , and denoted by A=(a , α , β ).Thus
~
A=(a , α , β )⊂(a , a , α , β).Hence, T r FN is a special case of the general case T p FN .
L U

Since the study is focused on the FFGP problems with T r FNS , the next subsection lists
fuzzy algebra properties specific to such
FNs ADDIN EN .CITE< EndNote >¿ Cite> ¿ Author > Mahdavi− Amiri<¿ Author >¿ Year> 2006<¿ Year>¿ Re

2.2 Fuzzy Algebra Properties of T r FNs

Since we are going to formulate the FMFP problem with T r FNs , we mention some
arithmetic properties on these FNs as follows.
~ ~ ~ ~
Let A 1 , A 2 ∈T r FNs , such that A 1=(a1 , α 1 , β1 ) and A 1=(a2 , α 2 , β2 ).Based on

[4, 5],[7],[8],[9],[10, 11],

Thus the following rules apply:

~ ~
1. Addition : A 1 ⨁ A 2=(a1 , α 1 , β 1)⨁ ( a 2 , α 2 , β 2 )=(a1 +a 2 , α 1+ α 2 , β1 + β 2) .
~ ~
2. Image: A 1=Image ( a1 , α 1 , β 1 )=− A 1=−( a 1 , α 1 , β 1 )=(−a1 , α 1 , β 1 ) .
3. Subtraction:
~ ~
A 1 ⊖ A 2=( a 1 , α 1 , β 1 ) ⊝ ( a2 ,α 2 , β 2 )= ( a1 , α 1 , β1 ) ⨁ (−a 2 , α 2 , β 2 )=(a1−a2 , α 1 + β 2 , β 1+ α 2 )
.
~ ~
4. Multiplication: A 1 ⨂ A 2=( a 1 , α 1 , β 1 ) ⨂ ( a2 , α 2 , β 2 ) ≅
~ ~~ ~
(i).(a 1 a 2 , a1 α 2+ a2 α 1 , a1 β 2 +a2 β 1 ¿ ; A1 > 0 , A2 > 0 .
 4

~ ~ ~ ~
(ii).(a 1 a 2 , a2 α 1−a1 β2 , a2 β 1−α 2 α 1 ; A1 > 0 , A2 > 0 .
~ ~ ~ ~
(iii). (a 1 a 2 ,−a2 α 1−a1 β 2 ,−a 2 α 1−a1 α 2 ¿ ; A1 < 0 , A2 < 0 .
~
5. Scalar Multiplication: δ ⨂ A 1=δ ⨂ ( a1 , α 1 , β 1 )=¿
(i).(δ a1 , δ a2 , δ β 1 ¿ ; δ >0
(ii).(δ a1 ,−δ β 1 ,−δ α 1 ¿ ; δ < 0.
~
6. Inverse: ( A ¿ ¿1)−1 ¿ ¿ ¿(a1 , α 1 , β 1 ) =( a1 , β 1 a1 , α 1 a1 ) .
−1 −1 −2 −2

~ ~ ~
7. Division: A 1 ⊘ A 2 ¿ A1 ⨂ ¿
β2 a1 +α 1 a2 α 2 a1 + β 1 a 2 ~ ~ ~
¿ ¿, , 2
¿ ; ∀ A1 , A 2 > 0 .
a2 a2
~ ~
Note that similar formulas hold when A 1∧¿ A2 are negative.

2.3 Linear Ranking Function and its Applications

One of the most convenient methods to defuzzify a fuzzy linear programming (FLP)
problem into its deterministic form and the comparison of fuzzy numbers is by using the
linear ranking function[10],[12],[13],[5, 7].

We describe the related ranking functions, their strengths and weaknesses.

Linear Ranking function is a map which transforms each fuzzy number into its
~ ~
corresponding real line, where a natural order exists , mathematically, R : A → R ; ∀ A∧R
is the set of all real numbers[14] , [15] , .Furthermore , the R' s fuzzy algebra rules on
~ ~ ~ ~
FNs are as follows [4], ∀ A 1 , A2 , A 3 , A 4 ∈ FN s , and δϵ R :

~ ~ ~ ~
1. A 1 ≥R A2 ⟺ R ( A 1 ) ≥ R ( A 2 ) .
~ ~ ~ ~
2. A 1 > A 2 ⟺ R ( A 1 ) > R ( A2 ) .
R

~ ~ ~ ~
3. A 1=
R
A 2 ⟺ R ( A1 ) =R ( A 2 ) .
~ ~ ~ ~
4. A 1 ≤ A2 ⟺ R ( A 1 ) ≤ R ( A 2 ) .
R

~ ~ ~ ~
5. δ A 1+R A 2=δ R ( A 1) + R ( A 2 ) .
~ ~ 0 ⟺ δ R (~
A 1 )=R ( ~
0 )=0.
6. A 1=
R
 5

~ ~ ~ ~ ~ ~
7. A 1 ≥R A2 ⟺ A 1− A2 ≥R 0 ⟺− A2 ≥R − A 1 ,and
~ ~ ~ ~ ~ ~ ~ ~
8. A 1 ≥R A2 Λ A 3 ≥R A 4 ⟹ A1 + A 3 ≥R A 2 + A 4 .

For an T p FN (~a )=(a L , a U , α , β ) , the special kind of R(~a) formulated as follows:

1
1
a ) = ∫ ¿¿
R (~
20

Based on the definition T r FN , and T p FN , (1) can converted into the following form for

the T r FN

β−α
R (~
a ) =a+
4
 6

4.Fuzzy Maximum Flow Problem

In this section linear programming formulation of maximal flow problems in crisp and
fuzzy environment are presented [16].

A. Linear Programming Formulation of Maximal Flow problems in Crisp

Environment
In this section the linear programming formulation of maximal flow problems in crisp
environment is presented.
Let us consider a directed graph G=(V , E) with capacities upper bound uij , and
associating to each arc (i , j) . Let f represent the amount of maximal flow in the network
from source node (say s ¿ to destination node (say t) . The maximal flow problems in
crisp environment may be formulated as follow:
Maximize f
Subject to

∑ x ij =∑ x ki + f ; i=s
j k j

∑ x ij=∑ x ki ; ∀ i≠ s ,t ∑ x ij +f =∑ x ki ; ∀ i=t
j k j k

0 ≤ xij ≤uij ∀ (i , j) ∈ E

B. Linear Programming Formulation of Maximal Flow problems in Fuzzy

Environment
In this section the linear programming formulation of maximal flow problems in
fuzzy environment is proposed.

Let us consider a directed graph G=(V , E) with fuzzy capacities upper bound ~ uij
, and associating to each arc (i , j) a value ~
x corresponding to the fuzzy flow in
ij
~
this arc (i , j). Let f represent the amount of fuzzy maximal flow in the network
from source node s to destination node t . The fuzzy maximal flow problems in
fuzzy environment may be formulated as follow:
~
Maximize f
Subject to
 7

∑ ~x ij =∑ ~x ki ⨁ ~f ; i=s
j k j

∑ ~x ij=∑ ~x ki ; ∀ i≠ s ,t ∑ ~x ij ⨁ ~f =∑ ~x ki ; ∀ i=t
j k j k

~
x ij ≤ ~
uij ∀(i , j )∈ E

5 LP Solution of Fuzzy Maximum –Flow Problems

Sunco Oil wants to ship the maximum possible amount of oil (per hour) via pipeline from
node so to node si in Figure below. On its way from node s0 to node si , oil must pass
through some or all of stations 1, 2, and 3. The various arcs represent pipelines of
 8

different diameters. The maximum number of barrels of oil (millions of barrels per hour)
that can be pumped through each arc is shown in Table 8. Each number is called an arc
capacity. Formulate an LP that can be used to determine the maximum number of barrels
of oil per hour that can be sent from s0 to si .

Arc Capacity (million barrels)

( s0,1) 1 1
(2, , )
8 7

( s0,2) (3,0,0)
 9

(1,2) (3,0,0)

(1,3) 1 1
(4, , )
5 4

(3 , s i) 1 1
(4, , )
8 9

(3 , s i) 1 1
(2, , )
8 7

~
x ij =milion of barrels of oil per hour pass through are (i , j¿ of pip line.

~ 1 1
x s , 1=¿(2, , )
0
3 4
~ ~
x 1,3= 0

~ 1 1
x 1 ,2=¿(2, , )
4 5
x 3 , s =¿ ~
~
i0
~
x =( 2,0,0 )
2 , si

x s , 2=¿ ~
~
0 0
~
x s s =(2,0,0)
i, 0

Barrels per hour, with about (1,0,0) million barrels sent by paths
s0 −1−2−si …1
s0 −1−3−s i                                                                            … 2
s0 −2−si … 3

~
x ij =milion of barrels of oil per hour pass through are (i , j¿ of pip line.
0 ≤ fuzzy flow throught each arc ≤ arc capacity … 1
fuzzy flow into i=fuzzy flow out of node i …2
 10

max . ~z=~
x0 [~
x 0=fuzzy total amount of oil entering the sink ]
s.t
~ 1 1
x s , 1 ≤ (2, , )
0
8 7
~
x s , 2 ≤(3,0,0)
0

~
x 1,2 ≤(3,0,0)
~ 1 1
x 2 , s ≤(2, , )
i
8 7
~ 1 1
x 1 ,3 ≤(4, , )
5 4
~ 1 1
x 3 , s ≤(2, , )
i
8 7

~
x 0=~
x s +~
xs 0,1 0,2

~
x s =~
x 12+~
x 13
0,1

~
x s +~
x 12=~
x 2 ,s
0,2 i

~
x 13=¿ ~x 3 , si ¿

~
x 3 , s +~x 2 , s =~x 0
i i

~ ~
x ij ≥ 0
 11

Ford-Fulkerson Method
A flow in the network is an integer-valued function f defined on the edges of G satisfying

0  f(i, j)  c(i, j), for every edge (i, j) in E.

Conservation Condition For every vertex j in V, where j is not the source S or the
destination D, the sum of the flow into j equals the sum of the flow out of j. A flow that
satisfies the conservation condition is called a feasible flow. Let f be a feasible flow in a
network G. The flow of the network, denoted by f(G) is the sum of flows coming out of
the source S. The Ford-Fulkerson algorithm is a popular method to find the solution of
maximization problem. It simply follows these steps: If an augmenting path exists in the
residual network,

 Assign the flow of amount equal to the capacity of augmenting path to the augmenting
path.

 Again, construct the residual network and repeat the same process until no augmenting
path remains.

The algorithm finds the augmenting path by using graph traversal methods like DFS or
BFS on the flow network.

It is quite difficult to compute the actual time complexity of Ford –Fulkerson algorithm
since it is not possible to determine exactly that how many augmenting paths exist in the
arbitrary flow network. In the worst case of the algorithm, the search for an augmenting
path takes time O(|E|) time, where |E| is the number of edges flow network G(V, E). So
worst-case, the complexity of the algorithm is O(|f|| E|)., where |f| is the amount of
maximal flow in the network [2].
 12

6 Illustrative Example

Sunco Oil wants to ship the maximum possible amount of oil (per hour) via pipeline from
node so to node si in Figure below. On its way from node s0 to node si , oil must pass
through some or all of stations 1, 2, and 3. The various arcs represent pipelines of
different diameters. The maximum number of barrels of oil (millions of barrels per hour)
that can be pumped through each arc is shown in Table 8. Each number is called an arc
capacity. Formulate an LP that can be used to determine the maximum number of barrels
of oil per hour that can be sent from s0 to si .

I =represent increasing of flow .

R=represent reduced of flow .

Flow from source to sink about 2.

Chain (s0 , 2)−( 1,2 )−( 1,3 )−¿ )

I Forward

R backward

β−α
Ranking Function R ¿= R ¿a,α , β ¿=a+ *
4
 13

1 1
R (2, , )¿ 2+¿ ¿R
8 7

R (3,0 , 0)¿ 3 I,R

1 1 1
R (2, , )¿ 1 R
8 7 224

1 1 1
R (4, , )¿ 4 I
5 4 80

1 1 1
R (1, , )¿ 1 I
10 9 360

1 1
i( s0 , 2)=3 , i(1,3)=4 , i ¿)=1 , ( r , 2 )=3
80 24

1
minr (x , y)(x , y) ϵC∩ R=2
224

1 1 1
min ⁡{1 ,2 }= 2
360 224 360

~ 1
x=¿ , )
9

Improved Fuzzy Flow from s0 to si

1 1 1 1 7 14
(2, , )⨁ (1, , )¿ (3, , )
4 5 10 9 20 45
 14

1 1 1 1 9 16
(2, , )⨁ (1, , )¿ (3, , ) Artifical
8 7 10 9 40 63

Conclusion

In this paper ,a new method based on fuzzy linear programming formulation of maximal
flow problems, is proposed.To illustrate the proposed method a numerical example is
solved and the obtianed results are discussed.Using the Ford-Fulkerson method the fuzzy
optimal solution of maximal flow problems occurring in real life situations can be easily
obtained.
 15

Reference

1. Kumar, A., N. Bhatia, and M. Kaur. A new approach for solving fuzzy maximal flow
problems. in International Workshop on Rough Sets, Fuzzy Sets, Data Mining, and
Granular-Soft Computing. 2009. Springer.
2. Dimri, S.C. and M. Ram, Flow maximization problem as linear programming problem
with capacity constraints. Applications and Applied Mathematics: An International
Journal (AAM), 2018. 13(1): p. 33.
3. Zadeh, L.A., Fuzzy sets, in Fuzzy sets, fuzzy logic, and fuzzy systems: selected papers by
Lotfi A Zadeh. 1996, World Scientific. p. 394-432.
4. Hamadameen, A.O. and N. Hassan, A compromise solution for the fully fuzzy
multiobjective linear programming problems. Ieee Access, 2018. 6: p. 43696-43711.
5. Mahdavi-Amiri, N. and S. Nasseri, Duality in fuzzy number linear programming by use of
a certain linear ranking function. Applied mathematics and computation, 2006. 180(1):
p. 206-216.
6. Ganesan, K. and P. Veeramani, Fuzzy linear programs with trapezoidal fuzzy numbers.
Annals of operations research, 2006. 143(1): p. 305-315.
7. Mahdavi-Amiri, N. and S.H. Nasseri, Duality results and a dual simplex method for linear
programming problems with trapezoidal fuzzy variables. Fuzzy sets and systems, 2007.
158(17): p. 1961-1978.
8. Dubois, D. and H. Prade, Operations on fuzzy numbers. International Journal of systems
science, 1978. 9(6): p. 613-626.
9. Dubois, D.J., Fuzzy sets and systems: theory and applications. Vol. 144. 1980: Academic
press.
10. Nasseri, S., et al., Simplex method for solving linear programming problems with fuzzy
numbers. World Academy of Science, Engineering and Technology, 2005. 10: p. 284-288.
11. Sakawa, M., Fuzzy sets and interactive multiobjective optimization. 2013: Springer
science & business media.
12. Maleki, H.R., M. Tata, and M. Mashinchi, Linear programming with fuzzy variables. Fuzzy
sets and systems, 2000. 109(1): p. 21-33.
13. Maleki, H., Ranking functions and their applications to fuzzy linear programming. 2003.
14. Hamadameen, A.O. and Z.M. Zainuddin, A reciprocated result using an approach of
multiobjective stochastic linear programming models with partial uncertainty.
International Journal of Mathematics in Operational Research, 2015. 7(4): p. 395-414.
15. Hamadameen, A.O. and N. Hassan, Pareto optimal solution for multiobjective stochastic
linear programming problems with partial uncertainty. International Journal of
Mathematics in Operational Research, 2018. 12(2): p. 139-166.
16. Kumar, A. and M. Kaur, Solution of fuzzy maximal flow problems using fuzzy linear
programming. A)= a, 2011. 2: p. 4.