International Journal of Scientific and Research Publications, Volume 8, Issue 12, December 2018 306

ISSN 2250-3153

Application of Ford-Fulkerson Algorithm to Maximum

Flow in Water Distribution Pipeline Network
Myint Than Kyi*, Lin Lin Naing**

*Department of Engineering Mathematics, Technological University (Mandalay). Mandalay Division, Myanmar.

**Faculty of Computing, University of Computer Studies, Hinthada, Ayeyarwady Division, Myanmar

DOI: 10.29322/IJSRP.8.12.2018.p8441
http://dx.doi.org/10.29322/IJSRP.8.12.2018.p8441

Abstract- In this paper, well known Ford-Fulkerson algorithm in edge in G is a non-negative number. This number represents the
graph theory is used to calculate the maximum flow in water capacity of the edge and it is denoted by 𝑐𝑖𝑗 for the directed edge
distribution pipeline network. The maximum flow problem is one from node 𝑖 to node 𝑗 in G [5]. We consider a network of
of the most fundamental problems in network flow theory and pipelines of water distribution system and the case of network of
has been investigated extensively. The Ford-Fulkerson algorithm pipes having values allowing flows only in one direction. This
is a simple algorithm to solve the maximum flow problem and type of network is represented by weighted connected digraph in
based on the idea of searching augmenting path from a started which District Metered Areas (DMA) are represented by vertices
source node to a target sink node. It is one of the most widely and lines which given water flows through by edges and
used algorithms in optimization of flow networks and various capacities by weights. To provide the maximum flow from
computer applications. The implementations for the detail steps source vertex to sink vertex is one of the most important things in
of algorithm will be illustrated by considering the maximum flow all transmission network. Source vertex is produced flows along
of proposed water distribution pipeline network in Pyigyitagon the edges of the digraph G and sink vertex is received [1]. The
Township, Mandalay, Myanmar as a case study. The goal of this residual graph can be defined by providing a systematic way in
paper is to find the maximum possible flow from the source node order to find the maximum flow. The residual capacity of an
s to the target node t through a given proposed pipeline network. edge can be obtained by using the formula 𝑐𝑓𝑖𝑗 = 𝑐𝑖𝑗 − 𝑓𝑖𝑗 and a
residual network can be defined by giving the amount of
Index Terms- Flow network, Ford-Fulkerson algorithm, Graph available capacity.
Theory, Maximum flow, Water distribution network.
A. Problem Definition
The maximum flow problems involve finding a possible
I. INTRODUCTION maximum flow through a single-source to single-sink flow

I n many fields of applications, graphs theory [1] plays vital role

for various modelling problems in real world such as
travelling, transportation, traffic control, communications, and
network. The maximum flow problem can be seen as a special
case of more complex network flow problems, such as the
circulation problem [6]. The problem is given by a directed graph
various computer applications and so on [2]. Graph is a G with a start node s and an end node t. Each edge has associated
mathematical representation of a network and it describes the with it a positive number called its capacity. A flow in a network
relationship among a finite number of points connected by lines. is a collection of chain flows which has the property that the sum
The maximum flow problem is a type of network optimization of the number of all chain flows that contain any edge is no
problem in the flow graph theory [3]. A flow network is a greater than the capacity of that edge [7]. The goal is to provide
directed graph where each edge has a capacity and receives a as much flow as possible from s to t in the graph. The flow in any
flow as weighted values. A flow must satisfy the restriction that edge is never exceeded its capacity, and for any vertex except for
the amount of flow into a node equals the amount of flow out of s and t, the flow in to the vertex must equal the flow out from the
it except source node, which has only outgoing flow, and sink vertex. On any edge we have 0 ≤ 𝑓𝑖𝑗 ≤ 𝑐𝑖𝑗 . This is called
node, which has only incoming flow [4]. In practice, the flow capacity constraint or edge condition. As a flow constraint, flow
problems can be seen in finding the maximum flows of currents into vertex is equal to flow out of vertex for any vertex. In a
in an electrical circuit, water in pipes, cars on roads, people in a formula,
public transportation system, goods from a producer to
0 if vertex 𝑖 ≠ 𝑠, 𝑖 ≠ 𝑡
consumers, and traffic in a computer network, and so on.
� 𝑓𝑘𝑖 − � 𝑓𝑖𝑗 = � −𝑓 at the source node
𝑘 𝑗 𝑓 at the target (sink)𝑡,
II. RESOURCES AND METHOD inflow − outflow
A simple, connected, weighted, digraph G is called a transport where 𝑓 is the total flow. Subject to these constraints, the total
network if the weighted value associated with every directed flow into target node t can be maximized.

http://dx.doi.org/10.29322/IJSRP.8.12.2018.p8441 www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 8, Issue 12, December 2018 307
ISSN 2250-3153

B. Ford-Fulkerson Algorithm for maximum flow

In 1955, Ford, L. R. Jr. and Fulkerson, D. R. created the Ford-
Fulkerson Algorithm [7]. This algorithm starts from the initial
flow and recursively constructs a sequence of flow of increasing
value and terminates with a maximum flow [8]. The idea behind
the algorithm is simple. We send flow along one of the paths
from source to sink in a graph with available capacity on all
edges in the path that is called an augmenting path and then we
find another path, and so on [6].
1). Notation and Steps of the Algorithm
For every edge connected from node 𝑖 to node 𝑗, we use the
notation (𝑖, 𝑗):
∆𝑖𝑗 = 𝑐𝑖𝑗 − 𝑓𝑖𝑗 for forward edges
∆𝑖𝑗 = 𝑓𝑖𝑗 for backward edges
∆= min ∆𝑖𝑗 taken over all edges of a path.
The Ford-Fulkerson algorithm has two main steps. The first
step is a labelling process that searches for a flow augmenting
path and the second step is to change the flow accordingly.
Otherwise, no augmenting path exists, and then we get the
maximum flow [4]. The detail steps of the algorithm are stated as
follows [1]:
Step 1. Assign an initial flow 𝑓𝑖𝑗 (or may be given), compute f.
Step 2. Label s by ∅. Mark the other vertices “unlabelled.”
Step 3. Find a labelled vertex i that has not yet been scanned.
Scan i as follows. For every unlabelled adjacent vertex
j, if 𝑐𝑖𝑗 > 𝑓𝑖𝑗 compute
∆𝑖𝑗 = 𝑐𝑖𝑗 − 𝑓𝑖𝑗 Fig. 1 Flowchart for maximum flow of water distribution network

∆𝑖𝑗 𝑖𝑓 𝑖 = 1
and ∆𝑗 = � III. RESULTS AND DISCUSSION
min(∆𝑖 , ∆𝑖𝑗 ) 𝑖𝑓 𝑖 > 1
In this paper, the application of graph theory to find the
and label j with a “forward label” (𝑖 + , ∆𝑗 )or if 𝑓𝑖𝑗 > 0, maximum flow of water distribution network has been illustrated
compute ∆𝑗 = min�∆𝑖 , 𝑓𝑗𝑖 � and label j by a “backward by using the Ford-Fulkerson algorithm [1] to the proposed water
label” (𝑖 − , ∆𝑗 ). distribution pipeline network of Pyigyitagon Township in
Mandalay, Myanmar. One who wants to know the possible
If no such j exists then OUTPUT f . Stop. maximum flow required the detail implementation of this
Else continue (i.e., go to step 4). algorithm. The weights on the links are referred as capacities and
current flows for corresponding edges.
Step 4. Repeat Step 3 until t is reached.
[This gives a flow augmenting path 𝑃: 𝑠 → 𝑡.]
If it is impossible to reach t, then OUTPUT ƒ. Stop. Else
continue (i.e., go to Step 5).
Step 5. Backtrack the path P, using the labels.
Step 6. Using P, augment the existing flow by∆𝑡 . Set 𝑓 = 𝑓 +
∆𝑡
Step 7. Remove all labels from vertices 2, … , 𝑛. Go to Step 3.
This is the end of the algorithm. The flowchart illustrating for
the maximum flow of water distribution network is as follow:

Fig. 2 Proposed network of water pipeline (yellow) in Pyigyitagon Township,

Mandalay, Myanmar [9]

http://dx.doi.org/10.29322/IJSRP.8.12.2018.p8441 www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 8, Issue 12, December 2018 308
ISSN 2250-3153

The flow network of pipeline is a directed graph with two Augmented flow 𝑓 = 70 + 15 = 85.
distinguished nodes; source and sink. Since the sizes of pipes
Step 7 Remove labels on vertices 2 … , 6. Go to Step 3.
used in this network may not be the same, the capacity for each
pipe may also be different. So it can only maintain a flow of a
certain amount of water. Anywhere those pipes meet, the total
amount of water coming into that junction must be equal to the
amount going out. Each edge between two nodes has a non-
negative capacity 𝑐 and receives a flow 𝑓 where amount of flow
on an edge cannot exceed its capacity [3]. The first number on
each edge represents capacity and the second number represents Fig. 4 Flow augmenting path 1-2-3-6
current flow (liter/sec). In this pipeline network, we denote the
source node 1 as “s” and the sink (target) node 6 as “t” as shown Step 3 Scan 1
in Fig. 3.
Compute ∆12 = 100 − 65 = 35 = ∆2 . Label 2 by (1+ , 35)
Compute ∆14 = 50 − 20 = 30 = ∆4 . Label 4 by (1+ , 30)
Step 4 Scan 2.
Compute ∆24 = 25 − 10 = 15, ∆4 = min(∆2 , ∆24 ).
∆4 = min(35, 15) = 15. Label 4 by (2+ , 15).
Scan 4.
Compute ∆43 = 20 − 15 = 5, ∆3 = min(∆4 , ∆43 )
Fig. 3 Representation of digraph for proposed network of water pipeline in
Pyigyitagon Township, Mandalay, Myanmar ∆3 = min(15, 5) = 5. Label 3 by (4+ , 5).
A. Implementation of the Algorithm Scan 3.
The implementation of the Ford-Fulkerson algorithm is Compute ∆36 = 65 − 60 = 5, ∆6 = min(∆3 , ∆36 ).
illustrated to find the maximum flow for proposed water
distribution pipeline network in Pyigyitagon Twonship, ∆6 = min(5, 5) = 5. Label 6 by (3+ , 5).
Mandalay from source node s to sink node t. We can choose any Step 5 𝑃: 𝑠 = 1 → 2 → 4 → 3 → 6 = 𝑡 is a flow augmenting
path from source to sink for each iteration step as an augmenting path.
path by using the edge with the non-zero residual capacity in
previous step. This means that the edge with maximum flow Step 6 ∆𝑡 = 5. Augmentation gives 𝑓12 = 70, 𝑓24 = 15,
cannot be used as a segment of augmenting path. 𝑓43 = 20, 𝑓36 = 65.
The detail implementation of algorithm can be seen as Augmented flow 𝑓 = 85 + 5 = 90.
follows:
Step 7 Remove labels on vertices 2 … , 6. Go to Step 3.
Step 1 An initial flow 𝑓 = 50 + 20 = 70(Given).
Step 2 Label (𝑠 = 1) by ∅. Mark 2, 3, 4, 5, 6 “unlabelled”
Step 3 Scan 1
Compute ∆12 = 100 − 50 = 50 = ∆2 . Label 2 by
(1+ , 50).
Compute ∆14 = 50 − 20 = 30 = ∆4 . Label 4 by (1+ , 30).
Step 4 Scan 2.
Compute∆23 = 55 − 40 = 15, ∆3 = min(∆2 , ∆23 )
Fig. 5 Flow augmenting path 1-2-4-3-6
∆3 = min(50, 15) = 1. Label 3 by (2+ , 15).
Step 3 Scan 1
Scan 3.
Compute ∆14 = 50 − 20 = 30 = ∆4 . Label 4 by (1+ , 30).
Compute∆36 = 65 − 45 = 20, ∆6 = min(∆3 , ∆36 )
Step 4 Scan 4.
∆6 = min(15, 20) = 15. Label 6 by (3+ , 15).
Compute ∆45 = 35 − 15 = 20, ∆5 = min(∆4 , ∆45 ).
Step 5 𝑃: 𝑠 = 1 → 2 → 3 → 6 = 𝑡 is a flow augmenting path.
∆5 = min(30, 20) = 20. Label 5 by (4+ , 20).
Step 6. ∆𝑡 = 15. Augmentation gives 𝑓12 = 65,
Scan 5.
𝑓23 = 55, 𝑓36 = 60.
Compute ∆56 = 50 − 25 = 25, ∆6 = min(∆5 , ∆56 ).

http://dx.doi.org/10.29322/IJSRP.8.12.2018.p8441 www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 8, Issue 12, December 2018 309
ISSN 2250-3153

∆6 = min(20, 25) = 20. Label 6 by (5+ , 20). structure [6]. The comparison for calculation times can be
checked in the following table [4].
Step 5 𝑃: 𝑠 = 1 → 4 → 5 → 6 = 𝑡 is a flow augmenting path.
Step 6 ∆𝑡 = 20. Augmentation gives 𝑓14 = 40, 𝑓45 = 35, TABLE I
COMPARISON FOR TIME COMPLEXITY OF MAXIMUM FLOW ALGORITHMS
𝑓56 = 45.
Augmented flow 𝑓 = 90 + 20 = 110. Maximum-flow Algorithm Time Complexity*

Ford-Fulkerson 𝑂(𝑛𝑚 𝑐)
Edmonds-Karp 𝑂(𝑛𝑚2 )
𝑛2
Goldberg-Tarjan 𝑂 �𝑚𝑛 log � ��
𝑚

* n, m and c are number of nodes, edges and capacities

respectively.
Fig. 6 Flow augmenting path 1-4-5-6 Although the calculation time for Ford-Fulkerson algorithm is
Finally, there are no augmenting paths possible from s to t, and higher than other algorithms, it is simple to implement in finding
then the flow is maximum flow. The maximum flows for all the maximum flow. The Edmonds-Karp algorithm is a small
pipelines in network have been calculated by iteration process of variation on the Ford-Fulkerson algorithm. Goldberg and Tarjan
Ford-Fulkerson algorithm and can be seen in Fig. 7. algorithm manipulates the preflow rather than an actual flow in a
graph. Other modified algorithms are developed to reduce
computation time [3], [6], but these algorithms are complicated
and required high speed processors such as supercomputer to
simulate the problem with 1000 nodes and over 94000 arcs and
run for optimal flow [6]. The Ford–Fulkerson algorithm is
guaranteed to terminate if the edge capacities are nonnegative
real values. For these reasons and above comparison, we strongly
recommend that Ford-Fulkerson algorithm is an algorithm to get
the solution for finding maximum flow with simple procedure.

IV. CONCLUSION AND FUTURE WORKS

Fig. 7 Maximum flows in pipeline network from source node s to sink node t
In this paper, Ford-Fulkerson algorithm in flow graph theory has
The maximum flow will be the total flow out of source node been implemented to find the maximum possible flow in water
which is also equal to total flow in to the sink node. In this case, distribution pipeline network of Pyigyitagon Township,
total flow out from source node s is 70 + 40 = 110 and total Mandalay, Myanmar. The main idea of the algorithm is to find a
flow in to the sink node t is 65 + 45 = 110 and hence maximum path through the graph from the source to the sink, in order to
possible flow for this network is 110 liters/sec. send a flow through this path without exceeding its capacity.
B. Comparison and Recommendation Then we find another path, and so on. A path with available
capacity is called an augmenting path. Maximum flow for
The maximum flow problem has been studied by many
researchers because of its importance for many areas of proposed water distribution pipeline network has been developed
applications [3]. There are many algorithms to find the maximum to implement the algorithm. Actually, the flow rate of water in
pipeline is depending upon the size of used pipes and pressure at
flow problems, among them the most popular algorithms are
Ford-Fulkerson algorithm, Edmonds-Karp algorithm, and each node. But we have considered only capacity and flow rate
Goldberg-Tarjan algorithm. Ford-Fulkerson uses depth-first- of water in the network and other factors such as size of pipes
and pressure at each node are not considered in this paper.
searches to find augmenting paths through a residual graph,
while Edmond-Karp uses breath-first-searches to achieve a As the future works, this algorithm can also be used to model
polynomial time complexity, and Goldberg-Tarjan solves the traffic in a road system, fluids in pipes, currents in an electrical
circuit, or anything similar in which something travels through a
problem by gradually pushing flow through the residual graph.
Although they all are algorithms to find maximum flow, the network of nodes. Moreover, based on the flowchart and detail
calculation procedure, one can create computer codes such as
special features of them are different. Ford-Fulkerson algorithm
C/C++, or JAVA, running these codes by using nonnegative real
can be applied to find the maximum flow between single source
and single sink in a graph, while Edmonds-Karp algorithm and weighted values from actual information and data to solve
general Ford-Fulkerson maximum flow problems.
Goldberg-Tarjan algorithm use breath-first-searches and are
performed from the sink, labelling each vertex with the distance
to the sink [10]. All the works above and some other works gave ACKNOWLEDGMENT
algorithms that have good computational performance. But the The author would like to acknowledge her thank to Dr. Sint
algorithms in these works are very complicated and some of Soe, Rector of Mandalay Technological University and
them are only suitable for the flow networks with specific Technological University (Mandalay) for his encouragement and
permission to submit this paper in this journals. And Dr. Lin Lin

http://dx.doi.org/10.29322/IJSRP.8.12.2018.p8441 www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 8, Issue 12, December 2018 310
ISSN 2250-3153

Naing, Professor and Head, Faculty of Computing, University of [6] Zhipeng Jiang1, Xiaodong Hu1, and Suixiang Gao. “A Parallel Ford-
Fulkerson Algorithm For Maximum Flow Problem”. Chinese Academy of
Computer Studies, Hinthada, for his invaluable comments and Sciences, Beijing, China.
perfect supervision throughout my research. Her special thanks
[7] L. R. Ford, JR. and D.R. Fulkerson, “Maximal Flow Through a Network”.
are due to U Myo Thant, Head of Water and Sanitation Rand Corporation, Santi Monica, California. September 20, 1955.
Department, Mandalay City Development Committee, Mandalay,
[8] M. Shokry, “New Operators on Ford-Fulkerson Algorithm”. IOSR Journal
Myanmar for his supports to get required data and information to of Mathematics (IOSR-JM). Vol. 11, Issue 2. Mar-Apr. 2015. pp. 58-67.
her research. Finally, the author’s thanks are to her beloved
[9] Tun Win U, “Proposed Water Supply Master Plan and Existing
mother, brothers and sister because her research cannot be done Wastewater Management System in Mandalay Industrial Hub”. Mandalay
successfully unless they support it. City Development Committee. 2016.
[10] Markus Josefsson, Martin Mützell, “Max Flow Algorithms, Comparison in
REFERENCES regards to practical running time on different types of randomized flow
networks”. Degree Project in Computer Science, DD143X. Sweden,
[1] Erwin Kreyszig, Advanced Engineering Mathematics, 10th edition.
Stockholm 2015.
Copyright © 2011 John Wiley & Sons, Inc. All rights reserved.
[2] Meysam Effati, Analysis of Uncertainty Considerations in Path Finding
Applications [Online] Available: http://5thsastech.khi. ac.ir/data1/
civil/1%20%2812%29.pdf. AUTHORS
[3] Ola M. Surakhi, Mohammad Qatawneh, Hussein A. al Ofeishat, “A First Author – Myint Than Kyi, Lecturer, Department of
Parallel Genetic Algorithm for Maximum Flow Problem,” (IJACSA)
International Journal of Advanced Computer Science and Applications,
Engineering Mathematics, Technological University (Mandalay).
Vol. 8, No. 6, 2017. [Online] Available: http://thesai.org Mandalay Division, Myanmar.
/Downloads/Volume8No6/Paper_20-A_Parallel_Genetic _Algorithm_ Email address: [email protected]
for_Maximum.pdf. Second Author – Dr. Lin Lin Naing, Professor and Head,
[4] Flow network [Online] Available: https://en.wikipedia.org/ wiki/ Faculty of Computing, University of Computer Studies,
Flow_network#Augmenting_paths. Hinthada, Ayeyarwady Division, Myanmar.
[5] Santanu Saha Ray, Graph Theory with Algorithms and its Applications In Email Address: [email protected]
Applied Science and Technology, Copyright Springer India. 2013. ISBN
978-81-322-0750-4.

http://dx.doi.org/10.29322/IJSRP.8.12.2018.p8441 www.ijsrp.org