mathematics

Article
Robust Design Problem for Multi-Source Multi-Sink Flow
Networks Based on Genetic Algorithm Approach
Sahbi Boubaker 1, * , Noha Hamdy Radwan 2 , Moatamad Refaat Hassan 2 , Faisal S. Alsubaei 3 ,
Ahmed Younes 4,5 and Hameda A. Sennary 2

1 Department of Computer & Network Engineering, College of Computer Science and Engineering,
University of Jeddah, Jeddah 21959, Saudi Arabia
2 Department of Mathematics and Computer Science, Faculty of Science, Aswan University,
Aswan 81528, Egypt; [email protected] (M.R.H.); [email protected] (H.A.S.)
3 Department of Cybersecurity, College of Computer Science and Engineering, University of Jeddah,
Jeddah 23218, Saudi Arabia; [email protected]
4 Department of Computer Science, College of Applied Studies and Community Service, Imam Abdulrahman
bin Faisal University, Dammam 34212, Saudi Arabia; [email protected]
5 Department of Computer Science, Faculty of Computer and Artificial Intelligence, Sohag University,
Sohag 82524, Egypt
* Correspondence: [email protected]

Abstract: Robust design problems in flow networks involve determining the optimal capacity assign-
ments that enable the network to operate effectively even in the case of events’ occurrence such as arcs
or nodes’ failures. Multi-source multi-sink flow networks (MMSFNs) are frequent in many real-life
systems such as computer and telecommunication, logistics and supply-chain, and urban traffic. Al-
though numerous studies on the design of MMSFNs have been conducted, the robust design problem
for multi-source multi-sink stochastic-flow networks (MMSFNs) remains unexplored. To contribute
to this field, this study addresses the robust design problem for MMSFNs using an approach of two
steps. First, the problem is mathematically formulated as an optimization problem and second, a
sub-optimal solution is proposed based on a genetic algorithm (GA) involving two components. The
Citation: Boubaker, S.; Radwan, first component, an outer genetic algorithm, is employed to search the optimal capacity assigned to
N.H.; Refaat Hassan, M.; Alsubaei, the network components with minimum sum. The second component, an inner genetic algorithm,
F.S.; Younes, A.; Sennary, H.A. Robust is used to find the optimal flow vectors that maximize the system’s reliability. Through extensive
Design Problem for Multi-Source experimentation on three different networks with different topologies, the proposed solution has
Multi-Sink Flow Networks Based on been found to be efficient.
Genetic Algorithm Approach.
Mathematics 2023, 11, 3902. https://
Keywords: flow network; robust design; MMSFNs; reliability optimization; genetic algorithm
doi.org/10.3390/math11183902

Academic Editor: Majid MSC: 68M10; 68M15; 68W50

Forghani-elahabad

Received: 12 August 2023

Revised: 7 September 2023
1. Introduction
Accepted: 8 September 2023
Published: 13 September 2023 Multi-source multi-sink stochastic-flow networks (MMSFNs) play a vital role in real-
life systems, such as telecommunication, computer, logistics, supply-chain, and transporta-
tion networks. The optimization of such networks has been a popular topic for applied
research for the last few years. The main concept behind these kinds of networks is related
Copyright: © 2023 by the authors. to optimizing both the network capacity and reliability under several constraints such as
Licensee MDPI, Basel, Switzerland. budget. As an example, Hassan [1] presented a strategy to maximize the reliability of the
This article is an open access article capacity vector for MMSFNs, considering an assignment budget. The strategy was divided
distributed under the terms and into two parts: the first part searches for the preferable components that can be specified
conditions of the Creative Commons in the network with minimum cost, while the second part searches for the flow arc with
Attribution (CC BY) license (https://
maximum reliability of the capacity arc for the assigned components. Hassan [2] solved the
creativecommons.org/licenses/by/
problem of system reliability optimization for MMSFNs, subject to transmission budget
4.0/).

Mathematics 2023, 11, 3902. https://doi.org/10.3390/math11183902 https://www.mdpi.com/journal/mathematics

Mathematics 2023, 11, 3902 2 of 16

constraints, by utilizing genetic algorithms to achieve maximum system reliability through

obtaining the best optimal set of lower boundary points. Forghani-elahabad and Kagan [3]
presented an algorithm to calculate the exact value of reliability in terms of minimal number
of paths in MMSFNs. Network reliability is defined as the probability that a gathering of
sinks is fed by a gathering of sources [4]. It refers to the ability of a network to fulfil its
purpose within the specified timeframe, despite environmental factors, as determined by
the connection between a source and a target [5].
Hamid et al. [6] employed a genetic algorithm (GA) to assign optimal capacities with
minimum total capacities and maximum reliability. The authors presented a GA to optimize
the reliability of MMSFN systems while solving flow allocation problems [7]. Liu et al. [8]
used the GA to find the optimal allocation pattern (vector and node) by the implementation
of the equations in the algorithm. The formulas sum up the flow quantity in each minimal
pass when the node or arc belongs to this minimal pass. Chen [9] introduced an algorithm
to achieve the optimal double-resource assignment for the robust design problem in multi-
state computer networks. Atzori and Raccis [10] used a genetic algorithm to assign capacity
for multicast services.
From the previous selected studies, it can be noticed that the concept of robustness
was not considered when solving the MMSFN problem. For this reason, the current
study will concentrate on solving these kinds of problems under constraints and robust-
ness considerations.
Robust Design is defined as a powerful mechanism for improving the reliability
of a system at a low cost in a short time [11,12]. It is a process used to find the best
configuration for a product design in order to fulfil the end-user’s needs while considering
feasibility requirements [13,14]. Robust Design (RD) is defined as an engineering approach
that concentrates on the development/improvement of new/existing products (physical
products or services), processes, and specific equipment. The robustness of a design refers
to the insensitivity to different disturbances and variations coming from the surrounding
environment [15]. Robust design, based on the concept of building quality into products or
processes, is considered one of the most important systems engineering design concepts
for improving quality and the optimization process. In this direction, Cho and Shin [16]
presented new robust design empirical and optimization models for time-oriented data
by developing a set of methods following three separate phases: modeling, estimation,
and optimization.
The robust design problem in a capacitated flow network aims to achieve the minimum
capacity allocated to each arc so that the network continues surviving even under the arc’s
failure and external disturbances that may affect its operation [17]. Hassan and Abdou [16]
conducted a study to evaluate the MMSFNs’ reliability under time constraints. The studies
by [18,19] solved the robust design problem for single and two-commodity flow networks
with node failure by using a genetic algorithm. Chen and Lin [20] proposed a solution to
find feasible flows that meet flow requirements while minimizing the maximum occurring
cost among all demand realizations. López-Prado [21] defined the robustness not only in
terms of the algorithm’s ability to supply “good” solutions to problem instances of different
sizes and numerical characteristics but also in terms of its ability to keep working well
even when constraints are added or removed. A successful robust design should ensure a
compromise between feasibility and optimality.
Reliability evaluation is the process of determining whether an existing system has
achieved a particular level of operational reliability. For example, the reliability evaluation
of electric power systems is a fundamental and vital problem in the planning, design, and
operation of power systems. The reliability assessment in a conventional power system
is the probability that it implements its functions properly, without any failure within a
stipulated period of time, when it is subject to normal operating conditions [22]. In transport
networks, reliability is a useful tool to represent its operational goodness [23]. Network
reliability is important to measure the degree of stability and quality of infrastructure
assembly of a network [24].
Mathematics 2023, 11, 3902 3 of 16

In order to carefully cope with the problems usually faced during the design phase of
a network, this study’s main contributions are the formulation of the robust design problem
for MMSFNs and its sub-optimal solution using efficient mathematical tools and proposing
suitable algorithms. The solution of the problem is built around the concept of robustness
defined as the ability of a network to operate even under uncertainties and failures. In
addition, an approach based on genetic algorithms (GAs) is developed to solve the problem
in MMSFNs by minimizing the sum of arcs capacities and searching for the best set of flow
vectors that check the conditions and achieve maximum reliability. The proposed approach
is illustrated through three networks having different topologies to show its efficiency. To
the best of the authors’ knowledge, this study is the first that considers the concept of
robustness in the design problem of MMSFNs.
The remaining parts of this paper are organized as follows: in Section 2, materials and
methods including the structural analysis, problem formulation, the proposed approach
for solving the problem, and the whole algorithm are presented in Section 2. The results
and discussion are provided in Section 3. Finally, Section 4 concludes the paper.

2. Materials and Methods

2.1. Notations
• Network notations:
np No. of paths
nc No. of arcs
v No. of nodes
σ No. of sources
θ No. of sinks
A The set of arcs {a_b|1 ≤ b ≤ nc}
S The set of source node {s_1,s_2,. . .,s_σ}
T The set of sink node {t_1,t_2,. . .,t_θ}
MP Minimal path
MPi, j, k The kth minimal path from source i to sink j
m No. of resources.
sdw,j The demand for resource w at sink node tj
DD The demand DD = {sdw,j|1 ≤ w ≤ m, 1≤ j ≤ θ}
srw,i The maximum quantity of resource w that source node si can supply
RR The resource RR = {srw,i|1≤ w ≤ m, 1 ≤ i ≤ σ}
Mi Maximum capacity of an arc i
Flow vector defined as F = (f _1,1,1,1,f _1,1,2,1 . . ., f _(i,j,k_(i,j,) 1), . . ., f_(i,j,k_(i,j),m),
F . . .,f_(σ,θ,k_(σ,θ),m)), where f_(i,j,k,w) represents the flow quantity of resource w on
MP_(i,j,k)
NF Length of the flow vector, NF = np × m
RS The system reliability

• Parameters of the outer GA:

NP1 Number of chromosomes
NG1 Number of genes equals to nc
MG1 The maximum number of generations
Cr1 The crossover rate
Mu1 The mutation rate

• Parameters of the inner GA:

NP2 Number of chromosomes
NG2 Number of genes equals to NF
MG2 The maximum number of generations
Cr2 The crossover rate
Mu2 The mutation rate
X Capacity vector, X = (x_1,x_2,. . .,x_neq)
RX The reliability of the capacity vectors
Mathematics 2023, 11, 3902 4 of 16

2.2. Assumptions
• The flow conservation law must apply to the flow network.
• The arcs have capacities that are statistically independent.
• The capacity of an arc is an integer-valued random variable, which takes values
0 < 1 < 2<. . .< Me according to a given distribution.
• The flow along a path does not exceed its maximum capacity of that path.

2.3. The Robust Design Problem for MMSFNs: Definitions

The analysis of a network structure is important to assign the optimal capacity to
network edges to guarantee that the network will continue surviving even under edges’
failures. In the following subsections, the progressive steps for getting a robust design for
MMSFNs as well as the main concepts used during the design will be presented.
• Coverage set
n o n o
Suppose that Γi = MPi,j,k ai ∈ MPi,j,k and Γ j = MPi,j,l a j ∈ MPi,j,l , the arc a j
belongs to the coverage set Qi of ai if and only if Γ j ⊆ Γi , [15].
• Structural impact
The structural impact (Si ) for ai is given by:
k Qi k
Si = (1)
nc
o
where Qi = a j Γ j ⊆ Γi . The edge ai has the strongest impact if Si = 1.


• Critical edge
The arc ai is said to be critical if Si = 1. The network reliability Rs is zero if and only if
ai has zero capacity [15].
• The capacity assignment
In a single-source single-sink network, the maximum capacity Mi of ai ranges from 0
to the demand value. However, in MMSFNs, there are multiple demands requested by the
sinks. As a result, the following expression is proposed here to set Mi values.
Let Ei = ∑ sdw,j ai ∈ MPi,j , then 0 ≤ Mi ≤ Ei + δ, δ ∈ [0, N ], N is a positive integer.
If ai is a critical edge, then Mi should not be less than Ei .
• Probabilities for each edge
The probability for the current capacity γ (0 ≤ γ ≤ Mi ) for ai , Pr{ γ}:
 
Mi γ
Pr{γ} = r ( 1 − r i ) Mi − γ (2)
γ i

where ri is the probability (availability) of ai [15].

2.4. MMSFN Problem Formulation

The problem is divided into two subproblems, as detailed in the next subsections. The
first subproblem is to determine the optimal capacities to be assigned to the network’s arcs.
The second one is to find the optimal feasible solutions for the flow vector to evaluate the
system’s reliability.

2.4.1. First Subproblem

Let us consider M = ( M1 , M2 , . . . , Mnc ) as assigned capacities to the set of arcs
( 1 2 , . . . , anc ). The mathematical formulation of the subproblem is provided as follows:
a , a

Minimize S (3)
Mathematics 2023, 11, 3902 5 of 16

Maximize RS (4)
where S = ∑inc Mi refers to the sum of assigned capacities and Rs is the corresponding
system reliability.

2.4.2. Second Subproblem

Searching the flow vectors that succeed to achieve Equations (5)–(7) (discussed in, [1,2,7,8]).
σ ki,j
∑ ∑ fi,j,k,w = sdw,j , w = 1, . . . , m; j = 1, . . . , θ (5)
i=1 k=1

θ ki,j
∑ ∑ fi,j,k,w ≤ srw,i , w = 1, . . . , m; i = 1, . . . , σ (6)
j=1 k=1

xl ≤ Ml l = 1, . . . , nc (7)
where,
σ θ ki,j m n o
xl = ∑∑ ∑ ∑ fi,j,k,w al MPi,j,k (8)
i=1 j=1 k=1 w=1

The reliability of the capacity vector is given by:

nc
∏ Pr

RX = x ≥ xl (9)
l=1

2.5. The Proposed Approach

The proposed approach to solve the design problem is based on a two genetic algorithms
(GAs) approach. The first GA (outer) is used to assign the optimal capacities to the arcs and
the second GA (inner) is employed to search for the flow vectors that satisfy the conditions
in (3)–(5) while maximizing the network reliability. The proposed approach (as required
in a GA solution) should include the representation of chromosomes, initial population
generation, fitness function, crossover process, and mutation process. Interested readers can
refer to [18,25] for details about the GA and its structure, components, and hyperparameters.

2.5.1. The Outer GA

Representation
The representation chromosome is used to characterize each chromosome in the GA.
The chromosome M is exemplified by a series of length (nc), where (nc) refers to the number
of arcs ( M1, M2, . . . , Mnc ).

Initial Population
The initial population is the first step in the genetic algorithm (GA) and consists of a
set of random potential solutions. Each solution is symbolized by a chromosome. Typically,
this initial population is generated randomly to produce a range of possible solutions [26].
Although random initialization may allow diversification of solutions over the search space,
other initialization methods such as pre-defined known solutions can be used.

Fitness Function
The fitness function is the heart of a genetic algorithm. This function makes it possible
to evaluate a given individual solution and determine how well it satisfies the optimization
criterion that the algorithm is developed for. The sum of the assigned capacities (S ) for the
ith individual is selected to be the fitness function of this candidate solution, i.e., f it(1) = S .
To normalize the fitness function for each individual i, the calculated value is divided by the
total sum of all fitnesses (∑i f it(1)) in the population. The following Algorithm 1 outlines
how to calculate the fitness for the ith solution in the population:
Mathematics 2023, 11, 3902 6 of 16

Algorithm 1: Fitness evaluation

f it(1) = S
f or 1= 1 : NP1
Normalize_ f it(1) = f it(1)/∑i f it(1)
End f or

Selection
The selection process is one of the important steps in genetic algorithms, used to
choose individuals based on their fitness value. Chromosomes with higher fitness values
are more likely to be selected for reproduction, while those with lower values have a lower
chance of being selected. In other words, the probability of a chromosome being chosen for
reproduction is proportional to its fitness value [27].
This study employs the roulette wheel selection method, a widely adopted technique
in genetic algorithms to identify promising individuals for the crossover and mutation
process (Algorithm 2). In the roulette wheel selection, like all other selection methods,
each prospective solution is assigned a solution through the fitness function. This measure
of physical fitness determines the likelihood of an individual’s selection. Solutions with
superior fitness have a higher probability of being selected, while inferior solutions may
still have the opportunity to endure through the selection process. This characteristic may
incorporate certain elements that could prove advantageous in subsequent recombination
processes. In this study, the roulette wheel mechanism employed for parent selection relies
on the cumulative sum of the respective individual’s fitness values.

Algorithm 2: Roulette wheel algorithm

Begin
f or j = 1 to NP1
f or i = j to NP1
CSum( j) = CSum( j) + Normalize_ f it(i ) //CSum stands for the commutative sum
End f or i
End f or j
Generate number random Î e [0, 1]
f or j = 1 to NP1
i f Î > CSum( j)
parent 1 = j − 1
end i f
end f or j
Generate number random Î e[0, 1]
f or j = 1 to NP1
i f Î > CSum( j)
parent 2 = j − 1
end i f
end f or j
End

Crossover
Crossover is a genetic operator that combines two chromosomes to produce a new
chromosome. The idea behind crossover is that the new chromosome may inherit the best
characteristics from each of the parent chromosomes.
In this study, a one-point crossover operator that randomly chooses one crossover
point is used. The process involves copying everything before this point from the first
parent and everything after the crossover points from the second parent, as follows [28].
h iα h inc h iα h inc
MC = M p1 ( j) + M p2 ( j) MD = M p2 ( j) + M p1 ( j) (10)
j =1 j = α +1 j =1 j = α +1
Mathematics 2023, 11, 3902 7 of 16

where MC and MD are the children and M p1 and M p2 parents selected based on Mc1 value
and α is the random cut-point.

Mutation
Mutation represents the final mechanism of evolution, wherein one or multiple genes un-
dergo genetic variation and are subsequently transmitted to offspring for potential adaptation.
The genetic algorithm (GA) typically employs a low mutation rate to maintain its functionality
since high rates of mutation may lead to a less efficient and more rudimentary random search
process (Algorithm 3). The introduction of the mutation operator in the algorithm adds
an additional level of randomness, thereby preserving the diversity of the population. The
utilization of the genetic algorithm (GA) technique is fruitful in circumventing the emergence
of analogous solutions and escalating the likelihood of evading local solutions [29].

Algorithm 3: Genetic algorithm

Generate a random number rm ∈ [0,1]
if rm≤ Mu1 then
{
for i = 1 to nc, do
MD (i ) = β, β ∈ {0, 1, . . . , Mi } .
End for
}

2.5.2. The Inner GA

The inner GA is used to locate the optimal set of lower boundary points that maximize
the system reliability. The flowing formulation represents the problem of searching for the
optimal lower boundaries to maximize the system reliability.

FindtheoptimalsetofX
s.t.
RS ismaximized

The equations needed to solve flow vectors searching problem are given in Section 2.4.2.

Representation
The chromosome F is represented by a series of length (NF), where (NF) refers to the
number resulting from multiplying a number to the minimal path (np) and the number of
resources (m) ( f 1,1,1,1 , f 1,1,2,1 , . . ., f i,j,ki,j, 1 , . . ., f i,j,ki,j ,m , . . ., f σ,θ,kσ,θ ,m ).

Fitness Function
The following Algorithm 4 shows how to calculate the fitness for each solution (i) in the
population, where X i corresponds to Fi , and each Fi satisfies Equations (5)–(7) presented
in Section 2.4.2.

Algorithm 4: Fitness function

Begin
Calculate the fitness value for X i = R Xi
f or all elements in the current population
The Normalized f itness f or each element i =
the f itness value f or X i /Sum o f all f itness
End f or
End

Selection
We used the roulette wheel mechanism in this work to select two parents (Algorithm 5).
Mathematics 2023, 11, 3902 8 of 16

Algorithm 5: Roulette wheel algorithm

Begin
f or j = 1 to NP2
f or i = j to NP2
CSum( j) = CSum( j) + Normalize_ f it(i )
End f or i
End f or j
Generate number random Î e[0, 1]
f or j = 1 to NP2
i f Î > CSum( j)
parent 1 = j − 1
end i f
end f or j
Generate number random Î e[0, 1]
f or j = 1 to NP2
i f Î > CSum( j)
parent 2 = j − 1
end i f
end f or j
End

Crossover
One-point crossover is used to generate new offspring for F as follows.
h iα h i NF h iα h i NF
FC = Fp1 ( j) + Fp2 ( j) FD = Fp2 ( j) + Fp1 ( j) (11)
j =1 j = α +1 j =1 j = α +1

where FC and FD are the new vectors generated by pairing up Fp1 and Fp2 based on Mc2
value and α is the random cut-point.

Mutation
A proposed a mutation mechanism to improve the flow vector F based on Mu2 is
provided as follows (Algorithm 6):

Algorithm 6: Mutation mechanism

Generate a random number rm ∈ [0,1]
if rm ≤ Mu2 then
{
for i = 1 to NF, do
FC (i ) = β, β ∈ {0, 1, I, Minc}, Minc the maximum value can be assigned to F(i), [2].
End for
}

Evaluating Rs
If X 1 , X 2 , . . . , X NP2 represent the generated set of capacity vectors that correspond to
the flow vectors, then by removing the non-minimal ones in X 1 , X 2 , . . . , X NP2 , we obtain all
lower boundary points if the network is cyclic (as discussed in, [30]). Next, if X 1 , X 2 , . . . , X l
are all lower boundary points, then the system reliability RS is calculated by Equation (12).

l n o
Z Z ≥ Xi
[
RS = pr (12)
i =1
 𝑅the=system {𝑍 𝑍 ≥ 𝑋 }
𝑝𝑟reliability (12)
all lower
all lower
boundary
boundary
points,
points,
then then
the system 𝑅 is𝑅calculated
reliability is calculated
by Equation
by Equation
(12). (12).

𝑝𝑟{𝑍} = 𝑝𝑟{𝑧 }. 𝑝𝑟{𝑧 }. … . 𝑝𝑟{𝑧 }. Then, we use the recursive sum of disjoint prod-
Mathematics 2023, 11, x FOR PEER REVIEW 𝑅 =𝑅𝑝𝑟= 𝑝𝑟{𝑍 𝑍 {𝑍 ≥ 𝑍𝑋 ≥} 𝑋 } 9 of(12)
17 (12)
ucts (RSDP) procedure presented in [31].
when TM = pr{Z ≥ X }
Mathematics 2023, 11, 3902 𝑝𝑟{𝑍}𝑝𝑟{𝑍} = 𝑝𝑟{𝑧 = 𝑝𝑟{𝑧 }. 𝑝𝑟{𝑧 }. 𝑝𝑟{𝑧
}. … . }. 𝑝𝑟{𝑧
… . 𝑝𝑟{𝑧 }. Then, }. Then,
we use we theuserecursive
the recursive sum sumof disjoint
of disjointprod- 9 of 16
prod-
Mathematics
Mathematics
2023, 2023,
11, x FOR
11, x PEER
FOR PEER
REVIEW REVIEW , }, 9 of 17 9 of 17
ucts ucts
(RSDP)
Evaluating (RSDP)
Rsprocedure and TM
procedure presented = pr{Zin≥[31].
presented }
Xin − [31].pr ⋃ {Z ≥X for i ≥ 2
If 𝑋 , 𝑋 , … , 𝑋 represent when the when TM TM
= pr{Z
generated = set
pr{Z
≥ Xof≥ }capacity
X } vectors that correspond to

the flow vectors,pr { Z }then = pr by {zremoving
1 }.pr { z2 }the .pr zneq . Then,
. . . . non-minimal ones wein 𝑋use , 𝑋 the
, …i,recursive
𝑋 , we sum ofall disjoint
Evaluating
Evaluating and
Rs and TM =
Rs TM𝑅𝑠 pr{Z== 𝑝𝑟 ≥ X }≥−
pr{Z {𝑍Xpr } −⋃
𝑍 pr𝑋{Z
≥[31]. ⋃} = ≥{Z ,
X ≥},𝑇𝑀 ,
X }, for for
≥ 2i ≥ 2 obtain(13)
products (RSDP) procedure presented in
lower boundary points if the network is cyclic (as discussed in, [30]). Next, if 𝑋 , 𝑋 , … , 𝑋 are
all lower boundary If 𝑋 If, 𝑋points,
𝑋, … , 𝑋, 𝑋, … then , 𝑋represent
the system represent the generated
the generated
reliability 𝑅 nset ofsetcapacity
is calculated oof capacity
by vectors
Equationvectors
that that
correspond
(12). correspond to to
𝑍 =flow
If the 𝑧 , flow
the 𝑧vectors,
, … ,vectors,
𝑧 ,… then , 𝑧 then
by removingby removing
when the
TM non-minimal
the= non-minimal
pr Z ≥ 1 ones
ones
X in 𝑋 ,
in 𝑋𝑋 , …
, 𝑋 , 𝑋, … , 𝑋
, we ,obtain
we obtain
all all
𝑅𝑠 =𝑅𝑠 𝑝𝑟 n= 𝑝𝑟{𝑍 o 𝑍 {𝑍≥1 𝑍𝑋n≥} =𝑋 } n=𝑇𝑀 𝑇𝑀o (13) (13)
lowerlower boundary boundary points points
if theifnetwork the network is cyclic
is cyclic
(as discussed
i−1 (as discussed in, [30]).
in, [30]).
Next, Next,
if 𝑋 ,
if 𝑋𝑋 , …, 𝑋, 𝑋, … , 𝑋
are are
≥ Xi − ≥ Xj,i , for i ≥ 2
S
and TMi = pr 𝑅 Z = 𝑝𝑟system {𝑍pr 𝑋1 } Z
𝑍 ≥reliability (12)
all lower
all lower boundary boundary points, points,then then
the
𝑝𝑟(𝑍) the
= system j=
reliability
𝑝𝑟(𝑧 ) 𝑅 is𝑅calculated
is calculatedby Equation
by Equation (12). (12).
(14)
If 𝑍 If = 𝑍𝑧=, 𝑧 𝑧, …, 𝑧, 𝑧, … , …, 𝑧, 𝑧, … , 𝑧 l n o l
𝑝𝑟{𝑍} = 𝑝𝑟{𝑧 }. 𝑝𝑟{𝑧 }. … . 𝑝𝑟{𝑧 Rs =}.prThen, ∑
i recursive sum of disjoint prod-
[
weZuse
Z ≥ the
X =
𝑅 =𝑅𝑝𝑟= 𝑝𝑟{𝑍 𝑍 {𝑍 ≥ 𝑍𝑋 ≥}TM 𝑋 i} (13)
(12) (12)
ucts (RSDP) procedure presented in [31]. 𝑝𝑟(𝑍) i𝑝𝑟(𝑍)
=1= =𝑝𝑟(𝑧𝑝𝑟(𝑧 ) ) i =1 (14) (14)
2.6. The Whole Algorithm of the when Proposed Approach

= pr{Z ≥ X }
Z =𝑝𝑟{𝑍}
If𝑝𝑟{𝑍} =z1𝑝𝑟{𝑧
, z=2 , .𝑝𝑟{𝑧
. . 𝑝𝑟{𝑧
}. .}.. ,…zTM
, ze}., .𝑝𝑟{𝑧 . }.
𝑝𝑟{𝑧
neq … . 𝑝𝑟{𝑧 }. Then,}. Then,
we use we theuserecursive
the recursive sum sum of disjoint
of disjointprod- prod-
, },
Please
ucts seeucts
(RSDP) and
Algorithm
(RSDP) TM
procedure 7=
procedureforpr{Z the ≥
presented whole X }−
presented pr ⋃ [31].
inalgorithm
[31].
in {Z
neq
≥
of X
the proposed for i ≥ 2
approach.
2.6. The
2.6. Whole
The Whole Algorithm Algorithm of theofProposed the Proposed ApproachApproach =∏ } X }
when prTM
when ( Z ) TM
= pr{Z=prpr{Z
≥(zXe ) ≥ (14)
Algorithm 7: The whole algorithm of the proposed e=1approach
, }, , },
and 𝑅𝑠 TM
and =for TM
=
𝑝𝑟 pr{Z
= {𝑍pr{Z
≥𝑍X ≥ }algorithm
≥−X𝑋pr }}−=⋃pr {Z ≥{Z
⋃proposed
𝑇𝑀 Xproposed
≥ X approach.
for i for
≥ 2i ≥ 2(13)
StartPlease Please
see Algorithm
see Algorithm 7 for 7the whole
the wholealgorithm of the of the approach.
2.6. The Whole Algorithm of the Proposed Approach
Input the network information, such as minimal path, demand, resources, MG1, NP1,
Algorithm
Algorithm 7: The 7: whole
The whole algorithm algorithm of the ofproposed
the proposed ofapproach
the approach
NG1,If Cr1, 𝑧 , 𝑧 ,MG2,
𝑍 =Algorithm
Mu1, … , 𝑧7 ,shows
… , 𝑧 NG2,
NP2, the whole Cr2, algorithm
Mu2. proposed approach.
StartStart 𝑅𝑠 = 𝑅𝑠𝑝𝑟= 𝑝𝑟{𝑍 𝑍 {𝑍 ≥ 𝑍𝑋 ≥} =𝑋 } =𝑇𝑀 𝑇𝑀 (13) (13)
Start outer GA
InputInput
the network
Algorithmthe network information,
7: The information,
whole algorithmsuchsuch as minimal
of the asproposed
minimal path, path,
demand,
approach demand, resources,
resources, MG1,MG1, NP1,NP1,
Generate the initial population randomly
If 𝑍 =
If 𝑍 𝑧 = , 𝑧 𝑧, … , 𝑧, 𝑧, …, … , 𝑧 , 𝑧 , … 𝑝𝑟(𝑍)
, 𝑧
with size
= 𝑝𝑟(𝑧𝑁𝑃1. ) (14)
NG1,NG1,Cr1,
Start Cr1,
Mu1, Mu1,
MG2, MG2,
NP2, NP2,
NG2, NG2,
Cr2, Cr2,
Mu2. Mu2.
Evaluate the initial population.
StartStart
outerouter GA networkGA
While Input ℊ ≤the 𝑀𝐺1, do information, such as minimal path, demand, resources, MG1, NP1, NG1, Cr1,
GenerateMu1,
Generate theMG2, theNP2,
initial initial NG2,
population Cr2, Mu2.
population randomly randomly with
𝑝𝑟(𝑍) with
size
= size
𝑝𝑟(𝑍) 𝑁𝑃1.
=𝑝𝑟(𝑧 𝑁𝑃1. )
𝑝𝑟(𝑧 ) (14) (14)
While
2.6. The Ᵽ ≤ outer
Whole 𝑁𝑃1, GAGenerate
do
EvaluateStart
Evaluatethe Algorithm
initial
the initial ofpopulation.
population. thethe Proposed Approach randomly with size NP1.
initial population
Select Evaluate
two chromosomes.
While While ℊ ≤ 𝑀𝐺1, ℊ the≤ 𝑀𝐺1, initial population.
do do
Generate
Please
While new
see offspring
ℊ Algorithm
≤ MG1, doafter 7 forapplying
the whole crossover
algorithm and of mutation.
the proposed approach.
While WhileⱣ
2.6.
While
≤
Ᵽ= The𝑁𝑃1,
Ᵽ
2.6. ≤
Ᵽ Whole
+The
≤
𝑁𝑃1,
do
1NP1,Whole do
Algorithm
do Algorithm of theofProposed the Proposed ApproachApproach
Select
end Select
do
Algorithmtwo chromosomes.
Select two
7:twoThe chromosomes.
chromosomes.
whole algorithm of the proposed approach
Generate
Generate new
Generate Please
Evaluate the current new
offspring
Please
new seeoffspring
seeafter
Algorithm
offspring
population. after
applying
Algorithm
after applying
7applying
for7the crossover
for crossover
whole
the whole
crossover and
algorithm
and mutation.
and
algorithm mutation.
mutation.of the ofproposed
the proposed approach.
approach.
Start
Ᵽ=Ᵽ+ == 1Ᵽ + + 11
Save the
Input thebestnetwork solution to M.
information, such as minimal path, demand, resources, MG1, NP1,
end doend
end do doAlgorithm
Algorithm 7: The 7: whole
The whole algorithm algorithm of the ofproposed
the proposed approach approach
ℊ = ℊ
NG1, Cr1, + 1 Mu1, MG2, NP2, NG2, Cr2, Mu2.
EvaluateEvaluate
Evaluatethe
StartStart the
current
the current
current population.
population.
population.
end
Startdoouter
Save the GAbest solution to M.
Save
ReportSave
the best
the the
Input best best
solution
Input
the solution
network
the tofound.
networkM. to M.
information,
information, suchsuch as minimal
as minimal path,path, demand,demand, resources,
resources, MG1,MG1, NP1,NP1,
Generate the +initial
ℊ = ℊ 1solution population randomly with size 𝑁𝑃1.
ℊ = ℊ NG1,+= 1ℊ NG1,
+Cr1, 1 Cr1, Mu1, Mu1,MG2, MG2,NP2, NP2,
NG2, NG2,Cr2, Cr2,
Mu2. Mu2.
Start inner
Evaluateendthe doGA initial population.
end doend do Start
Start outer outer
GA GA found.
While Report
Generate ℊ ≤the the
𝑀𝐺1, best
initial solution
population
dosolution randomly with size 𝑁𝑃2.
ReportReport
the inner
Start best
Generatethe best
Generate solution
GA
the initial
the found.
initial found.
populationpopulation randomly randomly withwith size size 𝑁𝑃1.𝑁𝑃1.
Evaluate
While the
ⱣGenerate
≤inner
𝑁𝑃1, initial population.
doinitial population randomly with size NP2.
StartStart
inner GA
EvaluateEvaluate GA
the
Whiletwo
Select 𝑀𝐺2,the
ℊ ≤chromosomes. doinitial
the initial population.
population.
Generate Evaluate
Generate
While the
While ℊthe initial
initial
the ≤ initial
𝑀𝐺1,
ℊ ≤ population.
population population
𝑀𝐺1,
do do randomly randomly withwith 𝑁𝑃2.𝑁𝑃2.
size size
While ⱣWhile
Generate ≤new 𝑁𝑃2,ℊ ≤ do
offspring
MG2, do after applying crossover and mutation.
Evaluate
Evaluate
Whilethe
While initial
Ᵽthe≤ initial
𝑁𝑃1,
Ᵽ population.
≤ 𝑁𝑃1,
do population.
do
Ᵽ=Ᵽ+1
While ≤ Select
NP2, do two chromosomes.
While ℊ ≤Select
While 𝑀𝐺2,
ℊtwo ≤ 𝑀𝐺2, do chromosomes.
do
end do Select chromosomes.
two
Generate Select new offspring after applying crossover and mutation.
two chromosomes.
While WhileⱣ ≤ 𝑁𝑃2,
Ᵽ
Generate ≤
Generate 𝑁𝑃2,
do
new do
new
offspring
Ᵽ = Ᵽ +population.
1
Generate offspring new after after
applying
offspring applying
aftercrossovercrossover
applying and mutation.
and mutation.
crossover and mutation.
Evaluate the current
Ᵽ= Ᵽ+
Select =
= 1two
Ᵽ+
Select + 11chromosomes.
two chromosomes.
Save the best solution to M.
Generate
end Generate
do new new offspring offspring afterafter
applyingapplying crossover
crossover and mutation.
and mutation.
ℊ = ℊ end + 1 do end do
Evaluate Ᵽ = theⱣ += 1Ᵽ
current + 1 population.
end do Evaluate Evaluate the current
the current
Save the best solution to F.
population. population.
Reportℊthe Save Save
the best
the best
solution solution to M. to M.
=ℊbest+ 1 solution found.
ℊ=ℊ+ = 1ℊ + 1
Start inner
end do GA
end do end do
Generate Reportthethe optimal
initial lowers found.
population randomly with size 𝑁𝑃2.
Report
End Report
inner theGA best
the best
solution
Calculate solution
the found.
system found.
reliability Rs.
Evaluate the initial population.
Startouter
End Start
inner GAinner
GA GA
While ℊ ≤ 𝑀𝐺2, do
End Generate Generate the initial
the initial population population randomly randomly withwith size size 𝑁𝑃2.𝑁𝑃2.
While Ᵽ ≤ 𝑁𝑃2, do
Evaluate Evaluate the initial
the initial population.population.
3. Results andSelect two chromosomes.
Discussion
While While ℊ ≤ 𝑀𝐺2,
ℊ ≤ 𝑀𝐺2, do do
Generate new offspring after applying crossover and mutation.
ⱣⱣ = ≤ Ᵽ𝑁𝑃2,
Ᵽ+ ≤1𝑁𝑃2,
In
While the
While following do subsections,
do the presented approach has been applied to three net-
works. The valid values Select of
Select
two the GA
chromosomes.
two parameters were as follows, MG1 = 100, NP1 = 20,
chromosomes.
Cr1 = 0.95, Mu1 = 0.05, GenerateMG2
Generate =new 100,new NP2offspring
offspring = 10,after15, after
and
applying 20. The
applying value
crossover ofand
crossover Cr2mutation.
= 0.95,
and and that
mutation.
Ᵽ=Ᵽ+
= 1Ᵽ + 1
 Mu2 = 0.05. The proposed approach was implemented in
with hardware specification that comprises a core i7 Intel p

Mathematics 2023, 11, 3902 3.1. Case1: Network with Three Sources and Two Sinks
10 of 16

This network consists of 10 arcs as shown in Figure 1

of Mu2 = 0.05. The proposed approach was implemented in MATLAB 2017a environment
MP = {a , a }, MP = {a , a }, MP = {a , a }, MP = {a4,
with 1,1,2 2 9
hardware specification that1,2,1
comprises1a core
8 i7 Intel processor
2,1,1 3 8 9GB of RAM.
with 2,2,1

=3.1.{aCase1:
6, a10 }. Where
Network with Three ncSources
= NG1 and Two=Sinks 10, NF = NG2 = 21, Resources:
r3,2, This
r3,3)network
= (5,consists2, 3, of5,103, arcs2,as 2,
shown2, in3). Demand:
Figure DMP
1; it has 7 MPs; = 1,1,1
(d=1,1{a,1,da71,2 }, , d2,1
MP1,1,2 = {a2 , a9 }, MP1,2,1 = {a1 , a8 }, MP2,1,1 = {a3 , a9 }, MP2,2,1 = {a4 , a10 }, MP3,1,1 = {a5 , a9 }
Table
and MP3,2,1 1 =shows thencbest
{a6 , a10 }. Where = NG1minimum
= 10, NF = NG2 = 21, 𝒮 Resources:
achieves R = (rmaximum
1,1 , r1,2 , r1,3 , 𝑅
𝑅 . S achieves maximum Rs when NP2
r2,1 , r2,2 , r2,3 , r3,1 , r3,2 , r3,3 ) = (5, 2, 3, 5, 3, 2, 2, 2, 3). Demand: D = (d1,1 , d1,2 , d2,1 , d2,2 , d3,1 ,
and
d3,2 ) =Table
(3, 1, 2, 2, 1, 23).clarifies
Table 1 shows Ftheandbest minimum
= 10, 15, and 20, and Table 2 clarifies F and R X .

Figure 1. Network
Figure 1. Network withand
with three sources three sources and two sinks.
two sinks.

Table 1. Case1: The best solution found while varying NP2 and its corresponding M, S, and Rs.

Table
NP2 1. Case1: The bestMsolution found whileSvarying
RS NP2 and i
10 [4 2 4 2 5 4 3 3 5 4] 36 0.973585
NP2
15
20
[4
[5
4
3
3
3
4
3
1
3
1
1
5
4
5
5
M 5
4
4]
5]
36
36
0.968533
0.968419
10 [4 2 4 2 5 4 3 3
Table 2. Case1: F and its corresponding RX for different NP2 (10, 15, and 20).
15 [4 4 3 4 1 1 5 5
F RX
1 020 0 0 0 1 [5 0 1 1 31 1 0 31 0 1 31 1 0 30 1 1 1 0.87494 5
0 1 0 1 0 0 0 1 1 1 1 0 1 0 1 1 0 1 0 1 1 0.8749
1 0 0 1 1 0 0 1 1 1 1 0 0 0 1 0 0 1 1 1 1 0.8745
1 0 0 0 0 0 1 1 1 1 1 1 0 1 1 1 1 0 0 0 1 0.8407
Table 2. Case1: F and its corresponding R for different NP2 (10,
0 0 1 1 1 0 0 1 1 1 1 0 1 0 1 1 0 0
X
0 1 1 0.8222
1 1 0 1 0 0 0 1 1 0 1 1 0 1 1 1 0 0 1 0 1 0.7904
1 0 0 1 0 0 0 1 1 0 1 0 0 0 1 1 1 1 1 1 1 0.7904
1 1 0 1 1 0 1 0 1 1 0 0 0 1 1 0 0 F
1 0 1 1 0.7899
1 1 0 1 0 1 0 1 1 0 1 0 0 0 1 1 0 0 1 1 1 0.7594
11 1 0
0 1 0 0 0
0 0 0
1 1 1
0 0
0 0 1
0 1 1
1 1 1 1 1
1 1 0
0 1 1 0
0.7413 1 1
0 1 0 1 0 0 0 1 1 1 1 0 1 0 1 1
1 0 0 1 1 0 0 1 1 1 1 0 0 0 1 0
1 0 0 0 0 0 1 1 1 1 1 1 0 1 1 1
0 0 1 1 1 0 0 1 1 1 1 0 1 0 1 1
1 1 0 1 0 0 0 1 1 0 1 1 0 1 1 1
Mathematics 2023, 11, 3902 11 of 16

Table 2. Cont.

F RX
1 1 0 1 0 1 0 1 1 1 0 0 1 1 1 0 1 0 0 0 1 0.9012
1 1 0 0 0 0 0 1 1 1 1 1 1 1 1 1 0 0 0 0 1 0.9012
1 0 0 0 1 0 1 1 1 1 1 1 0 1 1 1 0 0 0 0 1 0.8855
1 0 1 1 0 0 0 1 1 1 1 0 1 1 1 0 1 0 0 0 1 0.8855
1 0 0 1 1 1 1 1 1 1 1 0 0 1 1 0 0 0 0 0 1 0.8832
1 0 1 1 1 0 0 1 1 1 1 0 1 1 1 0 0 0 0 0 1 0.8832
0 0 1 1 1 0 1 1 1 1 1 0 0 1 1 1 0 0 0 0 1 0.8832
0 1 0 1 0 1 1 1 1 1 1 0 0 1 1 1 0 0 0 0 1 0.8120
1 0 1 1 1 0 0 1 1 1 0 0 1 1 1 0 1 0 0 0 1 0.8102
1 0 0 1 1 0 0 1 1 0 1 1 1 1 1 1 0 0 0 0 1 0.8082
1 0 0 1 1 1 0 1 1 1 1 0 1 1 1 0 0 0 0 0 1 0.7976
1 0 0 1 0 1 1 1 1 1 1 0 0 1 1 0 1 0 0 0 1 0.7976
1 1 0 1 1 0 0 1 1 1 0 1 1 1 1 0 0 0 0 0 1 0.7976
1 1 0 0 1 0 0 1 1 1 0 1 1 1 1 1 0 0 0 0 1 0.7958
1 1 0 1 0 0 0 1 1 1 0 1 1 1 1 0 1 0 0 0 1 0.7299
1 0 1 1 1 0 1 1 1 1 1 0 0 1 1 0 0 0 0 0 1 0.8760
1 1 0 0 1 0 1 1 1 1 0 0 0 1 1 1 0 1 0 0 1 0.8760
1 1 0 0 1 0 1 1 1 1 0 1 0 1 1 1 0 0 0 0 1 0.8760
1 1 0 0 0 0 1 1 1 1 0 0 0 1 1 1 1 1 0 0 1 0.8313
1 1 0 1 1 0 1 1 1 1 0 0 0 1 1 0 0 1 0 0 1 0.8313
1 1 1 1 0 0 0 1 1 0 1 0 1 1 1 1 0 0 0 0 1 0.8313
1 1 0 1 1 0 0 1 1 1 0 0 1 1 1 0 0 1 0 0 1 0.8301
0 1 1 1 0 0 0 1 1 1 1 0 1 1 1 1 0 0 0 0 1 0.8301
1 0 1 0 1 0 1 1 1 1 1 0 0 1 1 1 0 0 0 0 1 0.8301
1 0 1 1 1 0 0 1 1 1 0 0 1 1 1 0 1 0 0 0 1 0.8292
1 1 0 1 0 1 1 1 1 1 1 0 0 1 1 0 0 0 0 0 1 0.8292
1 1 0 1 1 0 0 1 1 1 0 1 1 1 1 0 0 0 0 0 1 0.8222
1 1 0 1 0 0 0 1 1 1 0 0 1 1 1 0 1 1 0 0 1 0.8222
0 1 1 1 1 0 0 1 1 1 0 0 1 1 1 1 0 0 0 0 1 0.8222
1 1 0 1 0 1 1 1 1 0 1 0 0 1 1 1 0 0 0 0 1 0.8222
1 1 1 0 1 0 0 1 1 1 0 0 1 1 1 1 0 0 0 0 1 0.8210
1 1 1 0 1 0 0 1 1 1 0 0 1 1 1 1 0 0 0 0 1 0.8210
1 1 0 1 0 0 1 1 1 0 0 0 0 1 1 1 1 1 0 0 1 0.8200
1 1 0 0 0 1 1 1 1 1 1 0 0 1 1 1 0 0 0 0 1 0.7834
1 0 1 1 0 0 0 1 1 1 1 0 1 1 1 0 1 0 0 0 1 0.7834

3.2. Case2: Network with Two Sources and Two Sinks

This network consists of 14 arcs as shown in Figure 2, it has 11 MPs; MP1,1,1 = {a1 ,
a5 }, MP1,1,2 = {a1 , a6 , a9 }, MP1,1,3 = {a2 , a7 , a9 }, MP1,2,1 = {a1 , a6 , a14 }, MP1,2,2 = {a2 , a7 , a14 },
MP2,1,1 = {a3 , a7 , a9 }, MP2,1,2 = {a4 , a8 , a13 , a9 }, MP2,2,1 = {a3 , a7 , a14 }, MP2,2,2 = {a4 , a8 , a13 ,
a14 }, MP2,2,3 = {a4 , a8 , a10 }, and MP2,2,4 = {a4 , a11 , a12 }. Where nc = NG1 = 14, NF = NG2=22,
Resources: R = (r1,1 , r1,2 , r2,1 , r2,2 ) = (15, 17, 10, 13). Demand: D = (d1,1 , d1,2 , d2,1 , d2,2 ) = (7,
8, 5, 8). Table 3 shows the best minimum S achieves maximum Rs when NP2 = 10, 15, and
20, and Table 4 clarifies F and R X .
Mathematics 2023, 11, x FOR PEER REVIEW
Mathematics 2023, 11, 3902 12 of 16

Figure 2. Network with two sources and two sinks.

Figure 2. Network with two sources and two sinks.
Table 3. Case2: The best solution found while varying NP2 and its corresponding M, S, and Rs.
Table 3. Case2: The best solution found while varying NP2 and its corresponding M
NP2 M S RS
10NP2
[15 16 4 17 4 18 9 11 14 9 9M 19 10 14] 169 0.996641 𝓢
15 [8 19 17 11 19 10 15 5 20 13 13 8 8 15] 181 0.989280
20 10
[19 6 [15
18 16
18 154 7 1711 416 18 19
13 9 9 9 11
4 14 17]9 189
17 19 10
0.999984 14] 16
15 [8 19 17 11 19 10 15 5 20 13 13 8 8 15] 18
Table20 [19
4. Case2: F and6its corresponding
18 18 15 7 11NP216(10, 15,
RX for different 13and19
20). 9 4 17 17] 18
F RX
Table
2 0 4.
1 Case2:
0 0 2F and
2 0 its1 corresponding
2 2 1 2 2 R0X for
1 different
2 2 1 1NP2
2 (10,
2 15,0.9916
and 20).
1 2 2 2 1 1 2 2 2 1 1 0 2 0 2 0 0 2 1 2 1 1 0.9645
1 1 2 1 1 1 1 1 1 2 1 0 2 2 2 F1 2 0 1 2 1 2 0.8583
2 0 1 2 2 1 2 1 2 2 2 0 0 2 0 0 2 2 1 1 1 2 0.8530
22 1 01 10 20 2 01 20 12 2 00 11 2 21 20 11 2 22 22 20 2 11 2 2 1
0.7601 1 2
2 1
2
1 2
222 212 1
2
2 0
0
121 112 2 1
2
2 2
210 2
1
2 1 1
1
1 0
100 1
2
0 2
2
221 001 2
0
2 2
2
011 0 0.7536
2 1
0.6973
2 1
01 2 12 22 11 0 12 12 11 2 12 01 2 21 10 20 2 22 21 02 2 10 2 0 1
0.6722 2 1
2 1 2 2 0 2 1 1 2 2 2 0 1 0 0 1 2 2 2 2 1 0 0.6591
02 2 02 11 22 0 22 11 12 1 12 02 1 22 21 00 2 02 20 20 2 02 2 2 1
0.6070 1 1
1 0
1
2 1
121 102 2
2
0 2
1
221 222 2 1
1
1 2
012 2
1
1 1 1
2
1 0
012 1
2
1 1
2
200 100 2
2
0 2
2
120 2 0.9234
2 2
0.9210
2 2
02 1 12 21 21 1 12 01 12 2 12 12 1 11 11 22 2 10 10 10 2 12 2 2 0
0.9206 2 2
1 0 2 0 1 2 2 2 2 0 1 2 2 1 2 2 0 0 0 2 2 2 0.8997
12 1 20 20 22 2 22 00 21 2 22 22 2 20 02 11 0 11 00 20 2 22 2 1 1
0.8997 0 2
2 0 0 0 2 2 2 1 2 2 1 2 2 1 1 2 1 1 0 0 2 2 0.8997
20 0 21 21 02 2 11 02 22 2 22 21 2 20 22 20 2 20 10 00 1 22 2 2 1
0.8996 0 2
1 0
2
2 1
110 200 2
1
2 2
0
022 222 2 2
1
2 2
112 2
2
2 2 1
2
2 2
221 1
2
0 0
1
110 000 0
0
0 2
2
122 2 0.8943
2 2
0.8370
2 1
10 2 22 20 21 2 21 01 22 1 12 11 0 10 22 20 0 12 21 11 2 01 2 2 0
0.8333 2 2
2 2 1 0 1 1 1 2 1 2 2 1 1 1 1 2 2 0 1 0 2 2 0.8331
21 2 01 21 20 0 22 20 22 1 22 02 0 12 12 22 0 12 10 11 2 12 1 0 0
0.8331 2 2
1 1 2 2 2 1 2 0 2 2 0 1 2 0 2 2 1 2 0 0 1 2 0.8330
11 0 12 12 22 1 21 10 21 2 21 21 1 20 22 11 1 12 00 12 2 22 2 0 0
0.8263 2 2
1 1 2 1 2 2 1 2 1 1 2 1 0 0 2 1 2 2 1 0 1 2 0.5717
0 1 2 1 2 1 2 1 1 2 2 1 1 1 1 2 2 0 0 1 2
1 0 2 0 1 2 2 2 2 0 1 2 2 1 2 2 0 0 0 2 2
1 1 0 0 2 2 2 0 2 2 2 2 2 0 2 1 0 1 0 2 2
2 0 0 0 2 2 2 1 2 2 1 2 2 1 1 2 1 1 0 0 2
2 0 1 1 0 2 1 2 2 2 2 2 2 0 2 2 2 0 0 0 1
Mathematics 2023, 11, x FOR PEER REVIEW 13 of 17

Mathematics 2023, 11, 3902 13 of 16

1 0 2 2 2 1 1 2 2 0 1 1 1 1 0 2 2 2 2 1 1 1 0.9992
1 1 2 1 2 0 0 2 2 1 1 1 1 2 2 2 2 0 1 2 1 1 0.9957
2 1 2 1 1 1 0 1 1 2 0 0 2 2 1 2 2 1 2 2 2 0 0.9957
Table 4. Cont.
2 0 2 0 1 1 2 2 1 2 2 2 0 2 1 1 2 1 1 1 1 1 0.9941
1 1 2 0 2 1 0 2 2 0 2F 2 0 1 0 1 2 2 2 2 2 1 RX
0.9906
21 20 02 12 22 1 12 22 20 01 11 10 12 11 02 20 2 12 22 21 11 11 2 0.9584
0.9992
21 21 22 21 12 00 0
0 2
1 2
2 1
2 1
0 1
0 1
2 2
1 2
2 2 2 0
2 0 21 22 11 21 0 0.9957
0.9566
2 1 2 1 1 1 0 1 1 2 0 0 2 2 1 2 2 1 2 2 2 0 0.9957
22 20 02 20 21 01 22 22 11 21 21 21 02 20 12 12 2 01 01 21 21 11 1 0.9526
0.9941
11 11 22 00 12 11 01 22 21 02 21 22 02 11 00 12 2 22 12 22 02 21 1 0.9906
0.9522
2 2 0 1 2 1 2 2 0 1 1 0 2 1 2 0 1 2 2 1 1 2 0.9584
12 12 22 02 21 20 0
2 1
2 2
0 2
2 0
1 0
2 2
1 1
0 2
2 0 0
2 0 2
22 21 02 20 2 0.9522
0.9566
22 12 10 02 02 00 21 22 12 11 12 12 22 02 22 22 0 10 12 12 01 11 2 0.9229
0.9526
21 11 22 10 11 11 10 22 12 20 10 22 22 10 02 22 2 21 12 10 22 11 1 0.9522
0.9176
1 1 2 0 2 2 2 2 0 2 1 2 1 0 2 0 0 2 2 0 2 2 0.9522
22 01 11 10 10 10 10 22 21 11 21 21 22 22 21 22 1 21 21 20 11 22 0 0.8803
0.9229
02 21 22 01 21 21 02 22 21 01 01 20 22 01 22 20 2 21 11 12 21 01 2 0.7910
0.9176
12 00 11 01 21 11 01 22 10 11 11 12 22 22 12 21 2 22 22 21 22 10 0 0.8803
0.7149
0 2 2 0 2 2 2 2 1 1 1 0 2 1 2 0 2 1 1 2 0 2 0.7910
11 10 01 20 22 01 11 22 02 11 12 20 22 22 22 12 2 12 02 02 21 20 1 0.7109
0.7149
21 11 00 02 12 10 12 22 21 10 22 01 22 22 21 22 1 20 10 12 12 11 2 0.7105
0.7109
12 21 20 10 11 11 22 22 12 02 21 11 22 20 10 22 2 21 01 11 01 12 2 0.7105
0.7098
1 2 2 1 1 1 2 2 2 2 1 1 2 0 0 2 2 0 1 0 1 2 0.7098
22 11 11 11 00 11 22 22 22 22 22 00 22 22 22 22 0 02 20 00 02 20 0 0.7095
0.7095
22 00 22 11 22 0 22 22 22 11 11 22 00 22 00 11 0 02 22 22 22 20 0 0.6072
0.6072

3.3.
3.3.Case3:
Case3:Network
Networkwith withTwo TwoSources
Sourcesand andThreeThreeSinks Sinks
This
Thisnetwork
networkconsists consists of of 11
11 arcsarcs as as shown
shown in in Figure 3, it has 13 MPs; MP1,1,1 1,1,1 =={a {a11, ,aa7},
7 },
MP 1,1,2=={a
MP1,1,2 {a2,2 ,aa5,5 ,aa77},},MP 1,2,1=={a
MP1,2,1 {a11, ,aa88},},MP
MP1,2,21,2,2=={a {a22, ,aa99},},MP 1,2,3=={a
MP1,2,3 {a22, ,aa55, , aa88},}, MP1,3,11,3,1=={a {a22, ,aa1010},},
MP
MP2,1,1 = {a
2,1,1 = 3{a, a35,, a57}, , aMP MP=2,1,2
7 }, 2,1,2 {a4, =a6{a , a45,, a76},, aMP5, a 7 },=MP
2,2,1 {a3,2,2,1
a9}, =MP 3 , a=9 },
{a2,2,2 {aMP 4, a62,2,2
, a9},=MP , a=6 , {a
{a42,3,1 a93,},
aMP
11}, 2,3,1 = {a=3 ,{aa411
MP2,3,2 , a},6,MPa10}, and
2,3,2 = {a
MP 4 , a
2,3,3
6 , a
={a 104 ,},a and
11 }. MP
Where nc
2,3,3 ={a
= ,
NG1
4 a 11 }.
= Where
11, NF = nc
NG2 = NG1
= 26, = Re-11,
NF = NG2
sources: R ==(r26, 1,1, rResources:
1,2, r2,1, r2,2) R = (r1,1
= (10, 19,, r14, , r
19).
1,2 2,1 2,2 , r
Demand:) = (10,D19,
= 14,
(d 1,119).
, d 1,2Demand:
, d 1,3 , d 2,1 , D
d =
2,2 , (d
d 1,1 1,2 ,
2,3 ) ,
= d(3,
d
2, 1,3 , d
2, 2, 2,1 , d
3, 3).2,2 ,
Tabled 2,3 5 shows the best minimum 𝒮 achieves maximum 𝑅 when NP2 = 10,
) = (3, 2, 2, 2, 3, 3). Table 5 shows the best minimum S achieves maximum
Rs and
15, when 20,NP2and=Table 10, 15, and 20, and
6 clarifies F and 𝑅 . 6 clarifies F and R X .
Table

Figure
Figure3.3.Network
Networkwith
withtwo
twosources
sourcesand
andthree
threesinks.
sinks.

Table 5. Case3: The best solution found while varying NP2 and its corresponding M, S, and Rs.

NP2 M S RS
10 [8 12 14 4 15 13 7 2 14 12 10] 111 0.995669
15 [12 6 12 14 5 9 12 6 12 14 9] 111 0.999923
20 [15 12 12 15 9 4 10 11 14 4 5] 111 0.999403
Mathematics 2023, 11, 3902 14 of 16

Table 6. Case3: F and its corresponding RX for different NP2 (10, 15, and 20).

F RX
0 1 1 1 0 1 1 0 0 1 1 0 1 0 1 0 1 1 0 0 0 1 0 1 1 1 0.9595
1 1 1 0 0 0 1 1 1 1 1 1 0 1 1 0 0 1 0 0 0 1 0 0 1 1 0.9595
1 0 1 1 1 0 1 1 0 1 1 1 0 1 1 0 0 1 0 0 1 1 0 1 0 0 0.9573
1 0 1 1 0 0 1 1 0 1 1 1 1 0 0 1 1 0 0 1 0 1 1 0 0 1 0.9573
1 1 1 1 0 1 0 1 1 0 1 1 0 0 1 0 0 1 1 0 0 1 1 1 0 0 0.9573
1 0 1 1 0 1 1 0 1 0 1 1 1 1 0 0 0 1 0 1 1 1 0 0 0 1 0.9212
1 0 1 1 0 1 0 1 1 0 0 1 1 0 0 1 0 1 1 0 1 1 1 1 0 0 0.8851
1 0 1 0 0 0 0 0 1 1 1 1 1 1 0 1 1 1 0 1 1 1 0 1 0 0 0.8851
0 0 1 1 0 1 1 0 1 0 1 1 1 0 1 1 0 1 0 1 1 1 0 1 0 0 0.8851
1 1 0 1 0 0 0 1 1 1 0 1 1 0 1 0 1 1 0 0 1 1 0 0 1 1 0.8851
1 1 0 0 1 0 0 0 0 1 1 1 1 0 1 1 1 1 0 1 1 1 0 1 0 0 0.9999
1 1 0 1 0 1 1 1 0 0 0 1 1 0 1 0 1 0 0 1 0 1 1 1 1 0 0.9998
1 1 1 0 1 1 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 1 1 0 0 1 0.9962
0 0 1 1 0 1 1 0 0 0 0 1 1 1 1 0 0 1 1 1 1 1 1 1 0 0 0.9962
1 1 0 0 0 0 0 0 0 1 0 1 1 1 1 0 1 1 1 1 0 1 1 0 1 1 0.9962
0 1 1 1 0 0 1 1 1 0 0 1 1 0 1 0 0 1 0 1 1 0 1 1 0 1 0.9962
1 0 1 1 1 0 0 1 0 1 0 1 1 0 0 1 1 0 0 1 1 1 1 1 0 0 0.9962
1 0 1 1 1 1 0 0 0 0 1 0 0 1 1 0 1 1 0 1 0 1 0 1 1 1 0.9960
1 0 1 1 1 1 1 1 0 0 0 0 0 0 1 1 0 0 0 1 1 1 0 1 1 1 0.9039
1 0 1 1 1 0 0 1 0 1 0 1 1 0 0 1 1 1 0 0 1 1 1 0 0 1 0.9038
0 1 1 0 0 0 0 1 0 1 1 1 1 0 1 1 1 0 1 1 0 1 1 1 0 0 0.9038
1 0 1 1 1 1 0 0 0 1 0 1 0 0 1 1 0 1 1 0 0 0 1 1 1 1 0.8988
1 0 1 1 0 1 0 0 0 0 1 0 0 1 1 0 1 1 1 1 0 1 1 1 0 1 0.8988
1 1 0 0 0 1 0 1 0 1 1 0 1 0 1 1 1 0 1 0 1 1 0 1 0 1 0.8988
0 1 1 0 1 1 0 1 1 0 0 0 1 0 1 1 0 1 0 0 0 1 1 1 1 1 0.8007
0 1 1 1 1 1 0 1 0 0 0 1 1 0 1 0 1 0 0 1 1 1 0 0 1 1 0.9994
1 1 1 0 1 1 1 0 0 0 0 1 1 0 0 1 0 1 0 1 1 0 0 1 1 1 0.9994
1 0 0 1 1 1 0 1 0 1 0 0 1 1 1 0 0 0 1 0 0 1 1 1 1 1 0.9988
1 1 1 0 1 0 0 1 0 0 0 1 1 0 0 1 1 1 0 1 0 1 1 0 1 1 0.9988
1 0 1 1 0 1 1 0 0 1 1 0 1 1 0 0 0 0 1 1 0 1 1 1 0 1 0.9988
1 0 1 1 1 0 0 1 0 0 1 0 1 0 0 1 1 1 0 1 1 1 0 1 0 1 0.9988
1 1 1 1 1 0 0 1 0 0 1 0 1 0 0 0 1 1 0 1 0 1 0 1 1 1 0.9988
0 0 1 1 0 0 0 1 1 1 1 0 1 1 1 0 0 1 1 0 0 1 0 1 1 1 0.9982
1 0 0 1 1 1 0 1 0 1 1 0 1 0 1 1 1 0 0 0 0 1 1 1 0 1 0.9606
1 0 1 1 1 1 0 1 1 0 1 1 1 0 0 1 0 0 0 1 0 0 1 1 0 1 0.9600
1 1 1 0 0 0 1 1 0 0 1 0 1 0 0 1 1 1 0 1 1 1 0 1 0 1 0.9600
1 0 1 1 0 0 1 1 0 0 1 1 0 1 1 0 1 1 0 1 0 1 0 1 1 0 0.9600
1 0 0 1 0 0 0 1 1 1 1 1 1 0 1 1 1 0 0 1 1 1 0 0 0 1 0.9600
0 1 1 0 0 0 1 1 1 1 0 1 1 1 1 0 0 0 0 1 0 1 1 0 1 1 0.9600
1 0 1 1 1 1 0 1 1 0 1 0 0 1 1 0 0 0 0 1 1 1 0 1 0 1 0.9600
0 1 1 0 1 1 0 0 0 1 0 1 1 1 1 0 1 0 0 1 1 0 0 1 1 1 0.9227
1 1 0 0 1 1 0 0 1 0 1 0 1 1 1 0 0 1 0 1 0 1 1 1 0 1 0.9227
1 0 1 1 0 0 1 1 0 1 0 1 1 1 0 0 1 1 0 0 0 1 1 1 1 0 0.9227
1 0 1 1 1 1 0 1 0 0 0 0 1 0 0 1 0 1 1 0 1 1 1 1 0 1 0.9227
0 1 1 0 1 1 0 0 1 1 1 0 1 0 1 1 0 1 0 0 0 1 0 1 1 1 0.9227

The main goal of the Robust Design Problem (RDP) for MMSFN is to find the optimal
set of capacities assigned to arcs, ensuring network survival even under arc failures,
i.e., RS > 0. Initially, we studied a network with three sources and two sinks. The best
values for S and RS were 36 and 0.973585, respectively, with NP2(NF) = 10. Next, in a
network with two sources and two sinks, the best values for S and RS were found to be,
respectively, 169 and 0.996641 with NP2(NF) = 10. Finally, in a network with two sources
and three sinks, the best values for S and RS were 111 and 0.999923, respectively, with
NP2(NF) = 15. The NP2 for the inner GA (corresponding to NF, number of flow vectors)
was selected to be 10, 15, and 20 for all studied cases to find the optimal or near-optimal
value for the system reliability with NG2 equal to 100. The number of flow vectors for
Mathematics 2023, 11, 3902 15 of 16

MMFNs cannot be anticipated. Therefore, we searched for the optimal flows that maximize
the system reliability. According to [1,2,7], the obtained maximum value for NF number
does not exceed 20 as summarized in Table 7. Furthermore, we follow up the reliability
values of capacity vectors RX for each example, [1], o identify the best flow strategy with
maximum RX .

Table 7. The number of NF found by [1,2,7].

Approach Problem Maximum No. of NF

Maximizing reliability of the capacity vector by
[1] searching the single optimal flow vector with minimum 1
assignment cost
[2] Optimal system reliability under transmission budget 10
[7] Optimal system reliability evaluation under DD 9
Proposed Optimal system reliability under maximum capacities
20
approach (Robust Design Problem)

4. Conclusions
This paper presented a structural analysis of MMSFN to discuss and formulate the
Robust Design Problem for this type of SFN. Additionally, it successfully solves the RDP for
MMSFN using a two-stage GA approach. We divided the problem into two subproblems:
first, searching for the optimal capacity assigned to network components with a minimum
sum, and second, searching for the optimal lower vectors to achieve the maximum system
reliability value. The proposed GA-based approach was applied to a group of networks,
and it yielded satisfactory results. No comparisons were made as the RDP for MMSFNs
has not been discussed previously.

Author Contributions: Conceptualization, N.H.R., M.R.H. and H.A.S.; methodology, N.H.R., M.R.H.
and H.A.S.; software, validation, N.H.R., A.Y., M.R.H. and H.A.S.; formal analysis, S.B. and F.S.A.;
investigation, S.B.; writing—original draft preparation, N.H.R., H.A.S. and S.B.; writing—review and
editing, A.Y., M.R.H., S.B. and F.S.A.; visualization, A.Y. and M.R.H.; supervision, S.B. and M.R.H.;
funding acquisition, S.B. and F.S.A. All authors have read and agreed to the published version of
the manuscript.
Funding: Deputyship for Research and Innovation, Ministry of Education in Saudi Arabia—project
number MoE-IF-UJ-22-4220772-4.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Acknowledgments: The authors extend their appreciation to the Deputyship for Research & Inno-
vation, Ministry of Education in Saudi Arabia, for funding this research work through the project
number MoE-IF-UJ-22-4220772-4.
Conflicts of Interest: The authors declare no conflict of interest in this work.

References
1. Hassan, M.R. Maximizing reliability of the capacity vector for multi-source multi-sink stochastic-flow networks subject to an
assignment budget. J. Ind. Manag. Optim. 2021, 17, 1253–1267. [CrossRef]
2. Hassan, M.R. System Reliability Optimization of Multi-Source Multi-Sink Stochastic Flow Networks with Budget Constraint. Int.
J. Reliab. Qual. Saf. Eng. 2021, 28, 2150025. [CrossRef]
3. Forghani-Elahabad, M.; Kagan, N. Reliability evaluation of a stochastic-flow network in terms of minimal paths with budget
constraint. IISE Trans. 2018, 51, 547–558. [CrossRef]
4. Bobbio, A.; Terruggia, R.; Ciancamerla, E.; Minichino, M. Reliability analysis of multi-source multi-sink critical interacting
systems. In Proceedings of the 2011 3rd International Workshop on Dependable Control of Discrete Systems, Saarbruecken,
Germany, 15–17 June 2011; pp. 127–132. [CrossRef]
Mathematics 2023, 11, 3902 16 of 16

5. Sahinoglu, M.; Rice, B. Network reliability evaluation. In WIREs Computational Statistics; John Wiley & Sons: Hoboken, NJ, USA,
2010; Volume 2, pp. 189–211.
6. Hamed, A.Y.; Alkinani, M.H.; Hassan, M.R. A Genetic Algorithm to Solve Capacity Assignment Problem in a Flow Network.
Comput. Mater. Contin. 2020, 64, 1579–1586. [CrossRef]
7. Hassan, M. Solving Flow Allocation Problems and Optimizing System Reliability of Multi-Source Multi-Sink Stochastic Flow
Network. Int. Arab. J. Inf. Technol. (IAJIT) 2016, 13, 477–483.
8. Liu, Q.; Zhang, H.; Ma, X.; Zhao, Q. Genetic Algorithm-based Study on Flow Allocation in a Multicommodity Stochastic-flow
Network with Unreliable Nodes. In Proceedings of the Eighth ACIS International Conference on Software Engineering, Artificial
Intelligence, Networking, and Parallel/Distributed Computing (SNPD 2007), Qingdao, China, 30 July 2007–1 August 2007;
Volume 1, pp. 576–581. [CrossRef]
9. Chen, S.-G. Optimal double-resource assignment for the robust design problem in multistate computer networks. Appl. Math.
Model. 2014, 38, 263–277. [CrossRef]
10. Atzori, L.; Raccis, A. Network Capacity Assignment for Multicast Services Using Genetic Algorithms. IEEE Commun. Lett. 2004,
8, 403–405. [CrossRef]
11. Yang, G. Life Cycle Reliability Engineering; John Wiley & Sons: Hoboken, NJ, USA, 2007. [CrossRef]
12. Büsing, C.; Koster, A.M.C.A.; Schmitz, S. Robust minimum cost flow problem under consistent flow constraints. Ann. Oper. Res.
2021, 312, 691–722. [CrossRef]
13. Fowlkes, W.Y.; Creveling, C.M. Engineering Methods for Robust Product Design: Using Taguchi Methods in Technology and Product
Development; Addison-Wesley Publishing Company: Boston, MA, USA, 1995.
14. Chabrier, A.; Danna, E.; Le Pape, C.; Perron, L. Solving a Network Design Problem. Ann. Oper. Res. 2004, 130, 217–239. [CrossRef]
15. Yong, K. Robust Design and Reliability. Trans. Nanjing Univ. Aeronaut. Astronaut. 1998, 15, 9–14.
16. Cho, B.R.; Shin, S. Quality Improvement and Robust Design Methods to a Pharmaceutical Research and Development. Math.
Probl. Eng. 2012, 2012, 193246. [CrossRef]
17. Chen, S.-G. An optimal capacity assignment for the robust design problem in capacitated flow networks. Appl. Math. Model. 2012,
36, 5272–5282. [CrossRef]
18. Hamdy, N.; Hassan, M.R.; Hussein, M.E. A genetic algorithm to solve the robust design problem for a Flow Network with Node
Failure. Trans. Networks Commun. 2020, 8, 1–10. [CrossRef]
19. Radwan, N.H.; Hassan, M.R.; Hussein, M.E. Solving the Robust Design Problem for a Two-Commodity Flow Network with Node
Failure. Am. J. Eng. Appl. Sci. 2020, 13, 837–845. [CrossRef]
20. Chen, S.-G.; Lin, Y.-K. An Approximate Algorithm for the Robust Design in a Stochastic-Flow Network. Commun. Stat.—Theory
Methods 2010, 39, 2440–2454. [CrossRef]
21. López-Prado, J.L.; Vélez, J.I.; Garcia-Llinás, G.A. Reliability Evaluation in Distribution Networks with Microgrids: Review and
Classification of the Literature. Energies 2020, 13, 6189. [CrossRef]
22. Al-Shaalan, A.M. Reliability Evaluation of Power Systems. In Reliability and Maintenance-An Overview of Cases; IntechOpen:
London, UK, 2020.
23. Niu, Y.-F.; Xu, X.-Z.; He, C.; Ding, D.; Liu, Z.-Z. Capacity Reliability Calculation and Sensitivity Analysis for a Stochastic Transport
Network. IEEE Access 2020, 8, 133161–133169. [CrossRef]
24. Satitsatian, S.; Kapur, K.C. An algorithm for lower reliability bounds of multistate two-terminal networks. IEEE Trans. Reliab.
2006, 55, 199–206. [CrossRef]
25. Hassan, M.R.; Abdou, H. Multi-Source Multi-Sink Stochastic-Flow Networks Reliability under Time Constraints. Indian J. Sci.
Technol. 2019, 12, 22. [CrossRef]
26. Abed, M.H.; Tang, A.Y. Hybridizing Genetic Algorithm and Record-to-Record Travel Algorithm for Solving Uncapacitated
Examination Timetabling Problem. electronic J. Comput. Sci. Inf. Technol. 2013, 4, 25–31.
27. Sharma, M. Role and Working of Genetic Algorithm in Computer Science. Int. J. Comput. Appl. Inf. Technol. 2013, 2, 27–32.
28. Malik, A. A Study of Genetic Algorithm and Crossover Techniques. Int. J. Comput. Sci. Mob. Comput. 2019, 8, 335–344.
29. Mirjalili, S. Studies in Computational Intelligence. In Evolutionary Algorithms and Neural Networks; Springer: Berlin/Heidelberg,
Germany, 2019; Volume 780.
30. Lin, Y.-K. A simple algorithm for reliability evaluation of a stochastic-flow network with node failure. Comput. Oper. Res. 2001, 28,
1277–1285. [CrossRef]
31. Zuo, M.J.; Tian, Z.; Huang, H.-Z. An efficient method for reliability evaluation of multistate networks given all minimal path
vectors. IIE Trans. 2007, 39, 811–817. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.