Yugoslav Journal of Operations Research

34 (2024), Number 2, 201-227

DOI: https://doi.org/10.2298/YJOR230417024J

A PYTHAGOREAN HESITANT FUZZY PROGRAMMING

APPROACH AND ITS APPLICATION TO MULTI
OBJECTIVE RELIABILITY OPTIMIZATION PROBLEM

Swarup JANA
Department of Mathematics,
University of Kalyani,
Kalyani, West Bengal 741235, India
[email protected]
Sahidul ISLAM
Department of Mathematics,
University of Kalyani,
Kalyani, West Bengal 741235, India
[email protected]

Received: April 2023 / Accepted: September 2023

Abstract: Decision-making problems can often be effectively solved using traditional

optimization methods that have a clearly defined configuration. However, in real-world
scenarios, decision-makers frequently encounter doubt or hesitation, making it challenging
to precisely specify certain parameters. As a result, they often seek input from different
experts, leading to conflicting values and varying levels of satisfaction among decision-
makers. This uncertainty and lack of crisp values make decision-making problems
inherently non-deterministic. In this paper, a novel Pythagorean hesitant fuzzy (PHF)
programming method is designed to address the challenges of optimization problems with
multiple objectives. Here PHF aggregation operators are used to aggregate the PHF
memberships and non-memberships of the objectives. Additionally, to account the
uncertainties of the parameters of the optimization problem Parabolic Pythagorean fuzzy
number is used and centroid method is applied for defuzzification. We solved an example
of multi objective optimization problem of manufacturing system to demonstrate our
proposed method and finally, presented a case study on reliability optimization model for
Life Support Systems, where the primary objectives are to maximize system reliability and
minimize cost. The result is compared with other existing methods by degree of closeness.
Keywords: Parabolic Pythagorean fuzzy number, Pythagorean hesitant fuzzy set,
Pythagorean hesitant fuzzy aggregation operator, multi objective reliability optimization.
MSC: 03B52, 90C29, 90B25, 90B30.
202 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

1. INTRODUCTION
In today's complex and uncertain world, decision-making procedures frequently entail
a number of competing goals that need to be optimized to achieve the most desirable
outcomes. Such situations are prevalent in various disciplines including engineering,
finance, transportation, and environmental management, where reliability is a critical
concern. Although it might be challenging to find optimal solution that successfully
addresses all goals, so decision makers (DMs) accept compromise solutions. Most of the
time the goal of the DMs and the parameter in the objective and constraints are not
precisely known. To deal with such issues like ambiguity and inaccuracy that inevitably
arise in decision-making problems, Zadeh [1] invented fuzzy sets (FSs). Later on, Bellman
and Zadeh [2] presented fuzzy decision set in uncertainty and ambiguity situations. Then
to solve uncertain optimization problem, Zimmerman [3] introduced fuzzy programming
approach. Various fuzzy programming methods have been proposed in literature to solve
multi-objective optimization problem (MOOP) including transportation, supplier
selection, inventory control, portfolio management, reliability optimization, etc. In order
to find the best compromise solution to a multi-objective transportation problem, El-
Wahed [4] introduced fuzzy programming approach. Parra et al. [5] found optimal
portfolio of investors using fuzzy goal programming. Mahapatra et al. [6] solved reliability
optimization model of series system by introducing max-min and max-additive operators
based fuzzy goal geometric programming. Majumder et al. [7] used uncertainty theory to
create three separate models allowed them to address the problem of multi-item fixed
charge solid transportation. Further, Majumder et al. [8] used a chance constraint model
and an expected value model to solve the multi-objective shortest route problem by treating
the parameters as uncertain variables. An uncertain variable-based multi-objective minimal
spanning tree problem is solved by Majumder et al. [9]. Afterword, Majumder et al. [10]
solved the multi-objective mean entropy portfolio selection issue by making use of the
investors' hazy securities.
In fuzzy environment, only membership degree of an element is considered which is
not fulfill DMs choice every time. Most of the time non-membership degree with
membership is required to express the uncertainty in optimization goal, parameters and
solution set involved in the problem. To overcome this, Atanassov [11] invented the idea
of intuitionistic fuzzy (IF) set (IFS) as generalized version of FS where the membership
and non-membership degrees added together are less than or equal to one. After that,
Angelov [12] proposed a programming approach in IF environment to solve MOOP
considering IF decision set. Further, lots of work has been done to solve MOOP in IF
environments. Dey and Roy [13] developed an IF programming approach to find the
optimal design of multi objective plane truss structure. Garg et al. [14] solved reliability
optimization problem (ROP) by IF optimization technique in interval environment. A
optimization model of biological treatment process on industrial waste water is solved by
Ghosh P. et al. [15] using IF goal geometric programming.
Reliability engineering is a crucial phase in the design and creation of a technological
system. Considering availability of system resources effective technique to improve
system reliability has always been the reliability engineer's top priority. The optimization
model includes a variety of coefficients and parameters that are constantly inaccurate and
ambiguous in nature due to the DM's uncertainty in everyday life. A fuzzy technique is
used to analyze this type of nature in MOOP. Ravi et al. [16] complex system reliability
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 203

optimization model as fuzzy MOOP. Sharma et al. [17] presented an analysis of system
reliability in IF environment. Islam and Kundu [18] applied neutrosophic optimization
technique to solve complex system reliability optimization model (ROM). A neutrosophic
goal geometric programming approach is presented by Kundu and Islam [19] to solve
multi objective ROP.
Yager [20] presented Pythagorean fuzzy (PF) set (PFS) as generalization of FS and IFS
where the sum of the squares of the membership and non-membership degrees is ≤ 1. In
last few years many decision-making problems has been solved in PF environment.
Recently Fei et al. [21] solved a decision making problem considering Pythagorean fuzzy
number (PFN) and interval valued PFN. Akram et al. [22] solved fully Pythagorean fuzzy
linear programming problem considering triangular PFN. In a PF environment, Luqman et
al. [23] designed digraph and matrix technique to assess risk in a Failure Mode and Effects
Analysis. A minimal spanning tree agglomerative hierarchical clustering technique is
proposed by Habib et al. [24] utilizing PF distance and similarity measurements. By taking
into account linguistic capacities and flows, Akram et al. [25] created a novel Pythagorean
fuzzy maximum flow method to handle various optimization problems using PF
information. To handle the multi-attribute decision-making problem in a PF environment,
Wan et al. [26] designed a Pythagorean fuzzy mathematical programming approach.

The main motivation of this study:

Most of the current approaches presented in the literature integrate multiple
objectives into a single objective by using real-valued and utility functions. The
utility functions can be expressed in many different ways, including the product form, max-
min, weighted max-min and weighted sum. Typically, these approaches employ the
preferences of the DM. Using a useful function, this choice is converted to a mathematical
expression. For instance, the weighted sum method can't guarantee that the levels of
achievement of fuzzy goals are consistent with desired relative weights or the expectations
of the DM. Zimmermann [3] employed the "min" operator in his suggested fuzzy
programming technique due to the ease of calculation, it does not guarantee a solution
which is free of dominance. Li and Lai [27] developed some aggregation operators based
on weighted root-power mean to solve the multi-objective transportation problem
worldwide in order to get over the aforementioned limitation. Further, Liu et al. [28] used
a modified s-curve membership function to address the supplier selection problem by
weighted root mean power aggregation operator.
In the decision-making process, hesitation is actually a trivial concern. It describes the
situation in which the decision-maker(s) are unsure of the precise values of the parameters,
though there is considerable uncertainty around a few possible outcomes. It may be
handled using a hesitant fuzzy (HF) set (HFS) [29] by assigning a set of varying
membership degrees to an element in the set. A hesitant mathematical programming
approach was created Zhang et al. [30] to address MCDM issues in the context of HF
elements. Bharati [31] suggested a HF computational technique for the production
planning problem. Additionally, aggregation procedures that were based on HFSs were
expanded. A novel set of hesitant aggregation operators was addressed by Xia et al. [32]
and their use in decision-making problems was provided. Additionally, Xia and Xu [33]
established number of hesitant aggregations operators and demonstrated how to employ
them to address decision-making issues. For hybrid multi-criteria group decision making
with incomplete criterion weight information and HF membership degrees, Xu et al. [34]
204 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

suggested a new HF mathematical programming technique. In order to build a HF

programming approach based on the linear programming methodology for
multidimensional analysis of preference, Wan et al. [35] investigated a hybrid problem of
multiple attribute decision making with unreliable attribute weight data. Recently F.
Ahmad et al. [36] proposed a optimization method to solve optimization problem taking
the advantage of HF aggregation operator.
Further, as an extension of hesitant FS, hesitant IFS, hesitant PFS are suggested to
solve MOOP. Recently Bharti [37] developed an interactive optimization algorithm in
hesitant IF environment to solve linear optimization problem. Adhami and Ahmad [38]
solved transportation problem by introducing an interactive optimization algorithm in
Pythagorean hesitant fuzzy (PHF) environment. Khan M.S.A et al. [39] introduced PHF
aggregation operator (PHFAO) to solve multi criterion decision making problem. Later on,
extension of PHFAO is established and applied to MCDM problem [40-41]. Best of our
knowledge still there are no use of PHFAO to solve MOOP. Fulfilling this gap in this study
we have presented an optimization algorithm using PHF weighted averaging (PHFWA)
operator and PHF weighted geometric (PHFWG) operator. Also, here the parameters of
the problem are taken as parabolic PFN (PPFN) and centroid method is used for
defuzzification. To illustrate the validity and acceptability of the proposed method, a
mathematical model of manufacturing system, a ROM for a life support system is solved,
and the outcomes are compared to the existing approach using HF aggregation operator by
closeness degree to ideal solution [42].
The main contribution of this study are as follows:
1) To handle the uncertain parameter of the optimization problem we presented
PPFN.
2) A PHF programming approach is developed to solve MOOP by using the
weighted sum of aggregated PHF membership and non-membership.
3) To demonstrate the effectiveness of the suggested approach, we addressed a
practical manufacturing system and a real-world multi-objective ROM
concerning the life support system within a space capsule. We then conducted a
comparative analysis of the outcomes against established methods.
4) By varying the weight, sensitivity analysis of the results is shown and compare
the results by degree of closeness to ideal solution.

The remaining portions of the paper are arranged as: section 2 describes the basic
definitions, PPFN and PHFAO. In section 3, a MOOP using PHFAO is formulated.
Application of the proposed method on multi objective ROP is shown in section 4. The
result and discussion part are in section 5, The limitation of this article is highlighted in
section 6 and at last the conclusion of the article is discussed in section 7.

2. PRELIMINARIES
2.1. Basic definitions
We have discussed some fundamental definitions of terms related to various fuzzy sets
in this section.
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 205

Definition 1 (HFS [27]). 𝑋 be a fixed set, a HFS, A on 𝑋 is defined as 𝐴 =

{⟨𝑥, ℎ𝐴 (𝑥)⟩ ∣ 𝑥 ∈ 𝑋} , where ℎ𝐴 (𝑥) is a collection of finite membership values in [0, 1].
ℎ𝐴 (𝑥) is referred to as a HF element also.
Definition 𝟐 [31]. The intersection of two HFSs ℎ1 and ℎ2 in a given set X is defined as
follows:
ℎ1 ∩ ℎ2 =∪𝛼1∈ℎ1 ,𝛼2∈ℎ2 min{𝛼1 , 𝛼2 }
Definition 3. PFS [20]. Given a set 𝑆, the following is the PFS, 𝑃̃ in 𝑆 using ordered
triplets:
𝑃̃ = {𝑠, 𝜇𝑃̃ (𝑆), 𝑣𝑃̃ (𝑆) ∣ 𝑠 ∈ 𝑆}
where 𝜇𝑃̃ (𝑠),𝑣𝑃̃ (𝑠): 𝑆 → [0,1] are membership and non-membership function of the
element 𝑥 into the set 𝑃̃ , respectively, with 0 ≤ 𝜇𝑃̃ (𝑠)2 + 𝑣𝑃̃ (𝑠)2 ≤ 1. Additionally, the
value of uncertainty is indicated by√1 − 𝜇𝑃̃ (𝑠)2 − 𝑣𝑃̃ (𝑠)2 . The PFS differs from IFS in
that it places limits on the squares of the membership and non-membership degrees added
together.
Definition 4. PHFS [38]. 𝑌 be an arbitrary set. The following describes a PHFS on
Y: 𝑃ℎ = {〈𝑦, 𝜇𝑃ℎ (𝑦), 𝑣𝑃ℎ (𝑦)〉 ∣ 𝑦 ∈ 𝑌}, where 𝜇𝑃ℎ (𝑦), 𝑣𝑃ℎ (𝑦) are finite values in [0, 1],
representing degrees of PHF membership and non-membership of 𝑦 to 𝑃ℎ , with 0 ≤
𝛼, 𝛽 ≤ 1 and 0 ≤ 𝛼 2 + 𝛽 2 ≤ 1, where 𝛼 ∈ 𝜇𝑃ℎ (𝑦), 𝛽 ∈ 𝑣𝑃ℎ (𝑦) for all 𝑦 ∈ 𝑌. For
simplicity, the 𝑃ℎ (𝑦) = {𝜇𝑃ℎ (𝑦), 𝑣𝑃ℎ (𝑦)} is called a PHF element.
Definition 5. [38]. The union of two PHFSs 𝑃ℎ1 and 𝑃ℎ2 in on an arbitrary set Y can be
defined as follows:
𝑃ℎ1 ∪ 𝑃ℎ2 = {𝜇ℎ ∈ (𝜇ℎ1 ∪ 𝜇ℎ2 ) ∣ 𝜇ℎ ≥ 𝑚𝑎𝑥(𝑚𝑖𝑛{𝜇ℎ1 ∪ 𝜇ℎ2 }),
𝑣ℎ ∈ (𝑣ℎ1 ∪ 𝑣ℎ2 ) ∣ 𝑣ℎ ≤ 𝑚𝑖𝑛(𝑚𝑎𝑥{𝑣ℎ1 ∪ 𝑣ℎ2 })}
Definition 6. [38]. The intersection of two PHFSs 𝑃ℎ1 and 𝑃ℎ2 in on an arbitrary set Y can
be defined as follows:
𝑃ℎ1 ∩ 𝑃ℎ2 = {𝜇ℎ ∈ (𝜇ℎ1 ∩ 𝜇ℎ2 ) ∣ 𝜇ℎ ≤ 𝑚𝑖𝑛(𝑚𝑎𝑥{𝜇ℎ1 ∩ 𝜇ℎ2 }),
𝑣ℎ ∈ (𝑣ℎ1 ∩ 𝑣ℎ2 ) ∣ 𝑣ℎ ≥ 𝑚𝑎𝑥(𝑚𝑖𝑛{𝑣ℎ1 ∩ 𝑣ℎ2 })}
2.2. Parabolic Pythagorean fuzzy number (PPFN) and defuzzification

Definition 7. A PPFN, 𝑃̃ = {(𝑎, 𝑏, 𝑐); 𝑝, 𝑞} is a PFS on ℝ, where 𝑝, 𝑞 are maximum degree

of membership and minimum degree of non-membership respectively, here membership
(𝜇𝑃̃ ) and non-membership(𝜗𝑃̃ ) are defined as:
206 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

1
𝑥−𝑎 2
𝑝( ) 𝑎≤𝑥<𝑏
𝑏−𝑎
𝑝 𝑥=𝑏
𝜇𝑃̃ = 1
𝑐−𝑥 2
𝑝( ) 𝑏<𝑥≤𝑐
𝑐−𝑏
{ 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
1
(𝑏 − 𝑥) + 𝑞 2 (𝑥 − 𝑎) 2
[ ] 𝑎≤𝑥<𝑏
𝑏−𝑎
𝑞 𝑥=𝑏
𝜗𝑃̃ = 1
(𝑥 − 𝑏) + 𝑞 2 (𝑐 − 𝑥) 2
[ ] 𝑏<𝑥≤𝑐
𝑐−𝑏
{ 1 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
̃
In particular, take 𝑝 = 1, 𝑞 = 0 then we get PPFN as 𝑃 = {(𝑎, 𝑏, 𝑐); (𝑎′ , 𝑏 ′ , 𝑐 ′ )} and
membership and non-membership are shown in Figure 1.
𝜇(𝑥)
𝜗(𝑥)

0 𝑎′ 𝑎 𝑏 𝑥
𝑐 𝑐′

Figure1: Parabolic Pythagorean fuzzy number

1
𝑥−𝑎 2
( ) 𝑎≤𝑥<𝑏
𝑏−𝑎
𝜇𝑃̃ (𝑥) = 1 𝑥=𝑏
1
𝑐−𝑥 2
( ) 𝑏<𝑥≤𝑐
𝑐−𝑏
{ 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
1
𝑏−𝑥 2
( ) 𝑎′ ≤ 𝑥 < 𝑏
𝑏−𝑎′

And 𝜗𝑃̃ (𝑥) = 0 𝑥=𝑏

1
𝑥−𝑏 2
( ) 𝑏 < 𝑥 ≤ 𝑐′
𝑐 ′ −𝑏
{ 1 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 207

By COA method, here the Centre of area of membership and non-membership are
calculated individually, i.e.
∫ 𝑥𝜇𝑃̃ (𝑥)𝑑𝑥 ∫ 𝑥𝜗𝑃̃ (𝑥)𝑑𝑥
𝑅1 (𝑃̃ ) = 𝑥 𝑎𝑛𝑑 𝑅2 (𝑃̃ ) = 𝑥
∫𝑥 𝜇𝑃̃ (𝑥)𝑑𝑥 ∫𝑥 𝜗𝑃̃ (𝑥)𝑑𝑥
𝑅1 (𝑃̃)+𝑅2 (𝑃̃)
Now the defuzzied value of the PPFN is defined by, 𝑅(𝑃̃) = .
2
𝑏 𝑥−𝑎 1/2 𝑐 𝑐−𝑥 1/2
∫𝑎 𝑥(𝑏−𝑎) 𝑑𝑥+∫𝑏 𝑥(𝑐−𝑏) 𝑑𝑥 2𝑎+𝑏+2𝑐
So, here from the membership function 𝑅1 (𝑃̃ ) = 𝑏 𝑥−𝑎 1/2 𝑐 𝑐−𝑥 1/2
=
5
∫𝑎 (𝑏−𝑎) 𝑑𝑥+∫𝑏 (𝑐−𝑏) 𝑑𝑥
𝑏 𝑏−𝑥 1/2 𝑐 𝑥−𝑏 1/2
∫𝑎 𝑥( ) 𝑑𝑥+∫𝑏 𝑥( ′ ) 𝑑𝑥 2𝑎′ +𝑏+2𝑐 ′
𝑏−𝑎′
And from non-membership function 𝑅2 (𝑃̃ ) = 𝑏 𝑏−𝑥 1/2
𝑐 −𝑏
𝑐 𝑥−𝑏 1/2
=
5
∫𝑎 ( ′ ) 𝑑𝑥+∫𝑏 ( ′ ) 𝑑𝑥
𝑏−𝑎 𝑐 −𝑏
(2𝑎+𝑏+2𝑐)+(2𝑎′ +𝑏+2𝑐 ′ )
Hence, 𝑅(𝑃̃ ) =
10
2.3. PHF aggregation operators

Definition 8(PHFWA operator [39]). Let ℎ̂𝑖 = ⟨𝐴𝑖 , 𝐵𝑖 ⟩ be a collection of all PHFNs, and
𝑤 = (𝑤1 , 𝑤2 , … , 𝑤𝑛 )𝑇 is weight vector of ℎ̂𝑖 with 𝑤𝑖 ≥ 0 , 𝑤𝑖 ∈ [0,1] and ∑𝑛𝑖=1 𝑤𝑖 = 1.
Then, a mapping ∅: PHFN 𝑛 → 𝑃𝐻𝐹𝑁 is called PHFWA operator and defined as
𝑃𝐻𝐹𝑊𝐴(ℎ̂1 , ℎ̂2 , … , ℎ̂𝑛 ) = 𝑤1 ℎ̂1 ⊕ 𝑤2 ℎ̂2 ⊕ … ⊕ 𝑤𝑛 ℎ̂𝑛

𝑛 𝑛
2 )𝑤𝑖 𝑤
= ⋃ {√1 − ∏(1 − 𝛼𝑖 } , {∏ 𝛽𝑖 𝑖 }
𝛼1 ∈𝐴1 ,𝛼2 ∈𝐴2 ,…….𝛼𝑛 ∈𝐴𝑛 𝑖=1 𝑖=1
𝛽1 ∈𝐵1 ,𝛽2 ∈𝐵2 ,…….𝛽𝑛 ∈𝐵𝑛 { }
Definition 9 (PHFWG operator) [39]. Let ℎ̂𝑖 = ⟨𝐴𝑖 , 𝐵𝑖 ⟩ be a collection of all PHFNs, and
𝑤 = (𝑤1 , 𝑤2 , … , 𝑤𝑛 )𝑇 is weight vector of ℎ̂𝑖 with 𝑤𝑖 ≥ 0 , 𝑤𝑖 ∈ [0,1] and ∑𝑛𝑖=1 𝑤𝑖 = 1.
Then, a mapping ∅: PHFN 𝑛 → 𝑃𝐻𝐹𝑁 is called PHFWG operator and defined as
𝑤 𝑤 𝑤
𝑃𝐻𝐹𝑊𝐺(ℎ̂1 , ℎ̂2 , … , ℎ̂𝑛 ) = ℎ̂1 1 ⨂ℎ̂2 2 ⨂ … . . . ⨂ℎ̂𝑛 𝑛

𝑛 𝑛
𝑤
= ⋃ {∏ 𝛼𝑖 𝑖 } , {√1 − ∏(1 − 𝛽𝑖2 )𝑤𝑖 }
𝛼1 ∈𝐴1 ,𝛼2 ∈𝐴2 ,…….𝛼𝑛 ∈𝐴𝑛 𝑖=1 𝑖=1
𝛽1 ∈𝐵1 ,𝛽2 ∈𝐵2 ,…….𝛽𝑛 ∈𝐵𝑛 { }

3. MULTI OBJECTIVE OPTIMIZATION MODEL USING

AGGREGATION OPERATORS BASED ON PHF SET
3.1. Multi objective optimization problem

A MOOP with 𝑙 objectives, 𝑚 constraints, and 𝑛 decision variables are formulated as

follows:
Minimize (𝑓1 , 𝑓2 , … … … . . 𝑓𝑙 )
208 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

Subject to, 𝑎𝑖 (𝑥) ≤ 𝑜𝑟 ≥ 𝑜𝑟 = 𝑏𝑖 , 𝑖 = 1, … … 𝑚.

Where 𝑥 = (𝑥1 , … … . . 𝑥𝑛 ) and 𝑥𝑗 ≥ 0 for all 𝑗 = 1, … … . 𝑛. (1)

Here 𝑓𝑙 are objectives, 𝑎𝑖 are real valued functions, 𝑏𝑖 are real numbers, 𝑥𝑗 are the decision
variables.

Using weighted sum with weight vector 𝑤 𝑇 = (𝑤1 , 𝑤2 , … … . . , 𝑤𝑙 ) and ∑𝑙𝑘=1 𝑤𝑘 = 1,

Problem (1) becomes,
Minimize (𝑤1 𝑓1 + 𝑤2 𝑓2 + ⋯ + 𝑤𝑙 𝑓𝑙 )

Subject to, 𝑎𝑖 (𝑥) ≤ 𝑜𝑟 ≥ 𝑜𝑟 = 𝑏𝑖 , 𝑖 = 1, … … 𝑚.

Where 𝑥 = (𝑥1 , … … . . 𝑥𝑛 ) and 𝑥𝑗 ≥ 0 for all 𝑗 = 1, … … . 𝑛. (2)

3.2. PHF membership and non-membership function
We have 𝑙 solutions for the unique solution of each objective function:(𝑋1 , 𝑋 2 , … . , 𝑋 𝑙 ).
Then, calculating value of each objective by all solutions we get the lower and upper bound
of objectives:
𝐿𝑘 = min[𝑓𝑘 (𝑋 𝑖 )] 𝑎𝑛𝑑 𝑈𝑘 = max[𝑓𝑘 (𝑋 𝑖 )] ∀𝑖 = 1,2,3, … , 𝑙 for all 𝑘 = 1,2, … . , 𝑙.
The linear membership and non-membership have huge applications in uncertainty and
a straightforward construction. Following is a definition of the linear membership function
of minimization objective under PHFS (Figure 2):
1, if 𝑓𝑘 < 𝐿𝑘 ;
𝐸 𝑈𝑘 − 𝑓𝑘
𝜇𝑝ℎ𝑛 (𝑍𝑘 (𝑥)) = 𝛿𝑛 , if 𝐿𝑘 ≤ 𝑓𝑘 ≤ 𝑈𝑘 + α;
(𝑈𝑘 + α) − 𝐿𝑘
{0, if 𝑓𝑘 > 𝑈𝑘 + 𝛼

Figure 2: Linear membership and non-membership function

and linear non-membership under PHFS is:

 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 209

0, if 𝑓𝑘 < 𝐿𝑘 ;
𝐸 𝑓𝑘 − 𝐿𝑘
𝜗𝑝ℎ𝑛 (𝑍𝑘 (𝑥)) = 𝛾𝑛 , if 𝐿𝑘 ≤ 𝑓𝑘 ≤ 𝑈𝑘 + α;
(𝑈𝑘 + α) − 𝐿𝑘
{1, if 𝑓𝑘 > 𝑈𝑘 + α
where 𝛿𝑛 , 𝛾𝑛 ∈ [0,1] are assigned to expert (𝐸𝑛 ).

Pareto optimal solution: pareto optimal solution to a MOOP is a basic feasible solution
𝑥 ∗ in feasible space of the problem, iff ∄ x s.t 𝑓𝑖 (𝑥) ≤ 𝑓𝑖 (𝑥 ∗ ) for all 𝑖 = 1, … . 𝑙 and 𝑓𝑖 (𝑥) <
𝑓𝑖 (𝑥 ∗ ) for at least one 𝑖 = 1, … . 𝑙.

PHF pareto optimal solution: a point 𝑥 ∗ in solution space(𝑋) is called PHF pareto
𝐸
optimal solution , iff ∄ 𝑥 in 𝑋 s.t 𝑓𝑖 (𝑥) ≤ 𝑓𝑖 (𝑥 ∗ ) for all 𝑖 = 1,2, … … 𝑙 with 𝜇𝑖 𝑘 (𝑓𝑖 (𝑥)) ≥
𝐸𝑘 𝐸 𝐸
𝜇𝑖 (𝑓𝑖 (𝑥 ∗ )) and 𝜗𝑖 𝑘 (𝑓𝑖 (𝑥)) ≤ 𝜗𝑖 𝑘 (𝑓𝑖 (𝑥 ∗ )) for all 𝑖 and inequality holds for at least one
𝐸 𝐸
𝑖, where 𝜇𝑖 𝑘 (𝑥), 𝜗𝑖 𝑘 (𝑥) are PF membership and non-membership respectively.

3.3. Evaluation of MOOP using PHFAO

PHFAOs, which are previously mentioned, may serve as the foundation for the
following type of utility function for MOOP:
𝑙
𝐸1 𝐸2 𝐸𝑛 𝐸1 𝐸2 𝐸𝑛 )
Max(PHFAO)(𝜇 , 𝜇 , … , 𝜇 ; 𝜗 , 𝜗 , … , 𝜗 = ∑ 𝑊𝑘 (PHFAO)
𝑘=1

Where, 𝑊𝑘 (∀𝑘 = 1,2, … , 𝑙), such that ∑𝑙𝑘=1 𝑊𝑘

= 1 is weight assigned by DMs to each
objective. PHFAO denotes the family of PHF aggregation operators.

The fuzzy set, according to Bellman and Zadeh [2], consists of three notions: fuzzy
goal (G), and fuzzy constraints (C), fuzzy decision (D). These concepts have been
integrated into numerous real-world decision-making applications within fuzzy
environments. A fuzzy decision set is formally defined as: 𝐷 = 𝐺 ∩ 𝐶 =
∪𝛼𝑛∈𝐺,𝛽𝑛∈𝐶 𝑚𝑖𝑛{(𝛼𝑛 , 𝛽𝑛 )}.
Now, the PHF decision set 𝐷𝑝ℎ is defined as follows, using PHF constraints (C) and
goals (Z):
𝐷𝑝ℎ = 𝑍 ∩ 𝐶 = (∩𝑙𝑘=1 𝑍𝑘 ) ∩ (∩𝑗=1 𝑚
𝐶𝑗 )
𝑥,∪ min{(𝛼1 , 𝛽1 ), (𝛼2 , 𝛽2 ), … , (𝛼𝑛 , 𝛽𝑛 )}(𝑥),
={ }
max{(𝜃1 , 𝜌1 ), (𝜃2 , 𝜌2 ), … , (𝜃𝑛 , 𝜌𝑛 )}(𝑥) ∣ 𝑥 ∈ 𝑋
= {𝑥, {𝜇 𝐸1 , 𝜇 𝐸2 , … , 𝜇 𝐸𝑛 ; 𝜗𝐸1 , 𝜗𝐸2 , … , 𝜗𝐸𝑛 }(𝑥) ∣ 𝑥 ∈ 𝑋}
𝑤ℎ𝑒𝑟𝑒, 𝜇 𝐸1 = min(𝛼1 , 𝛽1 ); 𝜗𝐸1 = 𝑚𝑎𝑥(𝜃1 , 𝜌1 )
𝜇 𝐸2 = min(𝛼2 , 𝛽2 ); 𝜗𝐸2 = 𝑚𝑎𝑥(𝜃2 , 𝜌2 )
⋮
𝜇 𝐸𝑛 = min(𝛼𝑛 , 𝛽𝑛 ); 𝜗𝐸𝑛 = 𝑚𝑎𝑥(𝜃𝑛 , 𝜌𝑛 )
where, 𝜇 𝐸𝑛 , 𝜗𝐸𝑛 are PHF membership and non-membership by 𝑛th experts.
210 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

Now MOOP in PHF environment become as:

Max PHFAO(𝜇 𝐸1 , 𝜇 𝐸2 , … , 𝜇 𝐸𝑛 ; 𝜗𝐸1 , 𝜗𝐸2 , … , 𝜗𝐸𝑛 ) = Max ∑𝑙𝑘=1 𝑊𝑘 (PHFAO)

subject to

𝜇 𝐸1 (𝑓𝑘 (𝑥)) ≥ 𝛼𝑘1 ; ϑE1 (𝑓𝑘 (𝑥)) ≤ 𝛽𝑘1

𝜇 𝐸2 (𝑓𝑘 (𝑥)) ≥ 𝛼𝑘2 ; ϑE2 (𝑓𝑘 (𝑥)) ≤ 𝛽𝑘2
… … ..

𝜇 𝐸𝑛 (𝑓𝑘 (𝑥)) ≥ 𝛼𝑘n , ϑE𝑛 (𝑓𝑘 (𝑥)) ≤ 𝛽𝑘𝑛

𝑎𝑖 (𝑥) ≤ 𝑜𝑟 = 𝑜𝑟 ≥ 𝑏𝑖 𝑖 = 1,2, … … . 𝑚
2 2
𝑥 ≥ 0 ,0 ≤ 𝛼𝑘𝑟 + 𝛽𝑘𝑠 ≤ 1 𝑎𝑛𝑑 𝛼𝑘𝑟 ≥ 𝛽𝑘𝑠 𝑟, 𝑠 ∈ {1, … … . 𝑛} (3)
𝑊𝑘 (∀𝑘 = 1,2, … , 𝑙), such that ∑𝑙𝑘=1 𝑊𝑘 = 1 are weights given to the objective by experts.
𝜔 = (𝜔1 , 𝜔2 , … , 𝜔𝑛 ) is the weight vector allocated to each PHF element, 𝜔𝑖 ∈ [0,1] and
∑𝑛𝑖=1 𝜔𝑖 = 1.𝛼𝑘𝑛 , 𝛽𝑘𝑛 are minimum membership and maximum non-membership to each
PHF membership function.

Here, 𝜇 𝐸i (𝑓𝑘 (𝑥)) : membership to the 𝑘 𝑡ℎ objective by 𝑖 𝑡ℎ expert.

𝜗𝐸i (𝑓𝑘 (𝑥)) : non-membership to the 𝑘 𝑡ℎ objective by 𝑖 𝑡ℎ expert.

Theorem1: For non-zero objective weight vector, a unique optimal solution (𝑥 ∗ , 𝛼 ∗ , 𝛽 ∗ )

of Problem (3) is a pareto optimal solution (𝑥 ∗ ) of problem (1). Where, 𝛼 ∗ =
(𝛼1∗ , 𝛼2∗ , … … . , 𝛼𝑙∗ ), 𝛽 ∗ = (𝛽1∗ , 𝛽2∗ , … … . , 𝛽𝑙∗ ) and 𝛼𝑖∗ = (𝛼𝑖1
∗ ∗
, 𝛼𝑖2 ∗
, … … . . , 𝛼𝑖𝑛 ),and 𝛽𝑖∗ =
∗ ∗ ∗
(𝛽𝑖1 , 𝛽𝑖2 , … … . . , 𝛽𝑖𝑛 ).

Proof: (for PHFWA operator)

𝜔
Consider 𝑓(𝛼, 𝛽) = ∑𝑙𝑘=1 𝑊𝑘 (√1 − ∏𝑛𝑖=1 (1 − 𝛼𝑘𝑖 2 )𝜔𝑖
− ∏𝑛𝑖=1 𝛽𝑘𝑖𝑖 )
Let (𝑥 ∗ , 𝛼 ∗ , 𝛽 ∗ ) is optimal solution of problem (3) then 𝑓(𝛼 ∗ , 𝛽 ∗ ) > 𝑓(𝛼, 𝛽) for all (𝑥, 𝛼, 𝛽)
in the feasible space of the problem (3).
Suppose 𝑥 ∗ is not a pareto optimal solution of problem (1) then there exist 𝑥 ∗∗ in feasible
space (𝑋) such that 𝑓𝑖 (𝑥 ∗∗ ) ≤ 𝑓𝑖 (𝑥 ∗ ) for all 𝑖 = 1,2, … . 𝑙 and 𝑓𝑖 (𝑥 ∗∗ ) < 𝑓𝑖 (𝑥 ∗ ) for at least
one 𝑖 = 1,2, … . 𝑙. (i)
𝑈𝑘 −𝑓𝑘 (𝑥 ∗∗ ) 𝑈𝑘 −𝑓𝑘 (𝑥 ∗ ) 𝑈𝑘 −𝑓𝑘 (𝑥 ∗∗ ) 𝑈𝑘 −𝑓𝑘 (𝑥 ∗ )
So, ≥ for all 𝑘 = 1,2 … . . 𝑙 and > for at least one
𝑈𝑘 −𝐿𝑘 𝑈𝑘 −𝐿𝑘 𝑈𝑘 −𝐿𝑘 𝑈𝑘 −𝐿𝑘
𝑘 = 1,2 … . . 𝑙
𝑈𝑘 −𝑓𝑘 (𝑥 ∗∗ ) 𝑈𝑘 −𝑓𝑘 (𝑥 ∗ ) 𝑈𝑘 −𝑓𝑘 (𝑥 ∗∗ ) 𝑈𝑘 −𝑓𝑘 (𝑥 ∗ )
Now, 𝛿𝑖 ≥ 𝛿𝑖 for all 𝑘 = 1,2 … . . 𝑝 and 𝛿𝑖 > 𝛿𝑖 for
𝑈𝑘 −𝐿𝑘 𝑈𝑘 −𝐿𝑘 𝑈𝑘 −𝐿𝑘 𝑈𝑘 −𝐿𝑘
at least one 𝑘 = 1,2 … . . 𝑙 and for all 𝑖 = 1,2, … . , 𝑛.
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 211

𝑈𝑘 −𝑓𝑘 (𝑥 ∗∗ ) 𝑈𝑘 −𝑓𝑘 (𝑥 ∗ )
That implies, 𝑀𝑎𝑥𝑘,𝑖 𝛿𝑖 ≥ 𝑀𝑎𝑥𝑘,𝑖 𝛿𝑖 for all 𝑘 = 1,2 … . . 𝑙 and
𝑈𝑘 −𝐿𝑘 𝑈𝑘 −𝐿𝑘
𝑈𝑘 −𝑓𝑘 (𝑥 ∗∗ ) 𝑈 −𝑓 (𝑥 ∗ )
𝑀𝑎𝑥𝑘,𝑖 𝛿𝑖 > 𝑀𝑎𝑥𝑘,𝑖 𝛿𝑖 𝑘 𝑘 for at least one 𝑘 = 1,2 … . . 𝑙 and for all 𝑖 =
𝑈𝑘 −𝐿𝑘 𝑈 −𝐿 𝑘 𝑘
1,2, … . 𝑛.
𝑈 −𝑓 (𝑥 ∗∗ ) ∗∗ 𝑈 −𝑓 (𝑥 ∗ ) ∗
Consider, 𝑀𝑎𝑥𝑘,𝑖 𝛿𝑖 𝑘 𝑘 = 𝛼𝑘𝑖 and 𝑀𝑎𝑥𝑘,𝑖 𝛿𝑖 𝑘 𝑘 = 𝛼𝑘𝑖
𝑈𝑘 −𝐿𝑘 𝑈𝑘 −𝐿𝑘
∗∗ ∗
𝛼𝑘𝑖 ≥ 𝛼𝑘𝑖 for all 𝑘 = 1,2 … . . 𝑙 and for all 𝑖 = 1,2, … 𝑛, the inequality occurs for at least
one 𝑘and 𝑖.
𝑛 𝑛
∗∗ 2 𝜔𝑖 ∗∗ 2 𝜔𝑖
∏(1 − 𝛼𝑘𝑖 ) < ∏(1 − 𝛼𝑘𝑖 )
𝑖=1 𝑖=1

∗∗ 2 𝜔𝑖 ∗ 2 𝜔𝑖
⇒ √1 − ∏𝑛𝑖=1(1 − 𝛼𝑘𝑖 ) > √1 − ∏𝑛𝑖=1(1 − 𝛼𝑘𝑖 ) (ii)
𝑓𝑘 (x∗∗ )−𝐿𝑘 𝑓𝑘 (x∗ )−𝐿𝑘 𝑓𝑘 (x∗∗ )−𝐿𝑘
Similarly, from Eq. (i) 𝛾𝑖 ≤ 𝛾𝑖 for all 𝑘 = 1,2 … . . 𝑙 and 𝛾𝑖 <
𝑈𝑘 +α−𝐿𝑘 𝑈𝑘 +α−𝐿𝑘 𝑈𝑘 +α−𝐿𝑘
𝑓𝑘 (x∗ )−𝐿𝑘
𝛾𝑖 for at least one 𝑘and 𝑖.
𝑈𝑘 +α−𝐿𝑘
𝑓𝑘 (x∗∗ )−𝐿𝑘 𝑓𝑘 (x∗ )−𝐿𝑘
That implies, 𝑀𝑖𝑛𝑘,𝑖 𝛾𝑖 ≤ 𝑀𝑖𝑛𝑘,𝑖 𝛾𝑖 for all 𝑘 and 𝑖 and
𝑈𝑘 +α−𝐿𝑘 𝑈𝑘 +α−𝐿𝑘
𝑓𝑘 (x∗∗ )−𝐿𝑘 𝑓 (x∗ )−𝐿𝑘
𝑀𝑖𝑛𝑘,𝑖 𝛾𝑖 < 𝑀𝑖𝑛𝑘,𝑖 𝛾𝑖 𝑘 for at least one 𝑘and 𝑖.
𝑈𝑘 +α−𝐿𝑘 𝑈 +α−𝐿 𝑘 𝑘
𝑓𝑘 (x∗∗ )−𝐿𝑘 ∗∗ 𝑓𝑘 (x∗ )−𝐿𝑘 ∗
Consider, 𝑀𝑖𝑛𝑘,𝑖 𝛾𝑖 = 𝛽𝑘𝑖 and 𝑀𝑖𝑛𝑘,𝑖 𝛾𝑖 = 𝛽𝑘𝑖
𝑈𝑘 +α−𝐿𝑘 𝑈𝑘 +α−𝐿𝑘
∗∗ ∗ ∗∗ ∗
So, 𝛽𝑘𝑖 ≤ 𝛽𝑘𝑖 for all 𝑘and 𝑖. and 𝛽𝑘𝑖 < 𝛽𝑘𝑖 for at least one 𝑘 and 𝑖.
∗∗ 𝜔𝑖 ∗ 𝜔𝑖
∏𝑛𝑖=1 𝛽𝑘𝑖 < ∏𝑛𝑖=1 𝛽𝑘𝑖 (iii)

∗∗ 2 𝜔𝑖 𝜔𝑖 ∗ 2 𝜔𝑖
Eqs. (ii) and (iii) ⇒ √1 − ∏𝑛𝑖=1(1 − 𝛼𝑘𝑖 ) − ∏𝑛𝑖=1 𝛽𝑘𝑖
∗∗
> √1 − ∏𝑛𝑖=1(1 − 𝛼𝑘𝑖 ) −
∗ 𝜔𝑖
∏𝑛𝑖=1 𝛽𝑘𝑖

𝑙 𝑛 𝑛
∗∗ 2 𝜔𝑖 ∗∗ 𝜔𝑖
⇒ ∑ 𝑊𝑘 √1 − ∏(1 − 𝛼𝑘𝑖 ) − ∏ 𝛽𝑘𝑖
𝑘=1 𝑖=1 𝑖=1
[ ]
𝑙 𝑛 𝑛
∗ 2 𝜔𝑖 ∗ 𝜔𝑖
> ∑ 𝑊𝑘 √1 − ∏(1 − 𝛼𝑘𝑖 ) − ∏ 𝛽𝑘𝑖
𝑘=1 𝑖=1 𝑖=1
[ ]
∗∗ ∗∗ ) ∗ ∗ ), (𝑥 ∗ ∗ ∗)
Hence, 𝑓(𝛼 , 𝛽 > 𝑓(𝛼 , 𝛽 which contradict , 𝛼 , 𝛽 is unique optimal solution
of Problem (3). So 𝑥 ∗ is a pareto optimal solution of Problem (1).
(For PHFWG operator) Similar as above.
Theorem2: For non-zero weight vector, a pareto optimal solution (𝑥 ∗ ) of Problem (1) is
an efficient solution of problem (3) if 𝑓𝑖 are convex.
Proof: 𝑥 ∗ is a pareto optimal solution of Problem 1.
Suppose, 𝑥 ∗ is not an efficient solution of Problem 3. Then there exists 𝑥 ∗∗ in feasible space
such that 𝑓(𝛼 ∗ , 𝛽 ∗ ) < 𝑓(𝛼 ∗∗ , 𝛽 ∗∗ )
212 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

𝑙 𝑛 𝑛
∗2 )𝜔𝑖 ∗𝜔
⇒ ∑ 𝑊𝑘 (√1 − ∏ (1 − 𝛼𝑘𝑖 − ∏ 𝛽𝑘𝑖 𝑖 )
𝑘=1 𝑖=1 𝑖=1

𝑙 𝑛 𝑛
∗∗2 )𝜔𝑖 ∗∗𝜔
< ∑ 𝑊𝑘 (√1 − ∏ (1 − 𝛼𝑘𝑖 − ∏ 𝛽𝑘𝑖 𝑖 )
𝑘=1 𝑖=1 𝑖=1

𝑈𝑘 −𝑓𝑘 (𝑥 ∗ ) ∗ 𝑓𝑘 (x∗ )−𝐿𝑘 ∗ 𝑈𝑘 −𝑓𝑘 (𝑥 ∗∗ ) ∗ 𝑓𝑘 (x∗∗ )−𝐿𝑘

Since, 𝛿𝑖 = 𝛼𝑘𝑖 , 𝛾𝑖 = 𝛽𝑘𝑖 𝑎𝑛𝑑 𝛿𝑖 = 𝛼𝑘𝑖 𝑎𝑛𝑑 𝛾𝑖 =
𝑈𝑘 −𝐿𝑘 (𝑈𝑘 +α)−𝐿𝑘 𝑈𝑘 −𝐿𝑘 (𝑈𝑘 +α)−𝐿𝑘
∗
𝛽𝑘𝑖
If 𝑓𝑖 are convex, then for all 𝑘 = 1,2, … 𝑙

𝑛 𝑛 𝑛 𝑛
∗2 )𝜔𝑖 ∗𝜔
∗∗2 )𝜔𝑖 ∗∗𝜔𝑖
(√1 − ∏ (1 − 𝛼𝑘𝑖 − ∏ 𝛽𝑘𝑖 𝑖 ) < √1 − ∏ (1 − 𝛼𝑘𝑖 − ∏ 𝛽𝑘𝑖
𝑖=1 𝑖=1 𝑖=1 𝑖=1

Taking, 𝜔1 = 1, 𝜔2 = 𝜔3 = ⋯ . = 𝜔𝑛 = 0

∗2 ) ∗ ∗∗2 ) ∗∗
√1 − (1 − 𝛼𝑘1 − 𝛽𝑘1 < √1 − (1 − 𝛼𝑘1 − 𝛽𝑘1
∗ ∗ ∗∗ ∗∗
⇒ 𝛼𝑘1 − 𝛽𝑘1 < 𝛼𝑘1 − 𝛽𝑘1
𝑈𝑘 − 𝑓𝑘 (𝑥 ∗ ) 𝑓𝑘 (x ∗ ) − 𝐿𝑘 𝑈𝑘 − 𝑓𝑘 (𝑥 ∗∗ ) 𝑓𝑘 (x ∗∗ ) − 𝐿𝑘
⇒ 𝛿1 − 𝛾1 < 𝛿1 − 𝛾1
𝑈𝑘 − 𝐿𝑘 (𝑈𝑘 + α) − 𝐿𝑘 𝑈𝑘 − 𝐿𝑘 (𝑈𝑘 + α) − 𝐿𝑘
⇒ −(𝛿1 + 𝛾1 )𝑓𝑘 (𝑥 ∗ ) < −(𝛿1 + 𝛾1 )𝑓𝑘 (𝑥 ∗∗ )
⇒ 𝑓𝑘 (𝑥 ∗∗ ) < 𝑓𝑘 (𝑥 ∗ ) 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑘 = 1,2, … . . 𝑙.
Which contradict that 𝑥 ∗ is pareto optimal solution of Problem (1)
Hence, 𝑥 ∗ is efficient solution of Problem (3)
The proposed method to solve MOOP is classified by two categories by two PHFAOs,
which are PHFWA, PHFWG operators. The following is the formulation of MOOP using
PHFAOs (weighted averaging and weighted geometric).

Model 1: by PHFWA operator

Max PHFWA(𝜇 𝐸1 , 𝜇 𝐸2 , … , 𝜇 𝐸𝑛 ; 𝜗𝐸1 , 𝜗 𝐸2 , … , 𝜗𝐸𝑛 )

𝑙 𝑛 𝑛
2 )𝜔𝑖 𝜔
= Max ∑ 𝑊𝑘 (√1 − ∏ (1 − 𝛼𝑘𝑖 − ∏ 𝛽𝑘𝑖𝑖 ).
𝑘=1 𝑖=1 𝑖=1

Subject to the constraints in Problem (3).

 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 213

Model 2: by PHFWG operator

Max PHFWG(𝜇 𝐸1 , 𝜇 𝐸2 , … , 𝜇 𝐸𝑛 ; 𝜗𝐸1 , 𝜗 𝐸2 , … , 𝜗𝐸𝑛 )

𝑙 𝑛 𝑛
𝜔 2 )𝜔𝑖
= Max ∑ 𝑊𝑘 (∏ 𝛼𝑘𝑖𝑖 − √1 − ∏ (1 − 𝛽𝑘𝑖 ).
𝑘=1 𝑖=1 𝑖=1

Subject to the constraints in Problem (3).

4. APPLICATION OF THE METHOD TO MULTI OBJECTIVE

OPTIMIZATION PROBLEM
A) Manufacturing system:
We have taken an example of multi objective optimization problem of manufacturing
system described in Singh and Yadav [43]. A manufacturing factory is planning to produce
three types of products: A, B, and C over a period of one month. To produce each product,
the factory requires three types of resources: R1, R2, and R3.
There are around 30, 20 and 20 units of tolerance is allowed for resource R1, R2, R3
respectively by manager. Unit cost and sale price of the product are 𝑐𝐴 , 𝑐𝐵 , 𝑐𝐶 and 𝑆𝐴 =
𝑠𝐴 𝑠𝐵 𝑠𝐶
1 , 𝑆𝐵 = 1 , 𝑆𝐶 = 1 respectively. Where 𝑎1 , 𝑎2 , 𝑎3 are positive real number. Company
𝑥1𝑎1 𝑥2𝑎2 𝑥1𝑎3
wants to maximize the profit and minimize the time required for production.

A B C
R1 2 4 3 ≤ 325
R2 4 2 2 ≤ 360
R3 3 2 3 ≥ 360
Time(h) 4 5 6

There are around 30, 20 and 20 units of tolerance is allowed for resource R1, R2, R3
respectively by manager. Unit cost and sale price of the product are 𝑐𝐴 , 𝑐𝐵 , 𝑐𝐶 and 𝑆𝐴 =
𝑠𝐴 𝑠𝐵 𝑠𝐶
1 , 𝑆𝐵 = 1 , 𝑆𝐶 = 1 respectively. Where 𝑎1 , 𝑎2 , 𝑎3 are positive real number. Company
𝑥1𝑎1 𝑥2𝑎2 𝑥1𝑎3
wants to maximize the profit and minimize the time required for production.
To handling the uncertainty, all parameters are taken as PPFN.
̃ = (95,100,103; 95, 100, 104), 𝑅(𝑠𝐴 ) = 99.4
𝑠𝐴 = 100
̃ = (118, 120,122; 115,120,123), 𝑅(𝑠𝐵 ) = 119.6
𝑠𝐵 = 120
̃ = (94, 95,96; 93,95,97), 𝑅(𝑠𝐶 ) = 95
𝑠𝐶 = 95
̃ = (7.5,7.5, 8.5; 7,7.5,9), 𝑅(𝑐𝐴 ) = 7.9
𝑐𝐴 = 7.5
̃ = (9,10,10.5; 9,10,11), 𝑅(𝑐𝐵 ) = 9.9
𝑐𝐵 = 10
𝑐𝐶 = 8̃ = (7.5,8,8.5; 7.5,8,9), 𝑅(𝑐𝐶 ) = 8.1
̃ = (325,325,340; 325,325,345), 𝑅(325
325 ̃ ) = 332
214 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

̃ = (360, 360,375; 360,360,380), 𝑅(360

360 ̃ ) = 367
̃ = (350, 365, 365; 345,365,365), 𝑅(365
365 ̃ ) = 358

2̃ = (1.5, 2,2.3; 1.2,2,2.5), 𝑅(2̃) = 1.9

3̃ = (2.5,3,3.4; 2.3,3,3.7), 𝑅(3̃) = 2.98
4̃ = (3.5,4,4.3; 3.2,4,4.5), 𝑅( 4̃) = 3.9
5̃ = (4.8, 5,5.6; 4.5, 5,5.8), 𝑅(5̃) = 5.14
6̃ = (5.8,6,6.2; 5.5,6,6.3), 𝑅( 6̃) = 5.96
3
𝑎1 = 𝑎2 = 2, 𝑎3 =
2
The optimization problem can be written as
̃ 𝑥1 1−1/𝑎1 ̃ 𝑥1 + 120
̃ 𝑥1 1−1/𝑎2 ̃𝑥
̃ 𝑥2 + 95 1−1/𝑎1
𝑚𝑎𝑥𝑖𝑚𝑖𝑧𝑒 𝑓1 = 100 − 7.5 − 10 1 − 8̃𝑥3
𝑚𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓2 = 4̃𝑥1 + 5̃𝑥2 + 6̃𝑥3
𝑆𝑢𝑏𝑗𝑒𝑐𝑡 𝑡𝑜
̃
2̃𝑥1 + 4̃𝑥2 + 3̃𝑥3 ≤ 325
4̃𝑥1 + 2̃𝑥2 + 2̃𝑥3 ≤ 360
̃
̃
3̃𝑥1 + 2̃𝑥2 + 3̃𝑥3 ≥ 365
𝑥1 , 𝑥2 , 𝑥3 ≥ 0.
The equivalent crisp multi objective optimization problem is
1 1 1
𝑚𝑎𝑥𝑖𝑚𝑖𝑧𝑒 𝑓1 = 99.4𝑥12 − 7.9𝑥1 + 119.6𝑥22 − 9.9𝑥2 + 95𝑥33 − 8.1𝑥3
𝑚𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓2 = 3.9𝑥1 + 5.14𝑥2 + 5.96𝑥3
𝑆𝑢𝑏𝑗𝑒𝑐𝑡 𝑡𝑜
1.9𝑥1 + 3.9𝑥2 + 2.98𝑥3 ≤ 332
3.9𝑥1 + 1.9𝑥2 + 1.9𝑥3 ≤ 367
2.98𝑥1 + 1.9𝑥2 + 2.98𝑥3 ≥ 358
𝑥1 , 𝑥2 , 𝑥3 ≥ 0.
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 215

Pay-off matrix is:

𝑋1 𝑋2

𝑓1 606.4352 220.4427

𝑓2 615.3611 573.0927

So, 𝑈1 = 606.4352, 𝐿1 = 220.4427, 𝑈2 = 615.3611, 𝐿2 = 573.0927

Here three experts assign their hesitant values 𝛿1 = .96, 𝛿2 = .98, 𝛿3 = 1 for membership
function of objective functions and hesitant values 𝛾1 = .96, 𝛾2 = .98, and 𝛾3 = 1 for non-
1
membership function. Tolerance 𝛼𝑓1 = 50 and 𝛼𝑓2 = 50. Take 𝜔1 = 𝜔2 = 𝜔3 = , Then
3
we can formulate the optimization problem in PHF environment as follows:

Using PHFWA operator:

Max PHFWA(𝜇 𝐸1 , 𝜇 𝐸2 , 𝜇 𝐸3 ; 𝜗𝐸1 , 𝜗𝐸2 , 𝜗𝐸3 )

2 3 3 1
1
2 )3 3
= Max ∑ 𝑊𝑘 √1 − ∏ (1 − 𝛼𝑘𝑖 − ∏ 𝛽𝑘𝑖 .
𝑘=1 𝑖=1 𝑖=1
( )
Subject to,
𝑓1 − 220.4427 606.4352 − 𝑓1
0.96 ≥ 𝛼11 ; 0.96 ≤ 𝛽11
606.4352 − 220.4427 606.4352 − 170.4427
𝑓1 − 220.4427 606.4352 − 𝑓1
0.98 ≥ 𝛼12 ; 0.98 ≤ 𝛽12
606.4352 − 220.4427 606.4352 − 170.4427
𝑓1 − 220.4427 606.4352 − 𝑓1
≥ 𝛼13 ; ≤ 𝛽13
606.4352 − 220.4427 606.4352 − 170.4427
615.3611 − 𝑓2 𝑓2 − 573.0927
0.96 ≥ 𝛼21 ; 0.96 ≤ 𝛽21
615.3611 − 573.0927 665.3611 − 573.0927
615.3611 − 𝑓2 𝑓2 − 573.0927
0.98 ≥ 𝛼22 ; 0.98 ≤ 𝛽22
615.3611 − 573.0927 665.3611 − 573.0927
615.3611 − 𝑓2 𝑓2 − 573.0927
≥ 𝛼23 ; ≤ 𝛽23
615.3611 − 573.0927 665.3611 − 573.0927
2 2 2 2
Where 0 ≤ 𝛼𝑘𝑟 + 𝛽𝑘𝑠 ≤ 1 𝛼𝑘𝑟 ≥ 𝛽𝑘𝑠 ∀𝑘 = 1,2 and 𝑟, 𝑠 ∈ {1,2,3}

1.9𝑥1 + 3.9𝑥2 + 2.98𝑥3 ≤ 332

3.9𝑥1 + 1.9𝑥2 + 1.9𝑥3 ≤ 367
2.98𝑥1 + 1.9𝑥2 + 2.98𝑥3 ≥ 358
216 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

𝑥1 , 𝑥2 , 𝑥3 ≥ 0.
𝑊1 + 𝑊2 = 1. (4)
Using PHFWG operator:

Max PHFWG(𝜇 𝐸1 , 𝜇 𝐸2 , 𝜇 𝐸3 ; 𝜗𝐸1 , 𝜗𝐸2 , 𝜗𝐸3 )

2 3 1 3
1
3 2 )3
= Max ∑ 𝑊𝑘 ∏ 𝛼𝑘𝑖 − √1 − ∏ (1 − 𝛽𝑘𝑖 .
𝑘=1 𝑖=1 𝑖=1
( )
Subject to, the same constraints as Problem (4). (5)
B) Life support system (LSS):

Figure 3: Life support system

The Figure 3 presents the Life support system, here the system cost (𝐶𝑠 ),system reliability
(𝑅𝑆 ) are presented by:
𝑅𝑆 = 1 − 𝑟3 [(1 − 𝑟1 )(1 − 𝑟4 )]2 − (1 − 𝑟3 )[1 − 𝑟2 {1 − (1 − 𝑟1 )(1 − 𝑟4 )}]2
𝑎 𝑎 𝑎 𝑎
𝐶𝑆 = 2𝐾1 . 𝑟1 1 + 2𝐾2 . 𝑟2 2 + 𝐾3 . 𝑟3 3 + 2𝐾4 . 𝑟4 4
The goal of this problem is to reduce system cost while maintaining component and
system reliability and taking cost coefficient as PPFN. Therefore, the problem is described
as:
Maximize 𝑅𝑆 , Minimize 𝐶̃𝑠
Subject to
. 5 ≤ 𝑟𝑖 ≤ 1 − 10−6 𝑖 = 1,2,3,4 (6)
Where, 𝑟𝑖 is reliability of 𝑖 𝑡ℎ component.
Here 𝐾1 = 100, 𝐾2 = 100, 𝐾3 = 200, 𝐾4 = 150 𝑎𝑛𝑑 𝑎𝑖 = 0.6 𝑓𝑜𝑟 𝑖 = 1,2,3,4.
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 217

Here cost coefficients 𝐾1 , 𝐾2 , 𝐾3 , 𝐾4 are taken as parabolic fuzzy number and the crisp
value is calculated in Table 1.

Table 1: Cost coefficient as parabolic Pythagorean fuzzy number

Parameter Parabolic Pythagorean fuzzy number Crisp value
{(𝑎, 𝑏, 𝑐), (𝑎′ , 𝑏, 𝑐 ′ )} (2𝑎 + 𝑏 + 2𝑐) + (2𝑎′ + 𝑏 + 2𝑐 ′ )
[ ]
10
𝐾1 {(98,100,102), (96,100,103)} 99.8
𝐾2 {(99,100,102), (98,100,103)} 100.4
𝐾3 {(198,200,201), (197,200,202)} 199.6
𝐾4 {(147,150,152), (147,150,153)} 149.8
Pay-off matrix is:

𝑅1 𝑅2

𝑅𝑠 0.999999 0.7734375

𝐶𝑠 1099.199 725.2015

So, 𝑈1 = .999999, 𝐿1 = 0.773437, 𝑈2 = 1099.199, 𝐿2 = 725.2015

Here three experts assign their hesitant values 𝛿1 = .96, 𝛿2 = .98, 𝛿3 = 1 for membership
function of objective functions and hesitant values 𝛾1 = .96, 𝛾2 = .98, and 𝛾3 = 1 for non-
1
membership function. Tolerance 𝛼𝑅𝑠 = .1 and 𝛼𝐶𝑠 = 50. Take 𝜔1 = 𝜔2 = 𝜔3 = , Then
3
we can formulate the optimization problem in PHF environment as follows:

Using PHFWA operator:

Max PHFWA(𝜇 𝐸1 , 𝜇 𝐸2 , 𝜇 𝐸3 ; 𝜗𝐸1 , 𝜗𝐸2 , 𝜗 𝐸3 )

2 3 3 1
1
2 )3 3
= Max ∑ 𝑊𝑘 √1 − ∏ (1 − 𝛼𝑘𝑖 − ∏ 𝛽𝑘𝑖 .
𝑘=1 𝑖=1 𝑖=1
( )

Subject to,

R s (r) − 0.7734375 0.999999 − R s (r)

0.96 ≥ 𝛼11 ; 0.96 ≤ 𝛽11
0.999999 − .7734375 0.999999 − 0.6734375
R s (r) − 0.7734375 0.999999 − R s (r)
0.98 ≥ 𝛼12 ; 0.98 ≤ 𝛽12
0.999999 − .7734375 0.999999 − 0.6734375
R s (r) − 0.7734375 0.999999 − R s (r)
≥ 𝛼13 ; ≤ 𝛽13
0.999999 − .7734375 0.999999 − 0.6734375
218 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

1099.199 − 𝐶𝑠 (𝑟) 𝐶𝑠 (𝑟) − 725.2015

0.96 ≥ 𝛼21 ; 0.96 ≤ 𝛽21
1099.199 − 725.2015 1149.199 − 725.2015
1099.199 − 𝐶𝑠 (𝑟) 𝐶𝑠 (𝑟) − 725.2015
0.98 ≥ 𝛼22 ; 0.98 ≤ 𝛽22
1099.199 − 725.2015 1.199 − 725.2015
1099.199 − 𝐶𝑠 (𝑟) 𝐶𝑠 (𝑟) − 725.2015
≥ 𝛼23 ; ≤ 𝛽23
1099.199 − 725.2015 1099.199 − 725.2015
2 2 2 2
Where 0 ≤ 𝛼𝑘𝑟 + 𝛽𝑘𝑠 ≤ 1 𝛼𝑘𝑟 ≥ 𝛽𝑘𝑠 ∀𝑘 = 1,2 and 𝑟, 𝑠 ∈ {1,2,3}
0.5 ≤ 𝑟𝑗 ≤ 1 − 10−6 𝑗 = 1,2,3,4.
𝑊1 + 𝑊2 = 1. (7)
Using PHFWG operator:

Max PHFWG(𝜇 𝐸1 , 𝜇 𝐸2 , 𝜇 𝐸3 ; 𝜗𝐸1 , 𝜗𝐸2 , 𝜗 𝐸3 )

2 3 1 3
1
3 2 )3
= Max ∑ 𝑊𝑘 ∏ 𝛼𝑘𝑖 − √1 − ∏ (1 − 𝛽𝑘𝑖 .
𝑘=1 𝑖=1 𝑖=1
( )
Subject to, the same constraints as Problem (7). (8)
Degree of closeness to Ideal solution:
A pivotal factor in ascertaining the compromise solution to demonstrate the
effectiveness of various solution strategies is how near the solutions are to the ideal ones.
Multi-objective optimization problem solution a variety of distance function families exist,
and these families may be used to evaluate the effectiveness of different solution methods.
In this section, we have considered a family of distance functions that may be described as
follows:
1
𝑙 𝑒
𝐷𝑒 (𝛾, 𝑙) = [∑ 𝛾𝑘𝑒 (1 − 𝑟𝑘 )𝑒 ]
𝑘=1

where 𝑟𝑘 is the fraction of compromise solution to the ideal solution of 𝑘 𝑡ℎ objective and
𝛾 = (𝛾1 , 𝛾2 , … , 𝛾𝑙 ) is the weight vector assigned to objectives.
𝑙

𝐷1 (𝛾, 𝑙) = 1 − ∑ 𝛾𝑘 𝑟𝑘 ,
𝑘=1
1
𝑙 2
𝐷2 (𝛾, 𝑙) = [∑ 𝛾𝑘2 (1 − 𝑟𝑘 )2 ]
𝑘=1

𝐷∞ (𝛾, 𝑙) = max [𝛾𝑘 (1 − 𝑟𝑘 )]

𝑘

In minimization type objective function,

 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 219

the ideal solution of 𝑓𝑘

𝑟𝑘 =
the desired compromise solution of 𝑓𝑘
In maximization type objective function,
the desired compromise solution of 𝑓𝑘
𝑟𝑘 =
the ideal solution of 𝑓𝑘

5. RESULTS AND DISCUSSION

The compromise solution of manufacturing system and life support system has been
achieved through the utilization of PHF linear membership functions and non-membership
functions under two distinct models, as illustrated in Table 2 and Table 3, respectively. In
each model, different crisp weights have been employed. The weights assigned to the first
objective have been selected randomly within the range of 0 to 1. These weights are
determined by the decision-maker(s) based on their level of satisfaction. Simultaneously,
the weight assigned to the second objective is the complement of weight to the first
objective function. The weight of PHF element in aggregation operator are taken
maintaining same weightage. Then the single objective optimization models are solved by
LINGO 18.0. To assess the sensitivity of the results, a thorough sensitivity analysis was
performed by modifying the weight parameter. To show the efficiency of the solution
degree of closeness to ideal solution are calculated and compared with solution obtained
by the existing method using HF weighted averaging (HFWA) and HF weighted geometric
(HFWG) operators in Table 5 and Table 6.
A. Compromise solutions of manufacturing system

In Table 2, by HFWA operator, for weight (.1, .9) to the objective function the optimal
solution is (220.7964, 573.0927), for weight (.3, .7) the solution is
(347.7069, 575.7889), for weight (.5, .5) the solution is (453.2292, 583.6776), for
weight (.7, .3) solution is (567.2987, 603.2049) and for (.9, .1) is
(606.4351, 615.3611).
So, when the weight of the first objective is increased then the solution goes from worst
to best solution and when the weight of second objective is decreased the solution goes
from best to the worst solution.
By PHFWA operator, for weight (.1, .9) to the objective function the optimal solution
is (292.2183, 573.9100), for weight (.3, .7) the solution is (427.8848, 581.1482), for
weight (.5, .5) the solution is (543.0254, 597.5321), for weight (.7, .3) solution is
(606.4313, 615.3597) and for (.9, .1) is (606.4332, 615.3604).
So, when the weight allocated to the first objective is increased, the solution transitions
from a relatively poorer solution to a better one. Conversely, when the weight allocated to
the second objective is decreased, the solution moves from a better solution to a relatively
poorer one.
The result obtained by using PHFWA operator is different from the result obtained by
using HFWAO, the comparison is presented by degree of closeness to ideal solution in
Table 5.
By HFWG operator, for weight (.1, .9) to the objective function the optimal solution is
(260.2183, 573.3294), for weight (.3, .7) the solution is (349.8160, 575.8869), for
220 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

weight (.5, .5) the solution is (453.3863, 583.6948), for weight (.7, .3) solution is
(566.1693, 602.9149) and for (.9, .1) is (606.4351, 615.3611).
So, when the weight allocated to the first objective is increased, the solution progresses
from its worst state to the best possible solution. Conversely, when the weight allocated to
the second objective is reduced, the solution regresses from its best state to the worst
possible solution.

Table 2: Optimal compromise objective values for manufacturing system

weight Objective values [(𝑓1∗ , 𝑓2∗ )]
(𝑊1 , 𝑊2 ) HFWA PHFWA HFWG PHFWG
(.1, .9) (220.7964, 573.0927) (292.9689, 573.9100) (260.2183, 573.3294) (270.9738, 573.4794)
(.2, .8) (296.2601, 573.9894) (362.3822, 576.5138) (303.2744, 574.1721) (323.4829, 574.8055)
(.3, .7) (347.7069, 575.7889) (427.8848, 581.1482) (349.8160, 575.8869) (377.7452, 577.3857)
(.4, .6) (399.2316, 578.8155) (488.4911, 588.0708) (399.9206, 578.8656) (433.2288, 581.6428)
(.5, .5) (453.2292, 583.6776) (543.0254, 597. 5321) (453.3863, 583.6948) (488.9955, 588.1421)
(.6, .4) (509.6103, 591.2903) (590.1488, 609.7415) (509.4388, 591.2621) (543.4850, 597.6295)
(.7, .3) (567.2987, 603.2049) (606.4313, 615.3597) (566.1693, 602.9149) (606.4351, 615.3610)
(.8, .2) (606.4351, 615.3611) (606.4322, 615.3600) (606.4351, 615.3611) (606.4351, 615.3610)
(.9, .1) (606.4351, 615.3611) (606.4332, 615.3604) (606.4351, 615.3611) (606.4351, 615.3610)

By PHFWG operator, for weight (.1, .9) to the objective function the optimal solution
is (270.9738, 573.4794), for weight (.3, .7) the solution is (377.7452, 577.3857), for
weight (.5, .5) the solution is (488.9955, 588.1421), for weight (.7, .3) solution is
(606.4351, 615.3610) and for (.9, .1) is (606.4351, 615.3610).
It is observed that as the weights assigned to the first objective function are increased,
the solution values progressively move closer to their best possible outcome. Similarly, As
the weights assigned to the second objective function decrease, the solution values tend to
converge toward their least favorable state.
The results obtained by PHFWG and HFWG operators are different; in Table 5, the
comparison is shown by how closely the results match the ideal solution.

B. Compromise solutions of reliability optimization model of Life support system

In Table 3, by HFWA operator, at weight (.1, .9) assigned to the objective functions
the optimal solution is (.773437, 725.2015), for weight (.3, .7) the solution is
(. 955693, 805.6241), at weight (.5, .5) the solution is (.999999, 861.3603), for weight
(.7, .3) solution is (. 999999, 861.3603) and for (.9, .1) is (.999999, 861.3603).
In summary, it can be concluded that increasing the weight of the first objective brings
the solution closer to the best outcome, while decreasing the weight of the second objective
moves the solution towards the worst possible result and vice versa.
By PHFWA operator, for weight (.1, .9) to the objective function the optimal solution
is (.868483, 759.8243), for weight (.3, .7) the solution is (. 938986, 794.4539), for
weight (.5, .5) the solution is (.992998, 840.6600), for weight (.7, .3) solution is
(. 999999, 861.3583) and for (.9, .1) is (.999999, 861.3583).
It is observed that altering the weight assigned to the first (second) objective leads the
solution closer to the best (worst) possible outcome.
By HFWG operator, for weight (.1, .9) to the objective function the optimal solution is
(.773440, 725.2023), for weight (.3, .7) the solution is (. 955632, 805.5805), for weight
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 221

(.5, .5) the solution is (.993971,842.2107), for weight (.7, .3) solution is
(. 998981, 853.6873), and for (.9, .1) is (.999934, 859.4861).
The results obtained by PHFWA and HFWA operators are different from one other,
and Table 6 compares them according to how closely they get to the ideal optimal solution.
So, when the weight allocated to the first objective is increased, the solution progresses
from its worst state to the best possible solution. Conversely, when the weight allocated to
the second objective is reduced, the solution regresses from its best state to the worst
possible solution.
By PHFWG operator, for weight (.1, .9) to the objective function the optimal solution
is (.868482, 759.8240), for weight (.3, .7) the solution is (. 939301,794.6547), for
weight (.5, .5) the solution is (.992432, 839.8040), for weight (.7, .3) solution is
(. 998735, 852.7920) and for (.9, .1) is (.999919, 859.2686).
The result produced by PHFWG operator differs from the result produced by HFWG
operator; the comparison is shown in Table 6 by the degree of similarity to the ideal
solution.
It is seen that the solution values gradually approach their ideal conclusion when the
weights applied to the first goal function are increased. Similar to this, when the weight
allocated to second objective is reduced, the solution values go to worse.

Table 3: Optimal compromise objective values for Life support system (LSS)
weight Objective values [(𝑓1∗ , 𝑓2∗ )]

(𝑊1 , 𝑊2 ) HFWA PHFWA HFWG PHFWG

(.1, .9) (.773437, 725.2015) (.868483, 759.8243) (.773440, 725.2023) (.868482, 759.8240)
(.2, .8) (.773437, 725.2015) (.868483, 759.8243) (.873537, 761.8927) (.868482, 759.8240)
(.3, .7) (.955693, 805.6241) (.938986,794.4539) (.955632, 805.5805) (.939301, 794.6547)
(.4, .6) (.985855, 831.4092) (.981391, 826.6951) (.985179, 830.6577) (.981096, 826.4021)
(.5, .5) (.999999, 861.3603) (.992998, 840.6600) (.993971, 842.2107) (.992432, 839.8040)
(.6, .4) (.999999, 861.3603) (.997634, 849.5475) (.997451, 849.0868) (.996825, 847.6161)
(.7, .3) (.999999, 861.3603) (.999999, 861.3583) (.998981, 853.6873) (.998735, 852.7920)
(.8, .2) (.999999, 861.3603) (.999999, 861.3583) (.999661, 856.9925) (.998581, 856.4895)
(.9, .1) (.999999, 861.3603) (.999999, 861.3583) (.999934, 859.4861) (.999919, 859.2686)

To summarize, increasing the weight of an objective enhances its influence, enabling

the solution to reach its optimal performance in that objective. Conversely, decreasing the
weight of an objective weakens its impact, potentially leading to a deterioration in
performance in that aspect while aiming for a better balance with other objectives. This
approach allows decision-makers to control the trade-offs between objectives and tailor the
solution to their preferences and priorities.
Using classical weighted sum method, the solution obtained for manufacturing system
and life support system are listed in Table 4. In this context, the solution falls within the
same coverage spectrum as the solution obtained through the proposed method.
222 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

Table 4: Optimal value of the objective by classical weighted sum method

weight Objective values [(𝑓1∗ , 𝑓2∗ )] by weighted sum method

(𝑊1 , 𝑊2 ) Manufacturing system Life support system

(.1, .9) (455.3876, 5839156) (.773437, 725.2015)
(.2, .8) (563.5913,602.2633) (.773437, 725.2015)
(.3, .7) (.606.4352,615.3611) (.773437, 725.2015)
(.4, .6) (606.4352,615.3611) (.773437, 725.2015)
(.5, .5) (606.4352,615.3611) (.773437, 725.2015)
(.6, .4) (606.4352,615.3611) (.773437, 725.2015)
(.7, .3) (606.4352,615.3611) (.773437, 725.2015)
(.8, .2) (606.4352,615.3611) (.773437, 725.2015)
(.9, .1) (606.4352,615.3611) (.773437, 725.2015)

It is shown that, in the weighted sum method, the DMs needs to provide fixed weights
for each objective in advance. These fixed weights may not reflect the DMs true
preferences accurately and may lead to biased or suboptimal solutions.
In Table 5, for each weight combination the closeness degree to ideal solution of
solutions obtained by PHFWA operator is better than the solution obtained by HFWA
operator, similarly the solution derived from the PHFWG operator is more proximate to
the ideal solution than the solution obtained by HFWG operator.
Table 5: Comparison of the optimal solution of manufacturing system by degree of closeness to
ideal solution
Weight Degree of closeness to HFWA PHFWA HFWG PHFWG
ideal solution
(.1, .9) 𝐷1 0.0636 0.0529 0.0574 0.0559
𝐷2 0.0636 0.0517 0.0571 0.0553
𝐷∞ 0.0636 0.0516 0.0571 0.0553
(.2, .8) 𝐷1 0.1035 0.0823 0.1015 0.0957
𝐷2 0.1023 0.0806 0.0999 0.0933
𝐷∞ 0.1023 0.0806 0.0998 0.0933
(.3, .7) 𝐷1 0.1313 0.0980 0.1303 0.1183
𝐷2 0.1280 0.0888 0.1269 0.1132
𝐷∞ 0.1279 0.0832 0.1269 0.1131
(.4, .6) 𝐷1 0.1426 0.0930 0.1422 0.1230
𝐷2 0.1368 0.0792 0.1363 0.1145
𝐷∞ 0.1366 0.0777 0.1362 0.1142
(.5, .5) 𝐷1 0.1354 0.0727 0.1353 0.1096
𝐷2 0.1266 0.0561 0.1265 0.0976
𝐷∞ 0.1263 0.0522 0.1261 0.0968
(.6, .4) 𝐷1 0.1081 0.0401 0.1082 0.0787
𝐷2 0.0966 0.0289 0.0967 0.0644
𝐷∞ 0.0958 0.0240 0.0959 0.0622
(.7, .3) 𝐷1 0.0602 0.0206 0.0613 0.0206
𝐷2 0.0476 0.0206 0.0488 0.0206
𝐷∞ 0.0451 0.0206 0.0464 0.0206
(.8, .2) 𝐷1 0.0137 0.0137 0.0137 0.0137
𝐷2 0.0137 0.0137 0.0137 0.0137
𝐷∞ 0.0137 0.0137 0.0137 0.0137
(.9, .1) 𝐷1 0.0069 0.0069 0.0069 0.0069
𝐷2 0.0069 0.0069 0.0069 0.0069
𝐷∞ 0.0069 0.0069 0.0069 0.0069
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 223

To show the efficiency of the proposed method closeness degree to ideal solution is
calculated and listed in Table 6. From this table, except the weight (.1, .9) the closeness
degree to ideal solution of the obtained solution by PHFWA operator is better than the
closeness of the solution obtained by HFWA operator and closeness degree of the solution
obtained by PHFWG operator is better than the solution obtained by HFWG operator.

Table 6: Comparison of the optimal solution of LSS by degree of closeness to ideal solution
Weight Degree of closeness to HFWA PHFWA HFWG PHFWG
ideal solution
(.1, .9) 𝐷1 0.0226 0.0541 0.0226 0.0541
𝐷2 0.0226 0.0430 0.0226 0.0430
𝐷∞ 0.0226 0.0410 0.0226 0.0410
(.2, .8) 𝐷1 0.0453 0.0627 0.0638 0.0627
𝐷2 0.0453 0.0449 0.0460 0.0449
𝐷∞ 0.0453 0.0364 0.0385 0.0364
(.3, .7) 𝐷1 0.0832 0.0793 0.0832 0.0794
𝐷2 0.0711 0.0637 0.0711 0.0638
𝐷∞ 0.0698 0.0610 0.0698 0.0611
(.4, .6) 𝐷1 0.0823 0.0811 0.0821 0.0810
𝐷2 0.0768 0.0740 0.0764 0.0738
𝐷∞ 0.0766 0.0736 0.0761 0.0734
(.5, .5) 𝐷1 0.0790 0.0721 0.0724 0.0720
𝐷2 0.0790 0.0687 0.0695 0.0683
𝐷∞ 0.0790 0.0687 0.0694 0.0682
(.6, .4) 𝐷1 0.0632 0.0631 0.0598 0.0598
𝐷2 0.0632 0.0631 0.0584 0.0581
𝐷∞ 0.0632 0.0631 0.0583 0.0581
(.7, .3) 𝐷1 0.0474 0.0473 0.0459 0.0458
𝐷2 0.0474 0.0473 0.0451 0.0450
𝐷∞ 0.0474 0.0473 0.0451 0.0450
(.8, .2) 𝐷1 0.0316 0.0316 0.0310 0.0310
𝐷2 0.0316 0.0316 0.0307 0.0307
𝐷∞ 0.0316 0.0316 0.0307 0.0307
(.9, .1) 𝐷1 0.0158 0.0158 0.0157 0.0157
𝐷2 0.0158 0.0158 0.0156 0.0156
𝐷∞ 0.0158 0.0158 0.0156 0.0156

6. LIMITATION OF OUR STUDY

We have identified certain limitations in our approach. While our method seeks to
obtain the Pareto optimal solution, it's important to note that the reverse may not hold true.
This is because we employ a weighted sum of the aggregated PHF membership and non-
membership as a utility function, which may not effectively explore solutions in the non-
convex region of the Pareto front, if one exists. To address this limitation and find Pareto
optimal solutions in the non-convex portion of the Pareto front, alternative approaches such
as metaheuristic methods or ϵ-constraint methods with suitable ϵ bounds on the objective
function are necessary. By explicitly discussing these limitations, we aim to offer a
transparent and balanced assessment of our work and provide guidance for future research
in this field.
224 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

7. CONCLUSIONS
In this paper, a multi-objective optimization approach is presented in PHF environment.
To convert the uncertain multi-objective optimization problem to single objective
optimization problem the PHFWA and PHFWG operators are used. So, here PHF
environments gives the option to choose memberships and non-memberships in much
range of area than FS and IFS to decision making experts. For uncertainty of the parameters
of objective functions, that are taken as PPFN. This method conclude that the result
obtained by using PHFWA and PHFWG operator is better than the result obtained by
HFWA and HFWG operator by degree of closeness to ideal solution. In order to
demonstrate the practicality of the suggested approach a numerical example of
manufacturing system and a real-life multi-objective ROM of LSS is solved. A sensitivity
analysis of the optimal solution is presented by different weights of the objectives and
giving the same weight to all DM’s preferences. A comparison of the solutions obtained
by proposed method and existing method is presented by closeness degree to ideal solution.
The proposed method is also useful when several experts are involved in decision making
process and they are confused about the exact value of parameters and goal of the
optimization problem.
In future, multi objective optimization model of various field such as engineering,
finance, inventory control, transportation, environmental management etc. can be solved
by the proposed method. We can make this method more reliable by using the extension
of PHF aggregation operator. At the place of linear membership, we can use non-linear
membership and non-membership functions (exponential, hyperbolic, etc.). Various
metaheuristic algorithm can be applied to the proposed model for more insightful result in
future.

Acknowledgements. The authors acknowledge the editors' and anonymous reviewers'

insightful remarks and recommendations, which helped to improve the paper's quality. The
CSIR-HRDG Fund, under grant EMR-09/106(0193)/2019-EMR-1, is gratefully
acknowledged by the first author for its financial assistance.

Funding. This research received no external funding.

REFERENCES

[1] L. A. Zadeh, "Fuzzy sets," Information and Control, vol. 8, no. 3, pp. 338-353, 1965/06/01/
1965, doi: https://doi.org/10.1016/S0019-9958(65)90241-X.
[2] R. E. Bellman and L. A. Zadeh, "Decision-making in a fuzzy environment," Management
science, vol. 17, no. 4, pp. B-141-B-164, 1970.
[3] H. J. Zimmermann, "Fuzzy programming and linear programming with several objective
functions," Fuzzy Sets and Systems, vol. 1, no. 1, pp. 45-55, 1978/01/01/ 1978, doi:
https://doi.org/10.1016/0165-0114(78)90031-3.
[4] W. F. Abd El-Wahed, "A multi-objective transportation problem under fuzziness," Fuzzy sets
and systems, vol. 117, no. 1, pp. 27-33, 2001.
[5] M. Arenas Parra, A. Bilbao Terol, and M. V. Rodrı́guez Urı́a, "A fuzzy goal programming
approach to portfolio selection," European Journal of Operational Research, vol. 133, no. 2,
pp. 287-297, 2001/01/01/ 2001, doi: https://doi.org/10.1016/S0377-2217(00)00298-8.
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 225

[6] G. Mahapatra, B. Mahapatra, and P. Roy, "Fuzzy decision-making on reliability of series

system: Fuzzy geometric programming approach," Annals of fuzzy Mathematics and
Informatics, vol. 1, no. 1, pp. 107-118, 2011.
[7] S. Majumder, M. B. Kar, S. Kar, and T. Pal, "Uncertain programming models for multi-
objective shortest path problem with uncertain parameters," Soft Computing, vol. 24, no. 12,
pp. 8975-8996, 2020/06/01 2020, doi: 10.1007/s00500-019-04423-3.
[8] S. Majumder, P. Kundu, S. Kar, and T. Pal, "Uncertain multi-objective multi-item fixed charge
solid transportation problem with budget constraint," Soft Computing, vol. 23, no. 10, pp. 3279-
3301, 2019/05/01 2019, doi: 10.1007/s00500-017-2987-7.
[9] S. Majumder et al., "On multi-objective minimum spanning tree problem under uncertain
paradigm," Symmetry, vol. 14, no. 1, p. 106, 2022.
[10] S. Majumder, S. Kar, and T. Pal, "Mean-Entropy Model of Uncertain Portfolio Selection
Problem," in Multi-Objective Optimization: Evolutionary to Hybrid Framework, J. K. Mandal,
S. Mukhopadhyay, and P. Dutta Eds. Singapore: Springer Singapore, 2018, pp. 25-54.
[11] K. T. Atanassov, "Intuitionistic fuzzy sets," Fuzzy Sets and Systems, vol. 20, no. 1, pp. 87-96,
1986/08/01/ 1986, doi: https://doi.org/10.1016/S0165-0114(86)80034-3.
[12] P. P. Angelov, "Optimization in an intuitionistic fuzzy environment," Fuzzy Sets and Systems,
vol. 86, no. 3, pp. 299-306, 1997/03/16/ 1997, doi: https://doi.org/10.1016/S0165-
0114(96)00009-7.
[13] S. Dey and T. K. Roy, "Multi-objective Structural Optimization Using Fuzzy and Intuitionistic
Fuzzy Optimization," 2015.
[14] H. Garg, M. Rani, S. P. Sharma, and Y. Vishwakarma, "Intuitionistic fuzzy optimization
technique for solving multi-objective reliability optimization problems in interval
environment," Expert Syst. Appl., vol. 41, pp. 3157-3167, 2014.
[15] P. Ghosh, T. K. Roy, and C. Majumder, "Optimization of Industrial Wastewater Treatment
Using Intuitionistic Fuzzy Goal Geometric Programming Problem," Fuzzy Information and
Engineering, vol. 8, no. 3, pp. 329-343, 2016/09/01 2016, doi: 10.1016/j.fiae.2016.09.002.
[16] V. Ravi, P. J. Reddy, and H. J. Zimmermann, "Fuzzy global optimization of complex system
reliability," IEEE Transactions on Fuzzy Systems, vol. 8, no. 3, pp. 241-248, 2000, doi:
10.1109/91.855914.
[17] M. K. Sharma, V. Sharma, and R. Dangwal, "Reliability Analysis of A System Using
Intuitionstic Fuzzy Sets," 2012.
[18] S. Islam and T. Kundu, "Application of Neutrosophic Optimization Technique on Multi-
objective Reliability Optimization Model," Peer Reviewers, p. 13.
[19] T. Kundu and S. Islam, "Neutrosophic Goal Geometric Programming Problem and Its
Application to Multi-objective Reliability Optimization Model," International Journal of
Fuzzy Systems, vol. 20, no. 6, pp. 1986-1994, 2018/08/01 2018, doi: 10.1007/s40815-018-
0479-2.
[20] R. R. Yager, "Pythagorean fuzzy subsets," in 2013 Joint IFSA World Congress and NAFIPS
Annual Meeting (IFSA/NAFIPS), 24-28 June 2013 2013, pp. 57-61, doi: 10.1109/IFSA-
NAFIPS.2013.6608375.
[21] L. Fei and Y. Deng, "Multi-criteria decision making in Pythagorean fuzzy environment,"
Applied Intelligence, vol. 50, no. 2, pp. 537-561, 2020/02/01 2020, doi: 10.1007/s10489-019-
01532-2.
[22] M. Akram, I. Ullah, T. Allahviranloo, and S. A. Edalatpanah, "Fully Pythagorean fuzzy linear
programming problems with equality constraints," Computational and Applied Mathematics,
vol. 40, no. 4, p. 120, 2021/04/24 2021, doi: 10.1007/s40314-021-01503-9.
[23] A. Luqman, M. Akram, and J. C. R. Alcantud, "Digraph and matrix approach for risk
evaluations under Pythagorean fuzzy information," Expert Systems with Applications, vol. 170,
p. 114518, 2021/05/15/ 2021, doi: https://doi.org/10.1016/j.eswa.2020.114518.
226 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach

[24] A. Habib, M. Akram, and C. Kahraman, "Minimum spanning tree hierarchical clustering
algorithm: A new Pythagorean fuzzy similarity measure for the analysis of functional brain
networks," Expert Systems with Applications, vol. 201, p. 117016, 2022/09/01/ 2022, doi:
https://doi.org/10.1016/j.eswa.2022.117016.
[25] M. Akram, A. Habib, and T. Allahviranloo, "A new maximal flow algorithm for solving
optimization problems with linguistic capacities and flows," Information Sciences, vol. 612,
pp. 201-230, 2022/10/01/ 2022, doi: https://doi.org/10.1016/j.ins.2022.08.068.
[26] S.-P. Wan, Z. Jin, and J.-Y. Dong, "Pythagorean fuzzy mathematical programming method for
multi-attribute group decision making with Pythagorean fuzzy truth degrees," Knowledge and
Information Systems, vol. 55, no. 2, pp. 437-466, 2018/05/01 2018, doi: 10.1007/s10115-017-
1085-6.
[27] L. Li and K. K. Lai, "A fuzzy approach to the multiobjective transportation problem,"
Computers & Operations Research, vol. 27, no. 1, pp. 43-57, 2000/01/01/ 2000, doi:
https://doi.org/10.1016/S0305-0548(99)00007-6.
[28] S. Liu, Z. Xu, and J. Gao, "A fuzzy compromise programming model based on the modified
S-curve membership functions for supplier selection," Granular Computing, vol. 3, no. 4, pp.
275-283, 2018/12/01 2018, doi: 10.1007/s41066-017-0066-1.
[29] V. Torra, "Hesitant fuzzy sets," International journal of intelligent systems, vol. 25, no. 6, pp.
529-539, 2010.
[30] X. Zhang, Z. Xu, and X. Xing, "Hesitant fuzzy programming technique for multidimensional
analysis of hesitant fuzzy preferences," OR Spectrum, vol. 38, no. 3, pp. 789-817, 2016/07/01
2016, doi: 10.1007/s00291-015-0420-0.
[31] S. K. Bharati, "Hesitant fuzzy computational algorithm for multiobjective optimization
problems," International Journal of Dynamics and Control, vol. 6, no. 4, pp. 1799-1806,
2018/12/01 2018, doi: 10.1007/s40435-018-0417-z.
[32] M. Xia, Z. Xu, and N. Chen, "Some Hesitant Fuzzy Aggregation Operators with Their
Application in Group Decision Making," Group Decision and Negotiation, vol. 22, no. 2, pp.
259-279, 2013/03/01 2013, doi: 10.1007/s10726-011-9261-7.
[33] M. Xia and Z. Xu, "Hesitant fuzzy information aggregation in decision making," International
Journal of Approximate Reasoning, vol. 52, no. 3, pp. 395-407, 2011/03/01/ 2011, doi:
https://doi.org/10.1016/j.ijar.2010.09.002.
[34] G.-L. Xu, S.-P. Wan, and J.-Y. Dong, "A hesitant fuzzy programming method for hybrid
MADM with incomplete attribute weight information," Informatica, vol. 27, no. 4, pp. 863-
892, 2016.
[35] S.-P. Wan, Y.-L. Qin, and J.-Y. Dong, "A hesitant fuzzy mathematical programming method
for hybrid multi-criteria group decision making with hesitant fuzzy truth degrees," Knowledge-
Based Systems, vol. 138, pp. 232-248, 2017/12/15/ 2017, doi:
https://doi.org/10.1016/j.knosys.2017.10.002.
[36] F. Ahmad, A. Y. Adhami, B. John, and A. Reza, "A novel approach for the solution of
multiobjective optimization problem using hesitant fuzzy aggregation operator," RAIRO-Oper.
Res., vol. 56, no. 1, pp. 275-292, 2022. [Online]. Available:
https://doi.org/10.1051/ro/2022006.
[37] S. K. Bharati, "Hesitant intuitionistic fuzzy algorithm for multiobjective optimization
problem," Operational Research, vol. 22, no. 4, pp. 3521-3547, 2022/09/01 2022, doi:
10.1007/s12351-021-00685-8.
[38] A. Y. Adhami and F. Ahmad, "Interactive Pythagorean-hesitant fuzzy computational algorithm
for multiobjective transportation problem under uncertainty," International Journal of
Management Science and Engineering Management, vol. 15, no. 4, pp. 288-297, 2020/10/01
2020, doi: 10.1080/17509653.2020.1783381.
 S. Jana and S. Islam / A Pythagorean Hesitant Fuzzy Programming Approach 227

[39] M. S. A. Khan, S. Abdullah, A. Ali, and K. Rahman, "Pythagorean Hesitant Fuzzy Information
Aggregation and Their Application to Multi-Attribute Group Decision-Making Problems,"
Journal of Intelligent Systems, vol. 29, no. 1, pp. 154-171, 2020, doi: doi:10.1515/jisys-2017-
0231.
[40] B. Batool, S. Abdullah, S. Ashraf, and M. Ahmad, "Pythagorean probabilistic hesitant fuzzy
aggregation operators and their application in decision-making," Kybernetes, vol. 51, no. 4, pp.
1626-1652, 2022.
[41] M. S. A. Khan, S. Abdullah, A. Ali, F. Amin, and F. Hussain, "Pythagorean hesitant fuzzy
Choquet integral aggregation operators and their application to multi-attribute decision-
making," Soft Computing, vol. 23, no. 1, pp. 251-267, 2019/01/01 2019, doi: 10.1007/s00500-
018-3592-0.
[42] W. F. Abd El-Wahed and S. M. Lee, "Interactive fuzzy goal programming for multi-objective
transportation problems," Omega, vol. 34, no. 2, pp. 158-166, 2006.
[43] S. K. Singh and S. P. Yadav, "Modeling and optimization of multi objective non-linear
programming problem in intuitionistic fuzzy environment," Applied Mathematical Modelling,
vol. 39, no. 16, pp. 4617-4629, 2015/08/15/ 2015, doi:
https://doi.org/10.1016/j.apm.2015.03.064.