International Journal of Computational Intelligence Systems

Volume 14, Issue 1, 2021, Pages 1809 - 1822

Harmonically Convex Fuzzy-Interval-Valued Functions and Fuzzy-Interval Riemann–Liouville Fractional Integral Inequalities

Authors
Gul Sana1, ORCID, Muhammad Bilal Khan1, Muhammad Aslam Noor1, Pshtiwan Othman Mohammed2, Yu-Ming Chu3, *
1Department of Mathematics, COMSATS University Islamabad, Islamabad, Pakistan
2Department of Mathematics, College of Education, University of Sulaimani, Sulaimani, Iraq
3Department of Mathematics, Huzhou University, Huzhou, P. R. China
*Corresponding author. Email: chuyuming@zjhu.edu.cn
Corresponding Author
Yu-Ming Chu
Received 5 April 2021, Accepted 14 June 2021, Available Online 28 June 2021.
DOI
10.2991/ijcis.d.210620.001How to use a DOI?
Keywords
Harmonically convex fuzzy interval-valued function; Fuzzy interval fractional integral operator; Hermite–Hadamard inequality; Hermite–Hadamard–Fejér inequality
Abstract

It is well known that the concept of convexity establishes strong relationship with integral inequality for single-valued and interval-valued function. The single-valued function and interval-valued function both are special cases of fuzzy interval-valued function. The aim of this paper is to introduce a new class of convex fuzzy interval-valued functions, which is called harmonically convex fuzzy interval-valued functions (harmonically convex fuzzy-IVFs) by means of fuzzy order relation and to investigate this new class via fuzzy-interval Riemann–Liouville fractional operator. With the help of fuzzy order relation and fuzzy-interval Riemann–Liouville fractional, we derive some integrals inequalities of Hermite–Hadamard (H-H) type and Hermite–Hadamard–Fejér (H-H Fejér) type as well as some product inequities for harmonically convex fuzzy-IVFs. Our results represent a significant improvement and refinement of the known results. We hope that these interesting outcomes may open a new direction for fuzzy optimization, modeling and interval-valued function.

Copyright
© 2021 The Authors. Published by Atlantis Press B.V.
Open Access
This is an open access article distributed under the CC BY-NC 4.0 license (http://creativecommons.org/licenses/by-nc/4.0/).

1. INTRODUCTION

The Hermite–Hadamard (H-H) inequality was firstly introduced by Hadamard [1] and Hermite [2] for convex functions. This inequality is used as a most useful tool in mathematical analysis and optimization because convex functions establish strong relationship with H-H inequality. Therefore, many authors have discussed the relation of H-H inequality with different kinds of convex and nonconvex functions and many papers have provided refinements, generalizations and extensions, see [38]. Besides, fractional integrals have played a critical role in different branches of sciences. It is also a familiar fact that inequalities have become a very popular method using fractional integrals, and that this method has been the driving force behind many studies in recent years. Many forms of inequality have been studied, resulting in the introduction of new trend in inequality theory. Firstly, by using fractional integrals, Sarikaya et al. [9] discovered fractional H-H inequality for classical convex function. After that, many scholars devoted their efforts to present fractional H-H type inequalities for different classes of convex and nonconvex functions see [1015].

It is well known that interval analysis provides tools to deal with data uncertainty. In general, interval analysis is typically used to deal with the models whose data are composed of inaccuracies that may occur from certain kinds of measurements. In 1966, the concept of interval analysis was firstly introduced by late American mathematician Ramon E. Moore in [16]. Since its inception, various authors in the mathematical community have paid close attention to this area of research. Interval analysis has been found to be useful in global optimization and constraint solution algorithms, according to experts. It has slowly risen in popularity over the last few decades. Scientists and engineers engaged in scientific computation have discovered that interval analysis is useful, especially in terms of accuracy, round-off error affects and automatic validation of results. After the invention of interval analysis, the researchers working in the area of inequalities wants to know whether the inequalities in abovementioned results can be found substituted with inclusions relation. In certain cases, the question is answered correctively. Recently, through interval Riemann integral, interval Riemann–Liouville fractional integrals and fuzzy Riemann integral, several authors presented new versions of various inequalities for interval and fuzzy-interval-valued functions like, as one can see Costa [17], Costa and Roman-Flores [18], Roman-Flores et al. [19,20], and Chalco-Cano et al. [21,22], An et al. [23], Zhao et al. [24], but also to more general set-valued maps by Nikodem et al. [25], Matkowski and Nikodem [26]. In particular, Zhang et al. [27] derived the new version of Jensen's inequalities for set-valued and fuzzy set-valued functions by means of a pseudo order relation and proved that these Jensen's inequalities generalized form of Costa Jensen's inequalities [17]. As a further extension, more and more, H-H type inequalities have been obtained through interval Riemann–Liouville fractional integrals, see for convex-IVFs [28,29], for harmonically convex-IVFs [30]. Moreover, recently, Khan et al. [31] introduced the new class of convex fuzzy mappings is known as h1,h2-convex fuzzy-IVFs by means fuzzy order relation and presented the following new version of H-H-type inequality for h1,h2-convex fuzzy-IVF involving fuzzy-interval Riemann integrals:

Let Q˜:u,νF0 be a h1,h2-convex fuzzy-IVF with h1,h2:0,1+ and h112h2120, whose β-levels define the family of IVFs Qβ:u,νKC+ are given by Qβz=Qz,β,Qz,β for all zu,ν, β0,1. If Q˜ is fuzzy-interval Riemann integrable (in sort, FR-integrable), then

12h112h212Q˜u+ν21νuFRuνQ˜zdzQ˜u+˜Q˜ν01h1ϱh21ϱdϱ.(1)

If h1ϱ=ϱ and h2ϱ1, then inequality (1) reduces to the following inequality:

Q˜u+ν21νuFRuνQ˜zdzQ˜u+˜Q˜ν2,(2)
where Q˜ is a convex fuzzy-IVF. We urge the readers to [3242] and the citations therein for further review of related literature on the implementations and characterization of fuzzy-interval, inequalities and generalized convex fuzzy mappings.

This study is organized as follows: Section 2 presents preliminary notions and results in interval space, and fuzzy-interval space. Moreover, Section 2 also discusses new class of convex fuzzy-IVF, which is known as harmonically convex fuzzy-IVF. Section 3 obtains fuzzy-interval H-H inequalities via convex fuzzy-IVFs. In particular, some intriguing examples are provided to support our outcomes. Conclusions and future plans are discussed in Section 4.

2. PRELIMINARIES

Let KC be the space of all closed and bounded intervals of and ηKC be defined by

η=η,η=z|ηzη,η,η.(3)

If η=η then, η is said to be degenerate. In this article, all intervals will be nondegenerate intervals. If η0, then η,η is called positive interval. The set of all positive interval is denoted by KC+ and defined as KC+=η,η:η,ηKC and η0.

Let ϱ and ϱη be defined by

ϱ.η=ϱη,ϱη if ϱ0,ϱη,ϱη if ϱ<0.(4)

Then the Minkowski difference ζη, addition η+ζ and η×ζ for η,ζKC are defined by

ζ,ζη,η=ζη,ζη,ζ,ζ+η,η=ζ+η,ζ+η,(5)
and
ζ,ζ×η,η=minζη,ζη,ζη,ζη,maxζη,ζη,ζη,ζη.(6)

The inclusion “” means that

ζη If and only if,ζ,ζη,η,if and only ifηζ,ζη.(7)

Remark 2.1.

[32] The relation “I” defined on KC by

ζ,ζIη,η if and only if ζη,ζη,(8)
for all ζ,ζ,η,ηKC, it is an order relation. For given ζ,ζ,η,ηKC, we say that ζ,ζIη,η if and only if ζη,ζη or ζη,ζ<η.

For ζ,ζ,η,ηI, the Hausdorff–Pompeiu distance between intervals ζ,ζ and η,η is defined by

dζ,ζ,η,η=max|ζη|,|ζη|.(9)

It is familiar fact that I,d is a complete metric space.

Assume is a set of real numbers. The membership function is a mapping ζ˜:0,1 that characterizes a fuzzy subset A of , for each fuzzy set and β(0,1], then β-level sets of ζ˜ is denoted and defined as follows: ζβ=u|ζ˜uβ. If β=0, thensuppζ˜=z|ζ˜z>0 is called support of ζ˜. By ζ˜0 we define the closure of suppζ˜.

Let F be the family of all fuzzy sets and ζ˜F denote the family of all nonempty sets. ζ˜F be a fuzzy set. Then we define the following:

  1. ζ˜ is said to be normal if there exists z and ζ˜z=1;

  2. ζ˜ is said to be upper semi continuous on if for given z, there exist ε>0 there exist δ>0 such that ζ˜zζ˜y<ε for all y with |zy|<δ;

  3. ζ˜ is said to be fuzzy convex if ζβ is convex for every β0,1;

  4. ζ˜ is compactly supported if suppζ˜ is compact.

A fuzzy set is called a fuzzy number or fuzzy interval if it has properties (1), (2), (3) and (4). We denote by F0 the family of all interval.

Let ζ˜F0 be a fuzzy-interval, if and only if, β levels ζ˜β is a nonempty compact convex set of . From these definitions, we have

ζ˜β=ζβ,ζβ,
where
ζβ=infz|ζ˜zβ,ζβ=supz|ζ˜zβ.

Proposition 2.2.

[18] If ζ˜,η˜F0 then relation “” defined on F0 by

ζ˜η˜ if and only if,ζ˜βIη˜β,for all β0,1,(10)
this relation is known as partial order relation.

For ζ˜,η˜F0 and ϱ, the sum ζ˜+˜η˜, product ζ˜×˜η˜, scalar product ϱ.ζ˜ and sum with scalar are defined by

Then, for all β0,1, we have

ζ˜+˜η˜β=ζ˜β+η˜β,(11)
ζ˜×˜η˜β=ζ˜β×η˜β,(12)
ϱ.ζ˜β=ϱ.ζ˜β.(13)
ϱ+˜ζ˜β=ϱ+ζ˜β.(14)

For ζ˜F0 such that ξ˜=η˜+˜ζ˜, then by this result we have existence of Hukuhara difference of ξ˜ and η˜, and we say that ζ˜ is the H-difference of ξ˜ and η˜, and denoted by ξ˜˜η˜. If H-difference exists, then

ζβ=ξηβ=ξβηβ,ζβ=ξηβ=ζβηβ.(15)

A partition of u,ν is any finite ordered subset P having the form

P=u=z1<z2<z3<z4<z5<zk=ν.

The mesh of a partition P is the maximum length of the subintervals containing P, that is,

mashP=maxzjzj1:j=1,2,3,k.

Let Pδ,u,ν be the set of all PPδ,u,ν such that mesh P<δ. For each interval zj1,zj, where 1jk, choose an arbitrary point μj and taking the sum

SQ,P,δ=j=1kQμjzjzj1,
where Q:u,νI. We call SQ,P,δ a Riemann sum of Q corresponding to PPδ,u,ν.

Definition 2.3.

[24] A function Q:u,νI is called interval Riemann integrable (IR-integrable) on u,ν if there exists BI such that, foe each ϵ, there exists δ>0 such that

dSQ,P,δ,B<ϵ,
for every Riemann sum of Q corresponding to PPδ,u,ν and for arbitrary choice of μjzj1,zj for 1jk. Then we say that B is the IR-integral of Q on u,ν and is denoted by B=IRuνQzdz.

Moore [16] firstly proposed the concept of Riemann integral for IVF and it is defined as follows:

Theorem 2.4.

[16] If Q:u,νI is an IVF on such that Qz=Q,Q. Then Q is Riemann integrable over u,ν if and only if, Q and Q both are Riemann integrable over u,ν such that

IRuνQzdz=RuνQudz,RuνQudz.(16)

Definition 2.5.

[33] A fuzzy map Q˜:KF0 is also known as fuzzy-IVF. For each β0,1, whose β levels characterize the family of IVFs Qβ:KKC are given by Qβz=Qz,β,Qz,β for all zK. Here, for each β0,1, the left and right real-valued functions Qz,β,Qz,β:K are also called lower and upper functions of Q˜.

Remark 2.6.

If Q˜:KF0 is a fuzzy-IVF then, Q˜z is called continuous function at zK, if for each β0,1, both left and right real-valued functions Qz,β and Qz,β are continuous at zK.

The following conclusion can be drawn from the above literature review, see [16,24,33].

Definition 2.7.

Let Q˜:c,dF0 be a fuzzy-IVF. Then, fuzzy Riemann integral of Q˜ over c,d, denoted by FRcdQ˜zdz, it is defined level by level

FRcdQ˜zdzβ=IRcdQβzdz=cdQz,βdz:Qz,βRc,d,(17)
for all β0,1, where Rc,d contains the family of left and right functions of IVFs. Q˜ is FR-integrable over c,d if FRcdQ˜zdzF0. Note that, if left and right real-valued functions are Lebesgue-integrable, then Q˜ is fuzzy Aumann-integrable over c,d, denoted by FAcdQ˜zdz, see [33].

Theorem 2.8.

Let Q˜:c,dF0 be a fuzzy-IVF, whose β levels characterize the collection of IVFs Qβ:c,dKC are defined by Qβz=Qz,β,Qz,β for all zc,d and for all β0,1. Then Q˜ is FR-integrable over c,d if and only if, Qz,β and Qz,β both are R-integrable over c,d. Moreover, if Q˜ is FR-integrable over c,d, then

FRcdQ˜zdzβ=RcdQz,βdz,RcdQz,βdz=IRcdQβzdz,(18)
for all β0,1. For each β0,1, QRc,d,β and Rc,d,β denote the collection of all FR-integrable fuzzy-IVFs and, R-integrable left and right functions over c,d.

Allahviranloo et al. [39] introduced the following fuzzy-interval Riemann–Liouville fractional integral operators:

Let α>0 and Lu,ν,F0 be the collection of all Lebesgue measurable fuzzy-IVFs on u,ν. Then the fuzzy-interval left and right Riemann–Liouville fractional integral of Q˜Lu,ν,F0 with order α>0 are defined by

Iu+αQ˜z=1Γαuzzϱα1Q˜ϱdϱ,z>u,(19)
and
IναQ˜z=1Γαzνϱzα1Q˜ϱdϱ,(z<ν),(20)
respectively, where Γz=0ϱz1uϱdϱ is the Euler gamma function. The fuzzy-interval left and right Riemann–Liouville fractional integral z based on left and right end point functions can be defined, that is,
Iu+αQzβ=1Γαuzzϱα1Qβϱdϱ=1Γαuzzϱα1Qϱ,β,Qϱ,βdϱ,z>u,(21)
where
Iu+αQz,β=1Γαuzzϱα1Qϱ,βdϱ,z>u,(22)
and
Iu+αQz,β=1Γαuzzϱα1Qϱ,βdϱ,z>u,(23)

Similarly, we can define right Riemann–Liouville fractional integral Q of z based on left and right end point functions.

Definition 2.9.

A set K=u,v+=0, is said to be harmonically convex set, if, for all z,yK,ϱ0,1, we have

zyϱz+1ϱyK.(24)

Definition 2.10.

[3] The fuzzy-IVF Q:u,νF0 is called harmonically convex fuzzy-IVF on u,ν if

Qzyϱz+1ϱy1ϱQz+ϱQy,(25)
for all z,yu,ν,ϱ0,1, where Qz0 for all zu,ν. If (25) is reversed then, Q is called harmonically concave fuzzy-IVF on u,ν.

Definition 2.11.

The fuzzy-IVF Q˜:u,νF0 is called harmonically convex fuzzy-IVF on u,ν if

Q˜zyϱz+1ϱy1ϱQ˜z+˜ϱQ˜y,(26)
for all z,yu,ν,ϱ0,1, where Q˜z0˜, for all zu,ν. If (26) is reversed then, Q˜ is called concave fuzzy-IVF on u,ν.

Theorem 2.12.

Let K be harmonically convex set, and let Q˜:KFC be a fuzzy-IVF whose β levels define the family of IVFs Qβ:KKC+KC are given by

Qβz=Qz,β,Qz,β,zK,(27)
for all zK, β0,1. Then Q˜ is harmonically convex on K, if and only if, for all β0,1,Qz,β and Qz,β are harmonically convex.

Proof.

Assume that for each β0,1,Qz,β and Qz,β are harmonically convex on K. Then from (25), we have

Qzyϱz+1ϱy,β1ϱQz,β+ϱQy,β,
and
Qzyϱz+1ϱy,β1ϱQz,β+ϱQy,β.

Then by (26), (19) and (21), we obtain

Qβzyϱz+1ϱy=Qϱz+1ϱy,β,Qϱz+1ϱy,β,I1ϱQz,β,Qz,β+ϱQy,β,Qy,β,
that is,
Q˜zyϱz+1ϱy1ϱQ˜z+˜ϱQ˜y,

z,yK,ϱ0,1. Hence, Q˜ is harmonically convex fuzzy-IVF on K.

Conversely, let Q˜ be harmonically convex fuzzy-IVF on K. Then for all z,yK, ϱ0,1, we have

Q˜zyϱz+1ϱy1ϱQ˜z+˜ϱQ˜y.

Therefore, from (26), for each β0,1, left side of above inequality, we have

Qβzyϱz+1ϱy=Qzyϱz+1ϱy,β,Qzyϱz+1ϱy,β.

Again, from (26), we obtain

1ϱQβz+ϱQβz=1ϱQz,β,Qz,β+ϱQy,β,Qy,β,
for all z,yK, ϱ0,1. Then by harmonically convexity of Q˜, we have for all z,yK, ϱ0,1 such that
Qzyϱz+1ϱy,β1ϱQz,β+ϱQy,β,
and
Qzyϱz+1ϱy,β1ϱQz,β+ϱQy,β,
for each β0,1. Hence, the result follows.

Remark 2.12.

If Qz,β=Qz,β and β=1 then from Definition 210, we obtain Definition 2.10.

Example 2.13.

We consider the fuzzy-IVFs Q˜:0,2FC defined by

Q˜zσ=σzσ0,z2σ2z2σ(z,2z]0otherwise.

Then, for each β0,1, we have Qβz=βz,2βz. Since end point functions Qz,β,Qz,β are harmonically convex functions for each β0,1. Hence Q˜z is harmonically convex fuzzy-IVF.

In next result, we will establish a relation between convex fuzzy-IVF and harmonically convex fuzzy-IVF.

Theorem 2.14.

Let Q˜:KFC be a fuzzy-IVF, where for all β0,1, whose β levels define the family of IVFs Qβ:KKC+KC are given by Qβz=Qz,β,Qz,β, for all zK. Then Q˜z is harmonically convex fuzzy-IVF on K, if and only if, Q˜1z is convex fuzzy-IVF on K.

Proof.

Since Q˜z is a harmonically convex fuzzy-IVF then, for z,yu,ν,ϱ0,1, we have

Q˜zyϱz+1ϱy1ϱQ˜z+˜ϱQ˜y.

Therefore, for each β0,1, we have

Qzyϱz+1ϱy,β1ϱQz,β+ϱQy,β,Qzyϱz+1ϱy,β1ϱQz,β+ϱQy,β.(28)

Consider φ˜z=Q˜1z. Taking m=1z and n=1y to replace z and y, respectively. Then for each β0,1, applying (28)

Q1zyϱ1z+1ϱ1y,β=Q11ϱz+ϱy,β=φ1ϱz+ϱy,βϱQ1y,β+1ϱQ1z,β=ϱφy,β+1ϱφz,β,Q1zyϱ1z+1ϱ1y,β=Q11ϱz+ϱy,β=φ1ϱz+ϱy,βϱQ1y,β+1ϱQ1z,β=ϱφy,β+1ϱφz,β.

It follows that

Q1zyϱ1z+1ϱ1y,β,Q1zyϱ1z+1ϱ1y,β=φ1ϱz+ϱy,β,φ1ϱz+ϱy,βϱφy,β,φy,β+1ϱφz,β,φz,β.
which implies that
φβ1ϱz+ϱyIϱφβy+1ϱφβz,
that is,
φ˜1ϱz+ϱyϱφ˜y+˜1ϱφ˜z.

This concludes that φ˜z is a convex fuzzy-IVF.

Conversely, let φ˜ is convex fuzzy-IVF on K. Then, for all z,yK, ϱ0,1, we have

φ˜ϱz+1ϱyϱφ˜z+˜1ϱφ˜y.

By using same steps as above, for each β0,1, we have

φϱ1z+1ϱ1y,β=Q1ϱ1z+1ϱ1y,β=Qzy1ϱz+ϱy,βϱφ1z,β+1ϱφ1y,β=ϱQz,β+1ϱQy,β
φϱ1z+1ϱ1y,β=Q1ϱ1z+1ϱ1y,β=Qzy1ϱz+ϱy,βϱφ1z,β+1ϱφ1y,β=ϱQz,β+1ϱQy,β.

It follows that

Qβzyϱz+1ϱyI1ϱQβz+ϱQβy,
that is,
Q˜zyϱz+1ϱy1ϱQ˜z+˜ϱQ˜y,
the proof the theorem has been completed.

Remark 2.15.

If Qz,β=Qz,β and β=1 then from Theorem 2.14, we obtain Lemma 2.1 of [13].

3. FUZZY-INTERVAL FRACTIONAL HERMITE–HADAMARD INEQUALITIES

In this section, we shall continue with the following the fractional HH inequality for harmonically convex fuzzy-IVFs and we also give fractional HH Fejér inequality for harmonically convex fuzzy-IVF through fuzzy order relation. In what follows, we denote by Lu,ν,F0 the family of Lebesgue measureable fuzzy-IVFs.

Theorem 3.1.

Let Q˜:u,νF0 be a harmonically convex fuzzy-IVF on u,ν, whose β levels define the family of IVFs Qβ:u,νKC+ are given by Qβz=Qz,β,Qz,β for all zu,ν, β0,1. If Q˜Lu,ν,F0, then

Q˜2uνu+νΓα+12νuαI1uαQ˜ψ1ν+˜I1ν+αQ˜ψ1uQ˜u+˜Q˜ν2.(29)

If Q˜z is concave fuzzy-IVF then

Q˜2uνu+νΓα+12νuαI1uαQ˜ψ1ν+˜I1ν+αQ˜ψ1uQ˜u+˜Q˜ν2.(30)
where ψz=1z.

Proof.

Let Q˜:u,νF0 be harmonically convex fuzzy-IVF. Then, by hypothesis, we have

2Q˜2uνu+νQ˜uνϱu+1ϱν+˜Q˜uν1ϱu+ϱν.

Therefore, for each β0,1, we have

2Q2uνu+ν,βQuνϱu+1ϱν,β+Quν1ϱu+ϱν,β,2Q2uνu+ν,βQuνϱu+1ϱν,β+Quν1ϱu+ϱν,β.

Consider φ˜z=Q˜1z. By Theorem 2.14 we have φ˜z is convex fuzzy-IVF then for each β0,1, above inequality, we have

2φu+ν2uν,βφϱu+1ϱνuν,β+φ1ϱu+ϱνuν,β.

Multiplying both sides by ϱα1 and integrating the obtained result with respect to ϱ over 0,1, we have

201ϱα1φu+ν2uν,βdϱ01ϱα1φϱu+1ϱνuν,βdϱ+01ϱα1φ1ϱu+ϱνuν,βdϱ.

Let z=1ϱu+ϱνuν and y=ϱu+1ϱνuν. Then we have

2αφu+ν2uν,βuννuα1ν1u1uyα1φy,βdy+uννuα1ν1uz1να1φz,βdz=ΓαuννuαI1uαφ1ν,β+I1ν+αφ1u,β.

Similarly, for Qz,γ, we have

2αφu+ν2uν,βΓαuννuαI1uαφ1ν,β+I1ν+αφ1u,β.

It follows that

2φu+ν2uν,β,φu+ν2uν,βIΓα+1uννuαI1uαφ1ν,β+I1ν+αφ1u,β,I1uαφ1ν,β+I1ν+αφ1u,β.

That is,

2φ˜u+ν2uνΓα+1uννuαI1uαφ˜1ν+˜I1ν+αφ˜1u.(31)

In a similar way as above, we have

ΓαuννuαI1uαφ˜1ν+˜I1ν+αφ˜1uφ˜1u+˜φ˜1να.(32)

Combining (31) and (32), we have

φ˜u+ν2uνΓα+1uννuα2I1uαφ˜1ν+˜I1ν+αφ˜1uφ˜1u+˜φ˜1ν2.

That is,

Q˜2uνu+νΓα+12νuαI1uαQ˜ψ1ν+˜I1ν+αQ˜ψ1uQ˜u+˜Q˜ν2.

Hence, the required result.

Remark 3.2.

If α=1, then inequality (29) reduces to the following inequality which is also new one:

Q˜2uνu+νuννuuνQ˜zz2dzQ˜u+˜Q˜ν2.(33)

If Qz,β=Qz,β with β=1 then, we obtain classical fractional H-H inequality for harmonically convex function which is given in [13]:

Q2uνu+νΓα+12νuαI1uαQψ1ν+I1ν+αQψ1uQu+Qν2.(34)

If Qz,β=Qz,β with β=1 and α=1 then, we obtain classical H-H inequality for harmonically convex function which is given in [3].

Q2uνu+νuννuuνQzz2dzQu+Qν2.(35)

Theorem 3.3.

(Second fuzzy fractional HH Fejér inequality) Let Q˜:u,νF0 be a harmonically convex fuzzy-IVF with u<ν, whose β levels define the family of IVFs Qβ:u,νKC+ are given by Qβz=Qz,β,Qz,β for all zu,ν, β0,1. If Q˜Lu,ν,F0 and Ω:u,ν,Ω11u+1ν1z=Ωz0, then

Iu+αQ˜Ωψν+˜IναQ˜ΩψuQ˜u+˜Q˜ν2I1ν+αΩψ1u+I1uαΩψ1ν.(36)

If Q˜ is concave fuzzy-IVF then, inequality (36) is reversed.

Proof.

Let Q˜ be a harmonically convex fuzzy-IVF and ϱα1Ωuνϱu+1ϱν0. Then, for each β0,1, we have

ϱα1Quνϱu+1ϱν,βΩuνϱu+1ϱνϱα11ϱQu,β+ϱQν,βΩuνϱu+1ϱν,(37)
and
ϱα1Quν1ϱu+ϱν,βΩuνϱu+1ϱνϱα1ϱQu,β+1ϱQν,βΩuνϱu+1ϱν.(38)

After adding (37) and (38), and integrating over 0,1, we get

01ϱα1Quνϱu+1ϱν,βΩuνϱu+1ϱνdϱ+01ϱα1Quν1ϱu+ϱν,βΩuνϱu+1ϱνdϱ01ϱα1Qu,βϱ+1ϱΩuνϱu+1ϱν+ϱα1Qν,β1ϱ+ϱΩuνϱu+1ϱνdϱ,=Qu,β01ϱα1Ωuνϱu+1ϱνdϱ+Qν,β01ϱα1Ωuνϱu+1ϱνdϱ,

Similarly, for Qz,γ, we have

01ϱα1Quνϱu+1ϱν,βΩuνϱu+1ϱνdϱ+01ϱα1Quν1ϱu+ϱν,βΩuνϱu+1ϱνdϱ=Qu,β01ϱα1Ωuνϱu+1ϱνdϱ+Qν,β01ϱα1Ωuνϱu+1ϱνdϱ.

From which, we have

ΓαuννuαIu+αQβΩψν+IναQβΩψuIΓαuννuαQβu+Qβν2I1ν+αΩψ1u+I1uαΩψ1ν,
that is,
Iu+αQ˜Ωψν+˜IναQ˜ΩψuQ˜u+˜Q˜ν2I1ν+αΩψ1u+I1uαΩψ1ν.(39)

Theorem 3.4.

(First fuzzy fractional HH Fejér inequality) Let Q˜:u,νF0 be a harmonically convex fuzzy-IVF with u<ν, whose β levels define the family of IVFs Qβ:u,νKC+ are given by Qβz=Qz,β,Qz,β for all zu,ν, β0,1. If Q˜Lu,ν,F0 and Ω:u,ν,Ω11u+1ν1z=Ωz0, then

Q˜2uνu+νI1ν+αΩψ1u+I1uαΩψ1νI1ν+αQ˜Ωψ1u+˜I1uαQ˜Ωψ1νQ˜u+˜Q˜ν2I1ν+αΩψ1u+I1uαΩψ1ν.(40)

If Q˜ is concave fuzzy-IVF then, inequality (40) is reversed.

Proof.

Since Q˜ is a harmonically convex fuzzy-IVF, then for β0,1, we have

Q2uνu+ν,β12Quνϱu+1ϱν,β+Quν1ϱu+ϱν,β.(41)

Multiplying both sides by (41) by ϱα1Ωuν1ϱu+ϱν and then integrating the resultant with respect to ϱ over 0,1, we obtain

Q2uνu+ν,β01ϱα1Ωuν1ϱu+ϱνdϱ1201ϱα1Quνϱu+1ϱν,βΩuν1ϱu+ϱνdϱ+01ϱα1Quν1ϱu+ϱν,βΩuν1ϱu+ϱνdϱ.(42)

Let z=uνϱu+1ϱν. Then, we have

2uννuαQ2uνu+ν,β1ν1uz1να1Ω1z,βdzuννuα1ν1uz1να1Q11u+1ν1z,βΩ1zdz+uννuαu1uz1να1Q1z,βΩ1zdz=uννuα1ν1u1uzα1Qz,βΩ11u+1ν1zdz+uννuα1ν1uz1να1Q1z,βΩ1zdz=Γαuννuα1ν+αQΩ1u+1uαQΩ1ν,(43)

Similarly, for Qz,γ, we have

2uννuαQ2uνu+ν,β1ν1uz1να1Ω1z,βdzΓαuννuαI1ν+αQΩ1u+I1uαQΩ1ν.(44)

From (43) and (44), we have

ΓαuννuαQ2uνu+ν,β,Q2uνu+ν,β                              .I1ν+αΩ1u+I1uαΩ1νIΓαuννuαI1ν+αQΩ1u+I1uαQΩ1ν,I1ν+αQΩ1u+I1uαQΩ1ν,
that is,
Q˜2uνu+νI1ν+αΩψ1u+I1uαΩψ1νI1ν+αQ˜Ωψ1u+˜I1uαQ˜Ωψ1ν.(45)

Similarly, if Q˜ be a harmonically convex fuzzy-IVF and ϱα1Ωuνϱu+1ϱν0, then, for each β0,1, we have

ϱα1Quνϱu+1ϱν,βΩuνϱu+1ϱνϱα11ϱQu,β+ϱQν,βΩuνϱu+1ϱν(46)
and
ϱα1Quν1ϱu+ϱν,βΩuνϱu+1ϱνϱα1ϱQu,β+1ϱQν,βΩuνϱu+1ϱν.(47)

After adding (46) and (47), and integrating the resultant over 0,1, we get

01ϱα1Quνϱu+1ϱν,βΩuνϱu+1ϱνdϱ+01ϱα1Quν1ϱu+ϱν,βΩuνϱu+1ϱνdϱ01ϱα1Qu,βϱ+1ϱuνϱu+1ϱν+ϱα1Qν,β1ϱ+ϱuνϱu+1ϱνdϱ,=Qu,β01ϱα1Ωuνϱu+1ϱνdϱ+Qν,β01ϱα1Ωuνϱu+1ϱνdϱ.

Similarly, for Qz,γ, we have

01ϱα1Quνϱu+1ϱν,βΩuνϱu+1ϱνdϱ+01ϱα1Quν1ϱu+ϱν,βΩuνϱu+1ϱνdϱ=Qu,β01ϱα1Ωuνϱu+1ϱνdϱ+Qν,β01ϱα1Ωuνϱu+1ϱνdϱ.

From which, we have

ΓαuννuαI1ν+αQβΩψν+I1uαQβΩψ1νIΓαuννuαQβu+Qβν2I1ν+αΩψ1u+I1uαΩψ1ν,
that is,
I1ν+αQ˜Ωψ1u+˜I1uαQ˜Ωψ1νQ˜u+˜Q˜ν2I1ν+αΩψ1u+I1uαΩψ1ν.(48)

By combining (45) and (48), we obtain the required inequality (40).

Remark 3.5.

Let α=1. Then from Theorems 3.3 and 3.4, we get following H-H inequality for harmonically convex fuzzy-IVF which is also new one:

Q˜2uνu+νuνΩzz2dzuνQ˜zz2ΩzdzQ˜u+˜Q˜ν2uνΩzz2dz.

Let Ωz=1. Then from Theorems 3.3 and 3.4, we obtain inequality (29).

Let Ωz=1 and α=1, then from Theorems 3.3 and 3.4, we get H-H inequality for harmonically convex fuzzy-IVF:

Q˜2uνu+νuννuuνQ˜zz2dzQ˜u+˜Q˜ν2.

If Qz,β=Qz,β with β=1 then from Theorems 3.3 and 3.4, we obtain classical fractional H-H Fejér inequality for harmonically convex function, given in [10].

Let Qz,β=Qz,β with β=1 and α=1. Then, from Theorems 3.3 and 3.4, we obtain classical H-H-Fejér inequality for harmonically convex function, given in [4].

If Qz,β=Qz,β with Ωz=β=1 then from Theorems 3.3 and 3.4, we obtain classical fractional H-H inequality for harmonically convex function.

If Qz,β=Qz,β and Ωz=β=α=1 then from Theorems 3.3 and 3.4, we obtain classical H-H inequality for harmonically convex function.

Now in next results, we will establish some H-H type inequalities for the products of two harmonically convex fuzzy-IVFs involving fuzzy-interval Riemann–Liouville fractional integral. These inequalities about harmonically convex fuzzy-IVFs are analogous generalization for some classical results provided by Noor [7], and Chen [6,13] for convex and generalized harmonically convex functions.

Theorem 3.6.

Let Q˜,P˜:u,νF0 be two harmonically convex fuzzy-IVFs on u,ν, whose β levels Qβ,Pβ:u,νKC+ are defined by Qβz=Qz,β,Qz,β and Pβz=Pz,β,Pz,β for all zu,ν, β0,1. If Q˜×˜P˜Lu,ν,F0, then

Γα+12uννuαI1ν+αQ˜ψ1uטP˜ψ1u+I1uαQ˜ψ1νטP˜ψ1ν12αα+1α+2M˜u,ν+αα+1α+2N˜u,ν.
where M˜u,ν=Q˜uטP˜u+˜Q˜νטP˜ν,N˜u,ν=Q˜uטP˜ν+˜Q˜νטP˜u, and Mβu,ν=Mu,ν,β,Mu,ν,β and Nβu,ν=Nu,ν,β,Nu,ν,β.

Proof.

Since Q˜,P˜ both are harmonically convex fuzzy-IVFs then, for each β0,1 we have

Quνϱu+1ϱν,β1ϱQu,β+ϱQν,β
and
Puνϱu+1ϱν,β1ϱPu,β+ϱPν,β.

From the definition of harmonically convex fuzzy-IVFs it follows that 0˜Q˜z and 0˜P˜z, so

Quνϱu+1ϱν,β×Puνϱu+1ϱν,β1ϱQu,β+ϱQν,β1ϱPu,β+ϱPν,β=1ϱ2Qu,β×Pu,β+ϱ2Qν,β×Pν,β+ϱ1ϱQu,β×Pν,β+ϱ1ϱQν,β×Pu,β(49)

Analogously, we have

Quν1ϱu+ϱν,βPuν1ϱu+ϱν,βϱ2Qu,β×Pu,β+1ϱ2Qν,β×Pν,β+ϱ1ϱQu,β×Pν,β+ϱ1ϱQν,β×Pu,β(50)

Adding (49) and (50), we have

Quνϱu+1ϱν,β×Puνϱu+1ϱν,β+Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,βϱ2+1ϱ2Qu,β×Pu,β+Qν,β×Pν,β+2ϱ1ϱQν,β×Pu,β+Qu,β×Pν,β(51)

Taking multiplication of (51) by ϱα1 and integrating the obtained result with respect to ϱ over (0, 1), we have

01ϱα1Quνϱu+1ϱν,β×Puνϱu+1ϱν,βdϱ+01ϱα1Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,βdϱMu,ν,β01ϱα1ϱ2+1ϱ2dϱ+2Nu,ν,β01ϱα1ϱ1ϱdϱ.

It follows that,

ΓαuννuαI1ν+αQ1u,β×P1u,β+I1uαQ1ν,β×P1ν,β2α12αα+1α+2Mu,ν,β+2ααα+1α+2Nu,ν,β

Similarly, for Qz,γ, we have

ΓαuννuαI1ν+αQ1u,β×P1u,β+I1uαQ1ν,β×P1ν,β2α12αα+1α+2Mu,ν,β+2ααα+1α+2Nu,ν,β,
that is,
ΓαuννuαI1ν+αQ1u,β×P1u,β+I1uαQ1ν,β×P1ν,β,I1ν+αQ1u,β×P1u,β+I1uαQ1ν,β×P1ν,βI2α12αα+1α+2Mu,ν,β,Mu,ν,β+2ααα+1α+2Nu,ν,β,Nu,ν,β.

Thus,

Γα+12uννuαI1ν+αQ˜ψ1uטP˜ψ1u+I1uαQ˜ψ1νטP˜ψ1ν12αα+1α+2M˜u,ν+αα+1α+2N˜u,ν.
and the theorem has been established.

Theorem 3.7.

Let Q˜,P˜:u,νF0 be two harmonically convex fuzzy-IVFs, whose β levels define the family of IVFs Qβ,Pβ:u,νKC+ are given by Qβz=Qz,β,Qz,β and Pβz=Pz,β,Pz,β for all zu,ν, β0,1. If Q˜×˜P˜Lu,ν,F0, then

Q˜2uνu+νטP˜2uνu+νΓα+14uννuαI1ν+αQ˜1uטP˜1u+I1uαQ˜1νטP˜1ν+1212αα+1α+2N˜u,ν+12αα+1α+2M˜u,ν.
where M˜u,ν=Q˜uטP˜u+˜Q˜νטP˜ν,N˜u,ν=Q˜uטP˜ν+˜Q˜νטP˜u, and Mβu,ν=Mu,ν,β,Mu,ν,β and Nβu,ν=Nu,ν,β,Nu,ν,β.

Proof.

Consider Q˜,P˜:u,νF0 are harmonically convex fuzzy-IVFs. Then by hypothesis, for each β0,1, we have

Q2uνu+ν,β×P2uνu+ν,β14Quνϱu+1ϱν,β×Puνϱu+1ϱν,β+Quνϱu+1ϱν,β×Puν1ϱu+ϱν,β+14Quν1ϱu+ϱν,β×Puνϱu+1ϱν,β+Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,β,14Quνϱu+1ϱν,β×Puνϱu+1ϱν,β+Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,β+14ϱQu,β+1ϱQν,β×1ϱPu,β+ϱPν,β+1ϱQu,β+ϱQν,β×ϱPu,β+1ϱPν,β,=14Quνϱu+1ϱν,β×Puνϱu+1ϱν,β+Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,β+14ϱ2+1ϱ2Nu,ν,β+ϱ1ϱ+1ϱϱMu,ν,β(52)

Multiplying inequality (52) by ϱα1 and integrating over 0,1,

Q2uνu+ν,β×P2uνu+ν,β1401ϱα1Quνϱu+1ϱν,β×Puνϱu+1ϱν,βdϱ+01ϱα1Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,βdϱ+14Nu,ν,β01ϱα1ϱ2+1ϱ2dϱ+2Mu,ν,β01ϱα1ϱ1ϱdϱ

Taking z=uνϱu+1ϱν and y=uν1ϱu+ϱν, then we get

1αQ2uνu+ν,β×P2uνu+ν,βΓα4uννuαI1ν+αQψ1u×Pψ1u+I1uαQψ1ν×Pψ1ν+12α12αα+1α+2Nu,ν,β+12ααα+1α+2Mu,ν,β,
1αQ2uνu+ν,β×P2uνu+ν,βΓα4uννuαI1ν+αQψ1u×Pψ1u++I1uαQψ1ν,β×Pψ1ν,β+12α12αα+1α+2Nu,ν,β+12ααα+1α+2Mu,ν,β,

Similarly, for Qz,γ, we have

1αQ2uνu+ν,β×P2uνu+ν,βΓα4uννuαI1ν+αQψ1u×Pψ1u+I1uαQψ1ν,β×Pψ1ν,β+12α12αα+1α+2Nu,ν,β+12ααα+1α+2Mu,ν,β,
that is,
Q˜2uνu+νטP˜2uνu+νΓα+14uννuαI1ν+αQ˜1uטP˜1u+I1uαQ˜1νטP˜1ν+1212αα+1α+2N˜u,ν+12αα+1α+2M˜u,ν.

Hence, the required result.

Theorem 3.8.

Let Q˜,P˜:u,νF0 be two harmonically convex fuzzy-IVFs, whose β levels define the family of IVFs Qβ,Pβ:u,νKC+ are given by Qβz=Qz,β,Qz,β and Pβz=Pz,β,Pz,β for all zu,ν, β0,1. If Q˜×˜P˜Lu,ν,F0, then

2Q˜2uνu+νטP˜2uνu+νΓα+121αuννuαIu+ν2uν+αQ˜ψ1uטP˜ψ1u+Iu+ν2uναQ˜ψ1νטP˜ψ1ν+12α2+3α4α+1α+2N˜u,ν+α2+3α4α+1α+2M˜u,ν.
where M˜u,ν=Q˜uטP˜u+˜Q˜νטP˜ν,N˜u,ν=Q˜uטP˜ν+˜Q˜νטP˜u, and Mβu,ν=Mu,ν,β,Mu,ν,β and Nβu,ν=Nu,ν,β,Nu,ν,β.

Proof.

Consider Q˜,P˜:u,νF0 are harmonically convex fuzzy-IVFs. Then by hypothesis, for each β0,1, we have

Q2uνu+ν,β×P2uνu+ν,β14Quνϱu+1ϱν,β×Puνϱu+1ϱν,β+Quνϱu+1ϱν,β×Puν1ϱu+ϱν,β+14Quν1ϱu+ϱν,β×Puνϱu+1ϱν,β+Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,β,14Quνϱu+1ϱν,β×Puνϱu+1ϱν,β+Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,β+14ϱQu,β+1ϱQν,β×1ϱPu,β+ϱPν,β+1ϱQu,β+ϱQν,β×ϱPu,β+1ϱPν,β,=14Quνϱu+1ϱν,β×Puνϱu+1ϱν,β+Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,β+14ϱ2+1ϱ2Nu,ν,β+2ϱ1ϱMu,ν,β(53)

Multiplying inequality (53) by 21+ααϱα1 and then integrating the obtain outcome over 0,12,

Q2uνu+ν,β×P2uνu+ν,β1401221+ααϱα1Quνϱu+1ϱν,β×Puνϱu+1ϱν,β+Quν1ϱu+ϱν,β×Puν1ϱu+ϱν,βdϱ.+14Nu,ν,β01221+ααϱα1ϱ2+1ϱ2dϱ+2Mu,ν,β01221+ααϱα1ϱ1ϱdϱ

Taking z=uνϱu+1ϱν and y=uν1ϱu+ϱν, then we get

2Q2uνu+ν,β×P2uνu+ν,βΓα+121αuννuαI1ν+αQψ1u×Pψ1u+I1uαQψ1ν×Pψ1ν+12αα+1α+2Nu,ν,β+αα+1α+2Mu,ν,β

Similarly, for Qz,γ, we have

2Q2uνu+ν,β×P2uνu+ν,βΓα+121αuννuαI1ν+αQψ1u×Pψ1u+I1uαQψ1ν×Pψ1ν,β+12α2+3α4α+1α+2Nu,ν,β+α2+3α4α+1α+2Mu,ν,β,
that is,
2Q˜2uνu+νטP˜2uνu+νΓα+121αuννuαIu+ν2uν+αQ˜ψ1uטP˜ψ1u+˜Iu+ν2uναQ˜ψ1νטP˜ψ1ν+12α2+3α4α+1α+2N˜u,ν+α2+3α4α+1α+2M˜u,ν.

4. CONCLUSION AND FUTURE STUDY

In this study, firstly we introduced the class of harmonically convex fuzzy-IVFs by means of fuzzy-order relation. Then we established H-H and H-H Fejér type inequalities for convex fuzzy-IVFs involving fuzzy Riemann–Liouville fractional integrals, and H-H inequalities are true for this concept of harmonically convex fuzzy-IVFs. As a future research, we try to explore this concept for generalized harmonically convex fuzzy-IVFs and some applications in fuzzy-interval nonlinear programing. By using this concept, the new direction of study can be found in optimization theory and convex analysis. We hope that this concept will be helpful for other authors to pay their roles in different fields of sciences.

AVAILABILITY OF DATA AND MATERIALS

Not applicable.

CONFLICTS OF INTEREST

The authors declare that they have no competing interests.

AUTHORS' CONTRIBUTIONS

All authors contributed equally to the writing of this paper. All authors read and approved the final manuscript.

Funding Statement

The research is supported by the National Natural Science Foundation of China (Grant No. 61673169).

ACKNOWLEDGMENTS

The authors would like to thank the Rector, COMSATS University Islamabad, Islamabad, Pakistan, for providing excellent research and academic environments.

REFERENCES

1.J. Hadamard, Étude sur les propriétés des fonctions entières et en particulier d'une fonction considérée par Riemann, J. Math. Pures Appl., Vol. 7, 1893, pp. 171-215.
2.C. Hermite, Sur deux limites d'une intégrale définie, Mathesis, Vol. 3, 1883, pp. 82-97.
3.I. Iscan, Hermite-Hadamard type inequalities for harmonically convex functions, Hacet. J. Math. Stat., Vol. 43, 2014, pp. 935-942.
5.M. Kunt, I. Iscan, N. Yazici, and U. Gözütok, Hermite-Hadamard type inequalities for product of harmonically convex functions via Riemann-Liouville fractional integrals, J. Math. Anal., Vol. 7, 2016, pp. 74-82.
7.M.A. Noor, K.I. Noor, M.U. Awan, and S. Costache, Some integral inequalities for harmonically h-convex functions, U. P. B. Sci. Bull. Ser. A, Vol. 77, 2015, pp. 5-16.
12.F. Chen, A note on Hermite-Hadamard inequalities for products of convex functions via Riemann-Liouville fractional integrals, Italian J. Pure Appl. Math., Vol. 33, 2014, pp. 299-306.
16.R.E. Moore, Interval Analysis, Prentice Hall, Englewood Cliffs, NJ, USA, 1966.
21.Y. Chalco-Cano, A. Flores-Franulic, and H. Román-Flores, Ostrowski type inequalities for interval-valued functions using generalized Hukuhara derivative, Comput. Appl. Math., Vol. 31, 2012, pp. 457-472.
25.K. Nikodem, J.L. Snchez, and L. Snchez, Jensen and Hermite-Hadamard inequalities for strongly convex set-valued maps, Math. Aterna, Vol. 4, 2014, pp. 979-987.
32.U. Kulish and W. Miranker, Computer Arithmetic in Theory and Practice, Academic Press, New York, NY, USA, 2014.
Journal
International Journal of Computational Intelligence Systems
Volume-Issue
14 - 1
Pages
1809 - 1822
Publication Date
2021/06/28
ISSN (Online)
1875-6883
ISSN (Print)
1875-6891
DOI
10.2991/ijcis.d.210620.001How to use a DOI?
Copyright
© 2021 The Authors. Published by Atlantis Press B.V.
Open Access
This is an open access article distributed under the CC BY-NC 4.0 license (http://creativecommons.org/licenses/by-nc/4.0/).

Cite this article

TY  - JOUR
AU  - Gul Sana
AU  - Muhammad Bilal Khan
AU  - Muhammad Aslam Noor
AU  - Pshtiwan Othman Mohammed
AU  - Yu-Ming Chu
PY  - 2021
DA  - 2021/06/28
TI  - Harmonically Convex Fuzzy-Interval-Valued Functions and Fuzzy-Interval Riemann–Liouville Fractional Integral Inequalities
JO  - International Journal of Computational Intelligence Systems
SP  - 1809
EP  - 1822
VL  - 14
IS  - 1
SN  - 1875-6883
UR  - https://doi.org/10.2991/ijcis.d.210620.001
DO  - 10.2991/ijcis.d.210620.001
ID  - Sana2021
ER  -