AAA Abstract and Applied Analysis 1687-0409 1085-3375 Hindawi Publishing Corporation 589694 10.1155/2012/589694 589694 Research Article Duality for Multitime Multiobjective Ratio Variational Problems on First Order Jet Bundle Postolache Mihai Peterson Allan Department of Mathematics and Informatics, Faculty of Applied Sciences University Politehnica of Bucharest Splaiul Independenţei No. 313 060042 Bucharest Romania pub.ro 2012 26 8 2012 2012 27 04 2012 29 06 2012 2012 Copyright © 2012 Mihai Postolache. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

We consider a new class of multitime multiobjective variational problems of minimizing a vector of quotients of functionals of curvilinear integral type. Based on the efficiency conditions for multitime multiobjective ratio variational problems, we introduce a ratio dual of generalized Mond-Weir-Zalmai type, and under some assumptions of generalized convexity, duality theorems are stated. We prove our weak duality theorem for efficient solutions, showing that the value of the objective function of the primal cannot exceed the value of the dual. Direct and converse duality theorems are stated, underlying the connections between the values of the objective functions of the primal and dual programs. As special cases, duality results of Mond-Weir-Zalmai type for a multitime multiobjective variational problem are obtained. This work further develops our studies in (Pitea and Postolache (2011)).

1. Introduction

The duality for scalar variational problems involving convex functions has been formulated by Mond and Hanson in . This study was further developed to various classes of convex and generalized convex functions. In , Hanson extended the duality results (in the sense of Wolfe) to a class of functions subsequently called invex. In order to weaken the convexity conditions, Bector et al. , introduced a dual to a variational problem (in the sense of Mond and Weir), different from that formulated by Mond and Hanson in . Mond et al. , extended the concept of invexity to continuous functions and used it to generalize earlier Wolfe duality results for a class of variational problems. Duality theory for scalar optimization can be found in many other works and see Mond and Husain  and Preda .

In , using a vector-valued Lagrangian function, Tanino and Sawaragi introduced a duality theory for multiobjective optimization. This idea is employed by Bitran , which associated a matrix of dual variables with the constraints in the primal problem. In , Mond and Weir considered a pair of symmetric dual nonlinear programs and developed a duality theory under assumptions of pseudoconvexity. Under various types of generalized convex functions, Mukherjee and Purnachandra , Preda  and Zalmai  established several weak efficiency conditions and developed different types of dualities for multiobjective variational problems. D. S. Kim and A. L. Kim  used the efficiency property of nondifferentiable multiobjective variational problems in duality theory. In a recent study , Pitea and Postolache considered a new class of multitime multiobjective variational problems of minimizing a vector of functionals of curvilinear integral type. Based on appropriate normal efficiency conditions, they studied duals of Mond-Weir type, generalized Mond-Weir-Zalmai type, and under appropriate assumptions of generalized convexity, stated duality theorems.

In time, several authors have been interested in the study of (vector) ratio programs in connection with generalized convexity. This study is motivated by many practical optimization problems whose objective functions are quotients of two functions. In , Jagannathan introduced a duality study using some results connecting solutions of a nonlinear ratio program with those of suitably defined parametric convex program. Concerning advances on single-objective ratio programs, see  by Khan and Hanson and  by Reddy and Mukherjee, which utilized invexity assumptions in the sense of Hanson  to obtain optimality conditions and duality results. As concerns vector ratio problems, Singh and Hanson  applied invex functions to derive duality results, while Jeyakumar and Mond  generalized these results to the class of V-invex functions. Later, Liang et al.  introduced a unified formulation of the generalized convexity to derive optimality conditions and duality results for vector ratio problems.

In this paper, we are motivated by previous published research articles (please refer to  by Antczak,  by Nahak and Mohapatra, and  by Pitea and Postolache) to consider a new class of multitime multiobjective variational problems of minimizing a vector of quotients of functionals of curvilinear integral type and, as a particular case, a vector of functionals of curvilinear integral type. We state and prove new duality results, of Mond-Weir-Zalmai type, for feasible solutions of our multitime multiobjective ratio variational problems, under assumptions of (ρ,b)-quasi-invexity. This study is encouraged by its possible application in mechanical engineering, where curvilinear integral objectives are extensively used due to their physical meaning as mechanical work. The objective vector function is of curvilinear integral type, the integrand depending on velocities; that is why we consider it adequate to introduce our results within the framework offered by the first order jet bundle,  by Pitea et al.,  by Sounders, and  by Udriste et al.

2. Preliminaries

Before presenting our results, we need the following background, which is necessary for the completeness of the exposition. For more details, we address the reader to .

2.1. The First Order Jet Bundle

To make our presentation self-contained and reader-friendly, we recall the notion of jet bundle of the first order, J1(T,M).

Let be given the smooth manifolds (M,g) and (T,h), of dimensions n, and p, respectively. The corresponding local coordinates are x=(xi), i=1,n¯, and t=(tα), α=1,p¯, respectively.

In the following, the set {1,2,,n} will be indexed by Latin characters, while the set {1,2,,p} will be indexed by Greek characters.

Denote by C(T,M) the set of mappings of class C, from T to M.

Consider (t0,x0) an arbitrary point of the product manifold T×M. On the set C(T,M), define the equivalence relation (2.1)f~(t0,x0)g{f(t0)=g(t0)=x0dft0=dgt0.

If f and g are arbitrary mappings from C(T,M), we denote (2.2)tβ(t0)=t0β,β=1,p¯,xi=xif,yi=xig,xi(x0)=x0i,  i=1,n¯.

Thus, the equivalence relation f~(t0,x0)g has the local expression (2.3)xi(t0β)=yi(t0β)=x0i,xitα(t0β)=yitα(t0β),i=1,n¯,α=1,p.¯

The equivalence class of a mapping fC(T,M) will be denoted by (2.4)[f](t0,x0)=  {gC(T,M)g~(t0,x0)f}. The quotient space obtained by the factorization of the space C(T,M) by the equivalence relation “~(t0,x0)”, (2.5)J(t0,x0)1(T,M)=C(T,M)~(t0,x0), is called 1-jet at the point (t0,x0).

The total space of the set of 1-jets, (2.6)J1(T,M)=(t0,x0)T×MJ(t0,x0)1(T,M), can be organized as a vector bundle over the base space T×M, endowed with the differentiable structure of the product space.

Let ς be the canonical projection, defined as (2.7)ς:J1(T,M)T×M,ς([f](t0,x0))=(t0,f(t0)).

The mapping ς is well defined, having the property to be onto.

For every local chart U×V of the product manifold T×M, we define the bijection (2.8)ϕU×V:ς-1(U×V)U×V×np,ϕU×V([f](t0,x0))=(t0,x0,xitα(t0β)),  x0=f(t0).

Therefore, the 1-jet space is endowed with a differentiable structure of dimension n+p+np, such that the mappings ϕU×V are diffeomorphisms.

The local coordinates on the space J1(T,M) are (tα,xi,xαi), where (2.9)tα([f](t0,x0))=tα(t0),xi([f](t0,x0))=xi(x0),xαi([f](t0,x0))=xitα(t0β),α=1,p¯,i=1,n¯.

Remark 2.1.

From physical viewpoint, the differentiable manifold T should be thought as a “temporal manifold" or a “multitime", while M is a “space manifold."

Remark 2.2.

From geometrical viewpoint, an element [f] of a fiber 1-jets J1(T,M) should be thought as a class of parametrized sheet. The sections of fiber 1-jets are “physical fields."

To simplify the notations, denote by xγ(t)=(x/tγ)(t), γ=1,p¯, the partial velocities. Also, in our subsequent theory, we will set πx(t)=(t,x(t),xγ(t)).

2.2. On Lagrange 1-Forms

A Lagrange 1-form of the first order on the jets space J1(T,M) has the form (2.10)ω=Lα(πx(t))dtα+Mi(πx(t))dxi+Niβ(πx(t))dxβi, where Lα, Mi and Niβ are Lagrangians of the first order. The pullback (2.11)x*ω=(Lα+Mixαi+Niβxβαi)dtα is a Lagrange 1-form of the second order on M. The coefficients Lα+Mixαi+Niβxβαi are second order Lagrangians, which are linear in the partial accelerations.

A smooth Lagrangian L(πx(t)), t+m produces two smooth closed (completely integrable) 1-forms:

(i) the differential (2.12)dL=Lxidxi+Lxγidxγi+Ltγdtγ  of components (L/tγ,L/xi,L/xγi), with respect to the basis (dtγ,dxi,dxγi);

(ii) the restriction of dL to πx(t), that is, the pullback (2.13)dL|πx(t)=(Lxixitβ+Lxγixγitβ+Ltβ)dtβ of components containing partial accelerations (2.14)DβL=Lxi(πx(t))xitβ(t)+Lxγi(πx(t))xγitβ(t)+Ltβ(πx(t)), with respect to the basis dtβ (for other ideas, see ).

3. Problem Description

Let (T,h) and (M,g) be Riemannian manifolds of dimensions p and n, respectively. Denote by t=(tα), α=1,p¯, and x=(xi), i=1,n¯, the local coordinates on T and M, respectively. Consider J1(T,M) the first order jet bundle associated to T and M.

To develop our theory, we recall the following relations between two vectors v=(vj) and w=(wj), j=1,δ¯: (3.1)v=wvj=wj,j=1,δ¯,v<wvj<wj,j=1,δ¯  ,vwvjwj,j=1,δ¯          (product  order  relation),vwvw,  vw.

Using the product order relation on p, the hyperparallelepiped Ωt0,t1, in p, with diagonal opposite points t0=(t01,,t0p) and t1=(t11,,t1p), can be written as being the interval [t0,t1]. Suppose γt0,t1 is a piecewise C1-class curve joining the points t0 and t1.

The closed Lagrange 1-forms densities of C-class (3.2)fα=(fα):J1(T,M)r,  kα=(kα):J1(T,M)r,=1,r¯,  α=1,p¯, determine the following path independent curvilinear functionals (actions, mechanical work): (3.3)F(x(·))=γt0,t1fα(πx(t))  dtα,K(x(·))=γt0,t1kα(πx(t))  dtα.

The closeness conditions (complete integrability conditions) are (3.4)Dβfα=Dαfβ,Dβkα=Dαkβ,α,β=1,p¯,αβ,=1,r¯, where Dβ is the total derivative.

Suppose K(x(·))>0, for all =1,r¯, and accept that the Lagrange matrix densities (3.5)g=(gab):J1(T,M)q¯s,a=1,s¯,  b=1,q¯¯,  q¯<n,h=(hab):J1(T,M)q~s,a=1,s¯,b=1,q~¯,q~<n, of C-class define the partial differential inequations (PDIs) (of evolution) (3.6)g(πx(t))0,tΩt0,t1, and the partial differential equations (PDE) (of evolution) (3.7)h(πx(t))=0,tΩt0,t1.

On the set C(Ωt0,t1,M) of all functions x:Ωt0,t1M of C-class, we set the norm x=x+α=1pxα.

For each =1,r¯, suppose K(x(·))>0, and consider (3.8)F(x(·))K(x(·))=(F1(x(·))K1(x(·)),,Fr(x(·))Kr(x(·))).

The aim of this work is to introduce and study the variational problem of minimizing a vector of quotients of functionals of curvilinear integral type: (MFP)minx(·)(F(x(·))K(x(·)))subject  tox(·)(Ωt0,t1), where (Ωt0,t1) denotes the set of all feasible solutions of problem (MFP), defined as (3.9)(Ωt0,t1)={xC(Ωt0,t1,M)  |  x(t0)=x0,x(t1)=x1,or  x(t)|Ωt0,t1=χ=given,g(πx(t))0,h(πx(t))=0,tΩt0,t1|  x(t0)=x0,x(t1)=x1,or  x(t)|Ωt0,t1}.

This kind of problems, of considerable interest, arises in various branches of mathematical, engineering, and economical sciences. We especially have in mind mechanical engineering, where curvilinear integral objectives are extensively used due to their physical meaning as mechanical work. These objectives play an essential role in mathematical modeling of certain processes in relation with robotics, tribology, engines, and much more.

4. Main Results

The following two definitions are crucial in developing our results. For more details, see  by Pitea et al., and , by Pitea and Postolache.

Definition 4.1.

A feasible solution x(·)(Ωt0,t1) is called an efficient solution of (MFP) if there is no x(·)(Ωt0,t1), x(·)x(·), such that F(x(·))/K(x(·))F(x(·))/K(x(·)).

Definition 4.2.

Let ρ be a real number and let b:C(Ωt0,t1,M)×C(Ωt0,t1,M)[0,) be a functional. To any closed 1-form a=(aα) we associate the path independent curvilinear functional (4.1)A(x(·))=γt0,t1aα(πx(t))dtα.

The functional A is called [strictly] (ρ,b)-quasi-invex at the point x(·) if there is a vector function η:J1(Ωt0,t1,M)×J1(Ωt0,t1,M)n, vanishing at the point (πx(t),πx(t)), and θ:C(Ωt0,t1,M)×C(Ωt0,t1,M)n, such that for any x(·) in C(Ωt0,t1,M), [x(·)x(·)], the following implication holds: (4.2)(A(x(·))A(x(·)))(b(x(·),x(·))γt0,t1[η(πx(t),πx(t)),aαx(πx(t))x((·),x(·))γt0,t1+Dγη(πx(t),πx(t)),aαxγ(πx(t))]dtα[<]-ρb(x(·),x(·))θ(x(·),x(·))2γt0,t1).

Several examples which illustrate our concept could be found in . However, the following example is interesting. It is a generalization of Example 1 in .

Example 4.3.

Let a:[0,1]×C([0,1])×C([0,1]), x(·)=(x1(·),x2(·),,xn(·)). With x(·)C([0,1]), denote (4.3)G(t)=ax(t,x(t),x˙(t))-Dax˙(t,x(t),x˙(t)),ax˙(t,x(t),x˙(t)),t[0,1], where by D we denoted the total derivative operator.

Suppose that G(1)-G(0)=0. Then, the functional (4.4)A(x(·))=01a(t,x(t),x˙(t))dt is (ρ,1)-quasi-invex, for ρ0 and any θ, at the point x(·), with respect to (4.5)η(πx(t),πx(t))=(A(x(·))-A(x(·)))(ax(t,x(t),x˙(t))-Dax˙(t,x(t),x˙(t))).

In their recent work , Pitea et al. established necessary efficiency conditions for problem (MFP). More accurately, they proved that if x(·) is an efficient solution of problem (MFP), then there are two vectors Λ10, Λ20 in r and the smooth functions μ and ν, the first from Ωt0,t1 to q¯sp, and the second from Ωt0,t1 to q~sp, such that (4.6)Λ10,fαx(πx(t))-Λ20,kαx(πx(t))+μα(t),gx(πx(t))+να(t),hx(πx(t))-Dγ(Λ10,fαxγ(πx(t))-Λ20,kαxγ(πx(t))DγDγ+μα(πx(t)),gxγ(πx(t))+να(t),hxγ(πx(t)))Dγ=0,tΩt0,t1,  α=1,p¯          (Euler-Lagrange  PDEs)(4.7)μα(t),g(πx(t))=0,μα(t)0,tΩt0,t1,α=1,p¯.

If Λ100 and Λ200, then x(·), from conditions (4.6), is called normal efficient solution.

Let x(·) be an efficient solution of primal (MFP), the scalars Λ10, Λ20 in r, and the smooth functions μ:Ωt0,t1q¯sp, ν:Ωt0,t1q~sp, given previously.

In order to use the idea of "grouping the resources," consider {P0,P1,,Pq} and {Q0,Q1,,Qq} partitions of the sets {1,,q¯} and {1,,q~}, respectively.

For each =1,r¯ and α=1,p¯, we denote (4.8)f¯α(πy(t))=fα(πy(t))+μαP0(t),gP0(πy(t))+ναQ0(t),hQ0(πy(t)),F¯(y(·))  =γt0,t1f¯α(πy(t))dtα.

Consider a function y(·)C(Ωt0,t1,M) and associate to (MFP) the multiobjective ratio variational problem (MFZD)maxy(·)(F¯1(y(·))K1(y(·)),,F¯r(y(·))Kr(y(·)))subject    toΛ10,fαy(πy(t))-Λ20,kαy(πy(t))+μα(t),gy(πy(t))+να(t),hy(πy(t))-Dγ(Λ10,fαyγ(πy(t))-Λ20kαyγ(πy(t))+μα(t),gyγ(πy(t))+να(t),hyγ(πy(t)))=0,α=1,p¯,tΩt0,t1,(4.9)μαPκ(t),gPκ(πy(t))+ναQκ(t),hQκ(πy(t))0,κ=1,q¯,α=1,p¯,tΩt0,t1Λ100, taking into account that the function y(t) has to satisfy the boundary conditions y(t0)=x0, y(t1)=x1, or y(t)|Ωt0,t1=χ=given.

ϖ ( x ( · ) ) is the value of the objective function of problem (MFP) at x(·)(Ωt0,t1), and δ¯(y(·),yγ(·),Λ10,Λ20,μ(·),ν(·)) is the maximizing functional vector of dual problem (MFZD) at the point (y(·),yγ(·),Λ10,Λ20,μ(·),ν(·))Δ, where Δ is the domain of problem (MFZD).

We prove three duality results in a generalized sense of Mond-Weir-Zalmai.

Theorem 4.4 (weak duality).

Let x(·) be a feasible solution of problem (MFP) and let (y(·),yγ(·),Λ10,Λ20,μ(·),ν(·)) be a feasible point of problem (MFZD). Assume that the following conditions are satisfied:

μP0(t),gP0(πy(t))+νQ0(t),hQ0(πy(t))0, for all tΩt0,t1;

Λ20>0 and Λ10F(y(·))-Λ20K(y(·))=0, for each =1,r¯;

for each =1,r¯, the functional Λ10F(x(·))-Λ20K(x(·)) is (ρ',b)-quasi-invex at the point y(·) with respect to η and θ, on (Ωt0,t1);

γt0,t1[μαPκ(t),gPκ(πx(t))+ναQκ(t),hQκ(πx(t))]dtα is (ρκ′′,b)-quasi-invex at y(·) with respect to η and θ, for each κ=1,q¯, on F(Ωt0,t1);

at least one of the functionals of (c), (d) is strictly quasi-invex;

=1rρ'+κ=1qρκ′′0.

Then, the inequality ϖ(x(·))δ¯(y(·),yγ(·),Λ10,Λ20,μ(·),ν(·)) is false.

Proof.

By reductio ad absurdum, suppose we have (4.10)F(x(·))K(x(·))F¯(y(·))K(y(·)),=1,r¯. From these inequalities, it follows (4.11)Λ10F(x(·))Λ20K(x(·))1+1F(y(·))γt0,t1[μαP0(t),gP0(πy(t))+ναQ0(t),hQ0(πy(t))]dtα, and taking into account the hypotheses (a) and (b), we get (4.12)Λ10F(x(·))Λ20K(x(·)),=1,r¯.

We obtain the following implications: (4.13)(Λ10,F(x(·))-Λ20,K(x(·))0)(=1rρ'b(x(·),y(·))×γt0,t1{Λ20,kαyγ(πy(t))η(πx(t),πy(t)),Λ10,fαy(πy(t))-Λ20,kαy(πy(t))+Dγη(πx(t),πy(t)),Λ10,fαyγ(πy(t))-Λ20,kαyγ(πy(t))}dtα-b(x(·),y(·))θ(x(·),y(·))2=1rρ').

Hypothesis (d) regarding the (ρκ′′,b)-quasi-invexity property of each functional implies (κ=1,q¯):(4.14)(γt0,t1[μαPκ(t),gPκ(πx(t))+ναQκ(t),hQκ(πx(t))]dtαγt0,t1[μαPκ(t),gPκ(πy(t))+ναQκ(t),hQκ(πy(t))]dtα)(γt0,t1b(x(·),y(·))×γt0,t1[η(πx(t),πy(t)),μαPκ(t),gPκy(πy(t))+ναQκ(t),hQκy(πy(t))+Dγη(πx(t),πy(t)),μαPκ(t),gPκyγ(πy(t))+ναQκ(t),hQκyγ(πy(t))]dtα-ρκ′′b(x(·),y(·))θ(x(·),y(·))2γt0,t1).

Now, we make the sum of implications (4.13) and (4.14) side by side and from κ=1 to κ=q. It follows (4.15)(γt0,t1Λ10,F(x(·))-Λ20,K(x(·))+γt0,t1[μα(t),g(πx(t))+να(t),h(πx(t))]dtα-γt0,t1[μα(t),g(πy(t))  +να(t),h(πy(t))]dtα0)((=1rρ'+κ=1qρκ′′)b(x(·),y(·))×γt0,t1[η(πx(t),πy(t)),Λ10,fαy(πy(t))-Λ20,kαy(πy(t))+μα(t),gy(πy(t))+να(t),hy(πy(t))  +Dγη(πx(t),πy(t)),Λ10,fαyγ(πy(t))-Λ20,kαyγ(πy(t))+μα(t),gyγ(πy(t))+να(t),hyγ(πy(t))]dtα-<b(x(·),y(·))θ(x(·),y(·))2(=1rρ'+κ=1qρκ′′)).

Since b(x(·),y(·))>0, we obtain (4.16)γt0,t1[η(πx(t),πy(t)),Vαy(πy(t),Λ10,Λ20,μ(·),ν(·))γt0,t1+Dγη(πx(t),πy(t)),Vαyγ(πy(t),Λ10,Λ20,μ(·),ν(·))]dtα<-θ(x(·),y(·))2(=1rρ'+κ=1qρκ′′), where (4.17)Vα(πy(·),Λ10,Λ20,μ(·),ν(·))=Λ10,fα(πy(t))-Λ20,kα(πy(t))Vα(πy(·),Λ10,Λ20,μ(·),ν(·))+μα(y(t)),g(πy(t))+να(t),h(πy(t)),tΩt0,t1,α=1,p¯.

The following relation holds: (4.18)Dγη(πx(t),πy(t)),Vαyγ(πy(t),Λ10,Λ20,μ(t),ν(t))=Dγη(πx(t),πy(t)),Vαyγ(πy(t),Λ10,Λ20,μ(t),ν(t))-η(πx(t),πy(t)),Dγ(Vαyγ)(πy(t),Λ10,Λ20,μ(t),ν(t)).

By replacing relations (4.18) and by using Euler-Lagrange PDE, relation (4.16) becomes (4.19)γt0,t1Dγη(πx(t),πy(t)),Vαyγ(πy(t),Λ10,Λ20,μ(t),ν(t))dtα<-θ(x(·),y(·))2(=1rρ'+κ=1qρκ′′).

For α,γ=1,p¯, let us denote by (4.20)Qαγ(t)=η(πx(t),πy(t)),Vαyγ(πy(t),Λ10,Λ20,μ(t),ν(t)),I=γt0,t1DγQαγ(t)  dtα.

According to , § 9, a total divergence is equal to a total derivative. Consequently, there exists Q(t), with Q(t0)=0 and Q(t1)=0 such that DγQαγ(t)=DαQ(t) and (4.21)I=γt0,t1DαQ(t)  dtα=Q(t1)-Q(t0)=0.

Replacing into inequality (4.19), it follows that (4.22)0<-θ(x(·),y(·))2(=1rρ'+a=1qρa′′), contradicting hypothesis (f).

From relation (4.15), it follows (4.23)0Λ10,F(x(·))-Λ20,K(x(·))+γt0,t1[μα(t),g(πx(t))+να(t),h(πx(t))]dtα-γt0,t1[μα(t),g(πy(t))+να(t),h(πy(t))]dtα. According to the constraints of problems (MFP) and (MFZD), the previously mentioned relation becomes Λ10F(x(·))-Λ20K(x(·))>0, that is: (4.24)K(x(·))K(y(·))[F(x(·))K(x(·))-F(y(·))K(y(·))]>0.

Because K(x(·))K(y(·))>0, =1,r¯, we conclude that (4.25)(F1(x(·))K1(x(·))-F1(y(·))K1(y(·)),,Fr(x(·))Kr(x(·))-Fr(y(·))Kr(y(·)))(0,,0), or (4.26)(F1(x(·))K1(x(·)),,Fr(x(·))Kr(x(·)))(F1(y(·))K1(y(·)),,Fr(y(·))Kr(y(·))).

Therefore, the inequality ϖ(x(·))δ¯(y(·),yγ(·),Λ10,Λ20,μ(·),ν(·)) contradicts relations (4.12), and this completes the proof.

Theorem 4.5 (direct duality).

Let x(·) be an efficient solution of primal (MFP). Suppose the hypotheses of Theorem 4.4 hold. Then there are Λ10 and Λ20 in r, and the smooth functions μ:Ωt0,t1qsp and ν:Ωt0,t1qsp, such that (x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)) is an efficient solution of dual program (MFZD). Moreover (4.27)ϖ(x(·))=δ¯(x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)).

Proof.

x ( · ) being efficient solution of primal (MFP), there are Λ10 and Λ20 in r and the smooth functions μ:Ωt0,t1qsp and ν:Ωt0,t1qsp, such that relations (4.6) are verified. Therefore, μα(t),g(πx(t))=0, α=1,p¯. It follows (x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)) is a feasible solution for program (MFZD). Obviously, (4.28)ϖ(x(·))=δ¯(x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)). The weak duality theorem assures the efficiency of x(·).

If we remark that the notion of efficient solution of problem (MFZD) is similar to those in Definition 4.1, we can state the result in the following.

Theorem 4.6 (converse duality).

Let (x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)) be an efficient solution to dual (MFZD) and suppose the following conditions are satisfied:

x¯(·) is a normal efficient solution of primal (MFP);

the hypotheses of Theorem 4.4 hold at (x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)).

Then x(·) is an efficient solution to (MFP). Moreover, one has the equality (4.29)ϖ(x(·))=δ¯(x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)).

Proof.

Suppose x¯(·)x(·). The efficiency of x¯(·) of primal (MFP) implies the existence of Λ¯10 and Λ¯20 in r and the smooth functions μ¯:Ωt0,t1qsp and ν¯:Ωt0,t1qsp, such that relations (4.6) are satisfied. It follows that (x¯(·),x¯γ(·),Λ¯10,Λ¯20,μ¯(·),ν¯(·)) is a feasible solution of (MFZD) and (4.30)ϖ(x¯(·))=δ¯(x¯(·),x¯γ(·),Λ¯10,Λ¯20,μ¯(·),ν¯(·)).

On the other hand, from the weak duality theorem, (4.31)ϖ(x¯(·))δ(x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)) holds, that is: (4.32)δ(x¯(·),x¯γ(·),Λ¯10,Λ¯20,μ¯(·),ν¯(·))δ(x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)).

This last relation contradicts the efficiency of (x(·),xγ(·),Λ10,Λ20,μ(·),ν(·)) for program (MFZD). Therefore x¯(·)=x(·), and the theorem is proved.

We would like to conclude the study in this section with the following important particular case.

In this respect, suppose that kα=positive  constant, =1,r¯, α=1,p¯. Denote F(x(·))=(F1(x(·)),,Fr(x(·))). Then, program (MFP) becomes (MP)          minF(x)subject  to        x(·)(Ωt0,t1), and its dual is (GMWD)max  abLe(πy(t))dtsubject  to  Λ10fαy(πy(t))+μα(t),gy(πy(t))+να(t),hy(πy(t))-Dγ(Λ10fαyγ(πy(t))+μα(t),gyγ(πy(t))+να(t),hyγ(πy(t)))=0,α=1,p¯,tΩt0,t1,μαPκ(t),gPκ(πy(t))+ναQκ(t),hQκ(πy(t))0,κ=1,q¯,α=1,p¯,tΩt0,t1,Λ100, where (4.33)Le(πy(t))=f(πy(t))+[μP0(t),gP0(πx(t))+νQ0(t),hQ0(πx(t))]e, with Le=(L1,,Lr).

It can be seen that we have obtained precisely the programs studied in the work . Therefore, the results in this paper are stronger than the ones mentioned before.

5. Conclusion

In our recent study , we initiated an optimization theory on the first order jet bundle. As natural continuation, in this paper we considered a new class of multitime multiobjective variational problems of minimizing a vector of quotients of functionals of curvilinear integral type. We derived duality results for efficient solutions of multitime multiobjective ratio variational problems under the assumptions of (ρ,b)-quasi-invexity. We proved our weak duality theorem for efficient solutions, showing that the value of the objective function of the primal cannot exceed the value of the dual. Direct and converse duality theorems are stated, underlying the connections between the values of the objective functions of the primal and dual programs. As special cases, duality results of Mond-Weir-Zalmai type for a multitime multiobjective variational problem are obtained.

Having in mind the physical significance of the objective function, this study is strongly motivated by its possible applications of nonlinear optimization to mechanical engineering and economics .

Mond B. Hanson M. A. Duality for variational problems Journal of Mathematical Analysis and Applications 1967 18 2 355 364 2-s2.0-0001080845 Hanson M. A. On sufficiency of the Kuhn-Tucker conditions Journal of Mathematical Analysis and Applications 1981 80 2 545 550 2-s2.0-49149139478 Bector C. R. Chandra S. Husain I. Generalized concavity and duality in continuous programming Utilitas Mathematica 1984 25 171 190 Mond B. Chandra S. Husain I. Duality for variational problems with invexity Journal of Mathematical Analysis and Applications 1988 134 2 322 328 2-s2.0-38249028979 Mond B. Husain I. Sufficient optimality criteria and duality for variational problems with generalized invexity Journal of the Australian Mathematical Society B 1989 31 1 108 121 Preda V. On Mond-Weir duality for variational problems Revue Roumaine de Mathématique Pures et Appliquées 1993 28 2 155 164 Tanino T. Sawaragi Y. Duality theory in multiobjective programming Journal of Optimization Theory and Applications 1979 27 4 509 529 2-s2.0-0018455378 10.1007/BF00933437 Bitran G. R. Duality for nonlinear multiple-criteria optimization problems Journal of Optimization Theory and Applications 1981 35 3 367 401 2-s2.0-0019632885 10.1007/BF00934908 Mond B. Weir T. Schaible S. Ziemba W. T. Generalized concavity and duality InGeneralized Convexity in Optimization and Economics 1981 New York, NY, USA Academic Press 263 280 Mukherjee R. N. Rao C. P. Mixed type duality for multiobjective variational problems Journal of Mathematical Analysis and Applications 2000 252 2 571 586 2-s2.0-0034670487 Preda V. Optimality and duality in fractional multiple objective programming involving semilocally preinvex and related functions Journal of Mathematical Analysis and Applications 2003 288 2 365 382 2-s2.0-0346973099 10.1016/S0022-247X(02)00460-2 Zalmai G. J. Generalized (F, b, ϕ, ρ, θ )-univex n-set functions and semiparametric duality models in multiobjective fractional subset programming International Journal of Mathematics and Mathematical Sciences 2005 2005 7 1109 1133 2-s2.0-27344446176 10.1155/IJMMS.2005.1109 Kim D. S. Kim A. L. Optimality and duality for nondifferentiable multiobjective variational problems Journal of Mathematical Analysis and Applications 2002 274 1 255 278 2-s2.0-0036815646 10.1016/S0022-247X(02)00298-6 Pitea A. Postolache M. Duality theorems for a new class of multitime multiobjective variational problems Journal of Global Optimization 2012 54 1 47 58 10.1007/s10898-011-9740-z Jagannathan R. Duality for nonlinear fractional programs Zeitschrift für Operations Research 1973 17 1 1 3 2-s2.0-0002870350 10.1007/BF01951364 Khan Z. A. Hanson M. A. On ratio invexity in mathematical programming Journal of Mathematical Analysis and Applications 1997 205 2 330 336 2-s2.0-0031567801 10.1006/jmaa.1997.5180 Reddy L. V. Mukherjee R. N. Some results on mathematical programming with generalized ratio invexity Journal of Mathematical Analysis and Applications 1999 240 2 299 310 2-s2.0-0007282009 10.1006/jmaa.1999.6334 Hanson M. A. A duality theorem in nonlinear programming with nonlinear constraints Australian Journal of Statistics 1961 3 64 72 Singh C. Hanson M. A. Multiobjective ratio programming duality theory Naval Research Logistics 1991 38 925 933 Jeyakumar V. Mond B. On generalized convex mathematical programming Journal of the Australian Mathematical Society B 1992 34 43 53 Liang Z. A. Huang H. X. Pardalos P. M. Efficiency conditions and duality for a class of multiobjective fractional programming problems Journal of Global Optimization 2003 27 4 447 471 2-s2.0-0242485178 10.1023/A:1026041403408 Antczak T. Modified ratio objective approach in mathematical programming Journal of Optimization Theory and Applications 2005 126 1 23 40 2-s2.0-23744486068 10.1007/s10957-005-2654-5 Nahak C. Mohapatra R. N. Nonsmooth ρ-(η, θ)-invexity in multiobjective programming problems Optimization Letters 2010 1 8 2-s2.0-77957731509 10.1007/s11590-010-0239-1 Pitea A. Integral Geometry and PDE Constrained Optimization Problems [Ph.D. thesis] 2008 Bucharest, Romania University “Politehnica” of Bucharest Pitea A. Udriste C. Mititelu S. New type dualities in PDI and PDE constrained optimization problems Journal of Advanced Mathematical Studies 2009 2 1 81 90 Pitea A. Udriste C. Mititelu S. PDI&PDE-constrained optimization problems with curvilinear functional quotients as objective vectors Balkan Journal of Geometry and Its Applications 2009 14 2 75 78 Sounders D. J. The Geometry of Jet Bundles 1989 Cambridge, UK Cambridge University Press Udriste C. Dogaru O. Ţevy I. Null Lagrangian forms and Euler-Lagrange PDEs Journal of Advanced Mathematical Studies 2008 1 1-2 143 156 2498896 Udriste C. Multitime maximum principle for curvilinear integral cost Balkan Journal of Geometry and Its Applications 2011 16 1 128 149 2-s2.0-79954531818