Next Article in Journal
Generalized Integral Inequalities of Chebyshev Type
Previous Article in Journal
Admissibility of Fractional Order Descriptor Systems Based on Complex Variables: An LMI Approach
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Homotopy Analysis Method to Solve Two-Dimensional Nonlinear Volterra-Fredholm Fuzzy Integral Equations

by
Atanaska Georgieva
1 and
Snezhana Hristova
2,*
1
Department of Mathematical Analysis, University of Plovdiv “Paisii Hilendarski”, 4000 Plovdiv, Bulgaria
2
Department of Applied Mathematics and Modeling, University of Plovdiv “Paisii Hilendarski”, 4000 Plovdiv, Bulgaria
*
Author to whom correspondence should be addressed.
Fractal Fract. 2020, 4(1), 9; https://doi.org/10.3390/fractalfract4010009
Submission received: 24 February 2020 / Revised: 13 March 2020 / Accepted: 13 March 2020 / Published: 18 March 2020

Abstract

:
The main goal of the paper is to present an approximate method for solving of a two-dimensional nonlinear Volterra-Fredholm fuzzy integral equation (2D-NVFFIE). It is applied the homotopy analysis method (HAM). The studied equation is converted to a nonlinear system of Volterra-Fredholm integral equations in a crisp case. Approximate solutions of this system are obtained by the help with HAM and hence an approximation for the fuzzy solution of the nonlinear Volterra-Fredholm fuzzy integral equation is presented. The convergence of the proposed method is proved and the error estimate between the exact and the approximate solution is obtained. The validity and applicability of the proposed method is illustrated on a numerical example.

1. Introduction

Fuzzy integral equations are one of the important branches of fuzzy analysis theory and they are applied as an adequate apparatus in mathematical modeling in biology, chemistry, physics, engineering, etc. (see, for example, [1,2,3,4]). One of the first applications of fuzzy integration was given by Wu and Ma [5], who investigated the fuzzy Fredholm integral equation of the second kind.
In connection with the application, it is very important to such kind of problems. If methods for solving such problems with uncertainty are developed, then, many real life models in different fields with imprecise variable can be solved easily and accurately. Some fixed point theorems for complete fuzzy metric space are given in [6,7,8]. In recent years, many mathematicians have studied a solution to fuzzy integral equations by numerical methods [9,10,11,12,13,14].
Liao employed the basic idea of the homotopy in topology to propose a general analytic method for nonlinear problems, namely HAM (see the monograph [15], and the papers [16,17,18]). This method is based on the concept of creating function series. If the series converges, its sum is the solution of this system of equations. Later, HAM has been successfully applied to solve many types of nonlinear problems such as multiple solutions of nonlinear boundary value problems ([19]), Abel fuzzy integral equations ([20]), partial differential equations ([21]), two-dimensional linear Volterra fuzzy integral equations ([22]), and fuzzy linear Volterra integral equations of the second kind ([23]).
The paper presents an application of HAM for solving the following nonlinear Volterra-Fredholm fuzzy integral equations with two variables (2D-NVFFIE)
u ( s , t ) = g ( s , t ) ( F R ) c t k 1 ( t , τ ) G 1 ( u ( s , τ ) ) d τ ( F R ) a b k 2 ( s , ξ ) G 2 ( u ( ξ , t ) ) d ξ ,
where g , u : A = [ a , b ] × [ c , d ] E 1 are continuous fuzzy-number valued functions, k 1 : [ c , d ] × [ c , d ] R + , k 2 : [ a , b ] × [ a , b ] R + and G 1 , G 2 : E 1 E 1 are continuous functions. By E 1 is denoted the set of all fuzzy numbers.
The integral Equation (1) is called the Fredholm equation with respect to the position and Volterra with respect to the time. This equation is used in many problems of mathematical physics, theory of elasticity, contact problems and mixed problems of mechanics of continuous media (see [2,24]). In [25], the Adomian decomposition method is applied for solving this equation.
The structure of this paper is organized as follows: In Section 2, some basic notations used in fuzzy calculus are introduced. In Section 3, we present HAM. In Section 4, we apply HAM for the parametric form of 2D-NVFFIE. In Section 5, we prove the convergence of the proposed method and we give an error estimate. In Section 6, a numerical example is illustrating the application of the presented above procedure for approximately solving of the studied equation.

2. Preliminaries

In this section, we review the fundamental notations of fuzzy set theory to be used throughout this paper.
Definition 1
([26]). A fuzzy number is a function u : R [ 0 , 1 ] satisfying the following properties:
(i)
u is upper semi-continuous on R ;
(ii)
u ( x ) = 0 outside of some interval [ c , d ] ;
(iii)
there are the real numbers a and b with c a b d , such that u is increasing on [ c , a ] , decreasing on [ b , d ] and u ( x ) = 1 for each x [ a , b ] ;
(iv)
u ( r x + ( 1 r ) y ) min { u ( x ) , u ( y ) } for any x , y R , r [ 0 , 1 ] .
By E 1 we denote the set of all fuzzy numbers. Any real number a R can be interpreted as a fuzzy number a ˜ = χ ( a ) and therefore R E 1 .
Denote R + = ( 0 , ) .
For any 0 < r 1 we denote the r-level set [ u ] r = { x R : u ( x ) r } that is a closed interval and [ u ] r = [ u r , u + r ] for all r [ 0 , 1 ] , where u , u + can be considered as functions u , u + : [ 0 , 1 ] R , such that u is increasing and u + is decreasing.
For u , v E 1 , k R the addition and the scalar multiplication are defined by [ u v ] r = [ u ] r + [ v ] r = [ u r + v r , u + r + v + r ] and [ k u ] r = k . [ u ] r = [ k u r , k u + r ] , k 0 [ k u + r , k u r ] , k < 0 .
The neutral element with respect to ⊕ in E 1 is denoted by 0 ˜ = χ { 0 } .
According to [27], we can summarize the following algebraic properties:
(i)
u ( v w ) = ( u v ) w and u v = v u for any u , v , w E 1 ;
(ii)
u 0 ˜ = 0 ˜ u = u for any u E 1 ;
(iii)
with respect to 0 ˜ , none u E 1 R , u 0 ˜ has opposite in ( E 1 , ) ;
(iv)
for any a , b R with a , b 0 or a , b 0 and any u E 1 we have ( a + b ) u = a u b u ;
(v)
for any a R and any u , v E 1 we have a ( u v ) = a u a v ;
(vi)
for any a , b R and any u E 1 we have a ( b u ) = ( a b ) u and 1 u = u .
As a distance between fuzzy numbers we use the Hausdorff metric.
Definition 2
([27]). For arbitrary fuzzy numbers u = ( u r , u + r ) and v = ( v r , v + r ) the quantity D ( u , v ) = sup r [ 0 , 1 ] max { | u r v r | , | u + r v + r | } is the distance between u, v.
Lemma 1
([27]). The following properties of the above distance hold:
(i)
( E 1 , D ) is a complete metric space;
(ii)
D ( u w , v w ) = D ( u , v ) , for all u , v , w E 1 ;
(iii)
D ( k u , k v ) = | k | D ( u , v ) , for all u , v E 1 , for all k R ;
(iv)
D ( u v , w e ) = D ( u , w ) + D ( v , e ) , for all u , v , w , e E 1 ;
(v)
D ( u v , 0 ˜ ) D ( u , 0 ˜ ) + D ( v , 0 ˜ ) , for all u , v E 1 ;
(vi)
D ( k 1 u , k 2 u ) = | k 1 k 2 | D ( u , 0 ˜ ) , for all k 1 , k 2 R with k 1 k 2 0 and u E 1 .
For any fuzzy-number-valued function f : I R E 1 we define the functions f ̲ ( . , r ) , f ¯ ( . , r ) : I R , by
f ̲ ( t , r ) = ( f ( t ) ) r and f ¯ ( t , r ) = ( f ( t ) ) + r for each t I , r [ 0 , 1 ] .
These functions are called the left and right r level functions of f .
We will use the notion of Henstock integral for fuzzy-number-valued functions defined as follows:
Let f : [ a , b ] E 1 . For Δ n : a = x 0 < x 1 < . . . < x n 1 < x n = b a partition of the interval [ a , b ] , we consider the points ξ i [ x i 1 , x i ] , i = 1 , . . . , n , and the function δ : [ a , b ] R + .
The partition P = { ( [ x i 1 , x i ] ; ξ i ) ; i = 1 , . . . , n } denoted by P = ( Δ n , ξ ) is called δ -fine iff [ x i 1 , x i ] ( ξ i δ ( ξ i ) , ξ i + δ ( ξ i ) ) .
Definition 3
([27]). For I E 1 , the function f is fuzzy-Henstock integrable on [ a , b ] if for any ε > 0 there is a function δ : [ a , b ] R + such that for any partition δ-fine P, D ( i = 1 n ( x i x i 1 ) f ( ξ i ) , I ) < ε . The fuzzy number I is named the fuzzy-Henstock integral of f and will be denoted by ( F H ) a b f ( t ) d t .
When the function δ : [ a , b ] R + is constant, then we obtain the Riemann integrability for fuzzy-number-valued functions ([26]). In this case, I E 1 is called the fuzzy-Riemann integral of f on [ a , b ] , being denoted by ( F R ) a b f ( t ) d t . Consequently, the fuzzy-Riemann integrability is a particular case of the fuzzy-Henstock integrability, and therefore the properties of the integral ( F H ) will be valid for the integral ( F R ) , too.
Lemma 2
([28]). Let f : [ a , b ] E 1 . Then f is ( F H ) integrable if and only if f ̲ ( . , r ) and f ¯ ( . , r ) are Henstock integrable for any r [ 0 , 1 ] . Furthermore, for any r [ 0 , 1 ] ,
( F H ) a b f ( t ) d t r = ( H ) a b f ̲ ( t , r ) d t , ( H ) a b f ¯ ( t , r ) d t .
Remark 1.
If f : [ a , b ] E 1 is continuous, then f ̲ ( . , r ) and f ¯ ( . , r ) are continuous for any r [ 0 , 1 ] and consequently, they are Henstock integrable. According to Lemma 1 we infer that f is ( F H ) integrable.
For any fuzzy-number-valued function f : A = [ a , b ] × [ c , d ] E 1 we can define the functions f ̲ ( . , . , r ) , f ¯ ( . , . , r ) : A R , by f ̲ ( s , t , r ) = ( f ( s , t ) ) r and f ¯ ( s , t , r ) = ( f ( s , t ) ) + r for all ( s , t ) A and r [ 0 , 1 ] . These functions are called the left and right r level functions of f.
Definition 4
([29]). A fuzzy-number-valued function f : A E 1 is called:
(i)
continuous in ( x 0 , y 0 ) A if for any ε > 0 there exists δ > 0 such that for any ( x , y ) A with | x x 0 | < δ , | y y 0 | < δ it follows that D ( f ( x , y ) , f ( x 0 , y 0 ) ) < ε .
(ii)
continuous on A if it is continuous in each ( x , y ) A ;
(iii)
bounded if there exists M 0 such that D ( f ( x , y ) , 0 ˜ ) M for all ( x , y ) A .
On the set C ( A , E 1 ) = { f : A E 1 ; f is continuous } the metric is defined by:
D * ( f , g ) = max ( s , t ) A D ( f ( s , t ) , g ( s , t ) ) for all f , g C ( A , E 1 ) .
We see that ( C ( A , E 1 ) , D * ) is a complete metric space.

3. The Homotopy Analysis Method

We will give a brief overview of the main used method—HAM. Homotopy analysis method transforms the considered equation into the corresponding deformation equation. Using this method, we solve the operator equation
N ( u ( z ) ) = 0 , z Ω ,
where N is the operator, u is the unknown function and Ω is any domain of the variable z.
Define the homotopy operator H as
H ( Φ , p ) ( 1 p ) L ( Φ ( z ; p ) u 0 ( z ) ) p h N ( Φ ( z ; p ) ) ,
where p [ 0 , 1 ] is an embedded parameter, h 0 denotes the convergence control parameter (see, for example [30,31]), u 0 represents the initial approximation of the solution of (2) and L is linear operator with property L ( 0 ) = 0 .
Solving equation H ( Φ , p ) = 0 , we get the zero-order deformation equation
( 1 p ) L ( Φ ( z ; p ) u 0 ( z ) ) = p h N ( Φ ( z ; p ) ) ,
Substitute p = 0 in (3) and obtain L ( Φ ( z ; 0 ) u 0 ( z ) ) = 0 . Therefore, Φ ( z ; 0 ) = u 0 ( z ) . If p = 1 , then N ( Φ ( z ; p ) ) = 0 , i.e., Φ ( z ; 1 ) is solution of the Equation (2). In this way, the change of parameter p from zero to one corresponds to the transition from a trivial task to the original task.
Taking the Maclaurin series of function Φ ( z ; p ) with respect to the parameter p, we obtain
Φ ( z ; p ) = u 0 ( z ) + m = 1 u m ( z ) p m ,
where u 0 ( z ) = Φ ( z ; 0 ) and
u m ( z ) = 1 m ! m Φ ( z ; p ) p m | p = 0 , m = 1 , 2 , 3 , . . . .
If the above series converges for p = 1 , we obtain the required solution
u ( z ) = m = 0 u m ( z ) .
In order to determine the function u m we differentiate m-times, with respect to parameter p, the left and right-hand side of Formula (3), then the obtained result is divided by m ! and substituted with p = 0 which gives the so-called mth-order deformation equation ( m 1 ) :
L ( u m ( z ) χ m u m 1 ( z ) ) = h R m ( u ˜ m 1 ( z ) ) ,
where u ˜ m 1 ( z ) = { u 0 ( z ) , u 1 ( z ) , . . . , u m 1 ( z ) } , and
χ m = 0 , m = 1 1 , m 2
and
R m ( u ˜ m 1 ( z ) ) = 1 ( m 1 ) ! m 1 N [ Φ ( z ; p ) ] p m 1 | p = 0 .
Apply L 1 to both sides of (7) and obtain
u m ( z ) = χ m u m 1 ( z ) ) + h L 1 ( R m ( u ˜ m 1 ( z ) ) ) .
If we are not able to determine the sum of the series in (6), then for the approximate solution of the considered equation we accept the partial sum of this series
s m ( z ) = i = 0 m u i ( z ) .
Choosing in an appropriate way the convergence control parameter h, we can influence the convergence region of the created series and the rate of this convergence ([15,32]). One of the methods to select the value of this parameter is the so-called “optimization method” ([15]). In this method, we define the squared residual of the governing equation
E n ( h ) = Ω ( N ( s n ( z ) ) 2 d z
where s n ( z ) and u i ( z ) are defined by (11) and (10) respectively and they depend on h.
The optimum value of the convergence control parameter is obtained by determining the minimum of this squared residual. The effective region of the convergence control parameter is additionally defined by
R h = { h : lim n E n ( h ) = 0 } .
Choosing a different value of the convergence control parameter than the optimal one, but still belonging to the effective region, we also obtain the convergent series, only the rate of convergence is lower. A version of the method with the above described selection of optimal value the convergence control parameter is called the basic optimal HAM ([15]).

4. Applying HAM to 2D-NVFFIE (1)

In this section we introduce the parametric form of the integral Equation (1) and then we will apply HAM for solving this equation.
Let u ( s , t , r ) = ( u ̲ ( s , t , r ) , u ¯ ( s , t , r ) ) and g ( s , t , r ) = ( g ̲ ( s , t , r ) , g ¯ ( s , t , r ) ) ; 0 r 1 and ( s , t ) A are parametric form of functions u ( s , t ) and g ( s , t ) , respectively. So the parametric form of Equation (1) is as follows:
u ̲ ( s , t , r ) = g ̲ ( s , t , r ) + c t k 1 ( t , τ ) G 1 ( u ( s , τ , r ) ) ̲ d τ + a b k 2 ( s , ξ ) G 2 ( u ( ξ , t , r ) ) ̲ d ξ , u ¯ ( s , t , r ) = g ¯ ( s , t , r ) + c t k 1 ( t , τ ) G 1 ( u ( s , τ , r ) ) ¯ d τ + a b k 2 ( s , ξ ) G 2 ( u ( ξ , t , r ) ) ¯ d ξ .
We define
H 1 ( u ̲ ( s , t , r ) , u ¯ ( s , t , r ) ) = m i n { G 1 ( β ) : u ̲ ( s , t , r ) β u ¯ ( s , t , r ) } , H 2 ( u ̲ ( s , t , r ) , u ¯ ( s , t , r ) ) = m i n { G 2 ( β ) : u ̲ ( s , t , r ) β u ¯ ( s , t , r ) } , F 1 ( u ̲ ( s , t , r ) , u ¯ ( s , t , r ) ) = m a x { G 1 ( β ) : u ̲ ( s , t , r ) β u ¯ ( s , t , r ) } , F 2 ( u ̲ ( s , t , r ) , u ¯ ( s , t , r ) ) = m a x { G 2 ( β ) : u ̲ ( s , t , r ) β u ¯ ( s , t , r ) } ,
where ( s , t ) A . Then,
k 1 ( t , τ ) G 1 ( u ( s , τ , r ) ) ̲ = k 1 ( t , τ ) H 1 ( u ̲ ( s , τ , r ) , u ¯ ( s , τ , r ) ) , if k 1 ( t , τ ) 0 k 1 ( t , τ ) F 1 ( u ̲ ( s , τ , r ) , u ¯ ( s , τ , r ) ) , if k 1 ( t , τ ) < 0 ,
k 2 ( s , ξ ) G 2 ( u ( ξ , t , r ) ) ̲ = k 2 ( s , ξ ) H 2 ( u ̲ ( ξ , t , r ) , u ¯ ( ξ , t , r ) ) , if k 2 ( s , ξ ) 0 k 2 ( s , ξ ) F 2 ( u ̲ ( ξ , t , r ) , u ¯ ( ξ , t , r ) ) , if k 2 ( s , ξ ) < 0 ,
k 1 ( t , τ ) G 1 ( u ( s , τ , r ) ) ¯ = k 1 ( t , τ ) F 1 ( u ̲ ( s , τ , r ) , u ¯ ( s , τ , r ) ) , if k 1 ( t , τ ) 0 k 1 ( t , τ ) H 1 ( u ̲ ( s , τ , r ) , u ¯ ( s , τ , r ) ) , if k ( t , τ ) < 0 ,
k 2 ( s , ξ ) G 2 ( u ( ξ , t , r ) ) ¯ = k 2 ( s , ξ ) F 2 ( u ̲ ( ξ , t , r ) , u ¯ ( ξ , t , r ) ) , if k 2 ( s , ξ ) 0 k 2 ( s , ξ ) H 2 ( u ̲ ( ξ , t , r ) , u ¯ ( ξ , t , r ) ) , if k 2 ( s , ξ ) < 0
for a s , ξ b , c τ t d and 0 r 1 .
Let for all a s , ξ b , c τ t d and 0 r 1 the functions G 1 ( β ) , G 2 ( β ) are increasing for β [ u ̲ ( s , t , r ) , u ¯ ( s , t , r ) ] and k 1 ( t , τ ) 0 , k 2 ( s , ξ ) 0 .
Then the parametric form of Equation (1) is
u ̲ ( s , t , r ) = g ̲ ( s , t , r ) + c t k 1 ( t , τ ) G 1 ( u ̲ ( s , τ , r ) ) d τ + a b k 2 ( s , ξ ) G 2 ( u ̲ ( ξ , t , r ) ) d ξ , s [ a , b ] , t [ c , d ] , r [ 0 , 1 ] ,
u ¯ ( s , t , r ) = g ¯ ( s , t , r ) + c t k 1 ( t , τ ) G 1 ( u ¯ ( s , τ , r ) ) d τ + a b k 2 ( s , ξ ) G 2 ( u ¯ ( ξ , t , r ) ) d ξ , s [ a , b ] , t [ c , d ] , r [ 0 , 1 ] .
We consider the operators
L ( u ̲ ( s , t , r ) ) = u ̲ ( s , t , r )
and
N ( u ̲ ( s , t , r ) ) = u ̲ ( s , t , r ) g ̲ ( s , t , r ) c t k 1 ( t , τ ) G 1 ( u ̲ ( s , τ , r ) ) d τ a b k 2 ( s , ξ ) G 2 ( u ̲ ( ξ , t , r ) ) d ξ .
Then we get
u ̲ m ( s , t , r ) = χ m u ̲ m 1 ( s , t , r ) + h R m ( u ̲ ˜ m 1 ( s , t , r ) ) ,
where χ m is determined by (8).
From (9) for the operator R m , m 1 we obtain
R m ( u ̲ ˜ m 1 ( s , t , r ) ) = 1 ( m 1 ) ! m 1 p m 1 N ( i = 0 u ̲ i ( s , t , r ) p i ) | p = 0 = 1 ( m 1 ) ! m 1 p m 1 i = 0 u ̲ i ( s , t , r ) p i g ̲ ( s , t , r ) | p = 0 1 ( m 1 ) ! m 1 p m 1 c t k 1 ( t , τ ) G 1 ( i = 0 u ̲ i ( s , τ , r ) ) p i d τ + a b k 2 ( s , ξ ) G 2 ( i = 0 u ̲ i ( ξ , t , r ) p i ) d ξ | p = 0 = u ̲ m 1 ( s , t , r ) 1 λ m ( m 1 ) ! g ̲ ( s , t , r ) 1 ( m 1 ) ! c t k 1 ( t , τ ) m 1 p m 1 G 1 ( i = 0 u ̲ i ( s , τ , r ) p i ) | p = 0 d τ 1 ( m 1 ) ! a b k 2 ( s , ξ ) m 1 p m 1 G 2 ( i = 0 u ̲ i ( ξ , t , r ) ) p i | p = 0 d ξ ,
By the definitions of the appropriate operators we obtain
u ̲ 1 ( s , t , r ) = h ( u ̲ 0 ( s , t , r ) g ̲ ( s , t , r ) c t k 1 ( t , τ ) G 1 ( u ̲ 0 ( s , τ , r ) ) d τ a b k 2 ( s , ξ ) G 2 ( u ̲ 0 ( ξ , t , r ) ) d ξ ) ,
where u ̲ 0 C ( A × [ 0 , 1 ] , R ) and for m 2
u ̲ m ( s , t , r ) = ( 1 + h ) u ̲ m 1 ( s , t , r ) h ( m 1 ) ! c t k 1 ( t , τ ) m 1 p m 1 G 1 ( i = 0 u ̲ i ( s , τ , r ) p i ) | p = 0 d τ h ( m 1 ) ! a b k 2 ( s , ξ ) m 1 p m 1 G 2 ( i = 0 u ̲ i ( ξ , t , r ) p i ) | p = 0 d ξ .

5. Convergence of the HAM

In this section we prove convergence of HAM and find an error estimate.
We introduce the following conditions:
( i ) g C ( A , E 1 ) , k 1 C ( [ c , d ] × [ c , d ] , R + ) , k 2 C ( [ a , b ] × [ a , b ] , R + ) and M 1 = max t , τ [ c , d ] k 1 ( t , τ ) > 0 and M 2 = max t , τ [ a , b ] k 2 ( s , ξ ) | > 0 .
( i i ) there exist constants L k > 0 , k = 1 , 2 such that
D ( G k ( u ) , G k ( v ) ) L k D ( u , v ) , for u , v E 1 , k = 1 , 2 ;
( i i i ) The inequality
α = M 1 L 1 ( d c ) + M 2 L 2 ( b a ) < 1
holds.
We will use the following result:
Theorem 1
([25]). Let the conditions (i)–(iii) be fulfilled. Then the integral Equation (1) has an unique solution.
Remark 2.
According to Theorem 1 if the conditions (i)–(iii) are fulfilled, then the Equations (14) and (15) have unique solutions.
Theorem 2.
Let
1.
the conditions (i)–(iii) are fulfilled.
2.
the functions u ̲ m ( s , t , r ) , m 1 , are defined by relations (17) and (18).
3.
L k < 1 for k = 1 , 2
4.
the series
i = 0 u ̲ i ( s , t , r )
converges.
Then the sum of the series (20) is the unique solution of Equation (14).
Proof. 
Introduce the notation
H ̲ k , m ( s , t , r ) = 1 m ! m p m G k ( i = 0 u ̲ i ( s , t , r ) p i ) | p = 0 , k = 1 , 2 .
From [33], if G k is the contraction mapping and the series (20) converges to u ̲ ( s , t , r ) , then the series m = 0 H ̲ k , m ( s , t , r ) , k = 1 , 2 , is respectively convergent to G k ( u ̲ ( s , t , r ) ) .
From condition 4, it follows that for any ( s , t , r ) A × [ 0 , 1 ] we have lim m u ̲ m ( s , t , r ) = 0 .
By applying the definition of the operator L we can write
m = 1 n L ( u ̲ m ( s , t , r ) χ m u ̲ m 1 ( s , t , r ) ) = m = 1 n ( u ̲ m ( s , t , r ) χ m u ̲ m 1 ( s , t , r ) ) = u ̲ n ( s , t , r ) .
Hence m = 1 L ( u ̲ m ( s , t , r ) χ m u ̲ m 1 ( s , t , r ) ) = lim n u ̲ n ( s , t , r ) = 0 .
Therefore, h m = 1 R m ( u ̲ ˜ m 1 ( s , t , r ) ) = m = 1 L ( u ̲ m ( s , t , r ) χ m u ̲ m 1 ( s , t , r ) ) and m = 1 R m ( u ̲ ˜ m 1 ( s , t , r ) ) = 0 .
Then we obtain
0 = m = 1 R m ( u ̲ ˜ m 1 ( s , t , r ) ) = m = 1 ( u ̲ m 1 ( s , t , r ) 1 λ m ( m 1 ) ! g ̲ ( s , t , r ) c t k 1 ( t , τ ) 1 ( m 1 ) ! m 1 p m 1 G 1 ( i = 0 u ̲ i ( s , τ , r ) ) p i | p = 0 d τ a b k 2 ( s , ξ ) 1 ( m 1 ) ! m 1 p m 1 G 2 ( i = 0 u ̲ i ( ξ , t , r ) ) p i | p = 0 d ξ = m = 1 u ̲ m 1 ( s , t , r ) g ̲ ( s , t , r ) c t k 1 ( t , τ ) m = 1 H ̲ 1 , m 1 ( s , τ , r ) d τ a b k 2 ( t , τ ) m = 1 H ̲ 2 , m 1 ( ξ , t , r ) d ξ .
Hence 0 = u ̲ ( s , t , r ) g ̲ ( s , t , r ) c t k 1 ( t , τ ) G 1 ( u ̲ ( s , τ , r ) ) d τ a b k 2 ( s , ξ ) G 2 ( u ̲ ( ξ , t , r ) ) d ξ . □
Theorem 3.
Let the condition (i)–(iii) be fulfilled.
Then the value of the convergence control parameter h can be selected such that the series (20) converges in A.
Proof. 
Let E = ( C ( A , R ) , . ) be the Banach space of all continuous functions on A and s ̲ n ( s , t , r ) = i = 0 n u ̲ i ( s , t , r ) for ( s , t ) A , r [ 0 , , 1 ] .
We will prove that { s ̲ n } is a Cauchy sequence in E.
We get
s ̲ 1 s ̲ 0 = u ̲ 1 | h | ( u ̲ 0 + g ̲ + M 1 ( d c ) ϕ 1 + M 2 ( b a ) ϕ 2 ) ,
where ϕ 1 = max ( s , t ) A | G 1 ( u ̲ 0 ( s , t , r ) | and ϕ 2 = max ( s , t ) A | G 2 ( u ̲ 0 ( s , t , r ) | .
Let m 1 and n > m . Then we obtain
s ̲ n s ̲ m = max ( s , t ) A | s ̲ n ( s , t , r ) s ̲ m ( s , t , r ) | = max ( s , t ) A | i = 0 n u ̲ i ( s , t , r ) i = 0 m u ̲ i ( s , t , r ) | = max ( s , t ) A | i = m + 1 n u ̲ i ( s , t , r ) | = max ( s , t ) A | i = m n 1 u ̲ i + 1 ( s , t , r ) | = max ( s , t ) A | ( 1 + h ) i = m n 1 u ̲ i ( s , t , r ) h ( i = m n 1 ( c t k 1 ( t , τ ) H ̲ 1 , i d τ + a b k 2 ( s , ξ ) H ̲ 2 , i d ξ ) ) | = max ( s , t ) A | ( 1 + h ) i = m n 1 u ̲ i ( s , t , r ) h { c t k 1 ( t , τ ) i = m n 1 H ̲ 1 , i d τ + a b k 2 ( s , ξ ) i = m n 1 H ̲ 2 , i d ξ ) } |
From [34] we have
i = m n 1 H ̲ 1 , i = G 1 ( s ̲ n 1 ) G 1 ( s ̲ m 1 ) , and i = m n 1 H ̲ 2 , i = G 2 ( s ̲ n 1 ) G 2 ( s ̲ m 1 ) .
Consequently, from conditions ( i i ) and ( i i i ) we obtain
s ̲ n s ̲ m = max ( s , t ) A | ( 1 + h ) ( s ̲ n 1 s ̲ m 1 ) h { c t k 1 ( t , τ ) ( G 1 ( s ̲ n 1 ) G 1 ( s ̲ m 1 ) ) d τ + a b k 2 ( s , ξ ) ( G 2 ( s ̲ n 1 ) G 2 ( s ̲ m 1 ) ) d ξ } | | 1 + h | s ̲ n 1 s ̲ m 1 + | h | { M 1 L 1 ( d c ) s ̲ n 1 s ̲ m 1 + m 2 L 2 ( b a ) s ̲ n 1 s ̲ m 1 } = ( | 1 + h | + | h | α ) s ̲ n 1 s ̲ m 1 = β h s ̲ n 1 s ̲ m 1
where β h = | 1 + h | + | h | α .
Let n = m + 1 then
s ̲ m + 1 s ̲ m β h s ̲ m s ̲ m 1 β h 2 s ̲ m 1 s ̲ m 2 β h m s ̲ 1 s ̲ 0 = β h m u ̲ 1 .
Using the triangle inequality and (22) we have
s ̲ n s ̲ m s ̲ m + 1 s ̲ m + s ̲ m + 2 s ̲ m + 1 + + s ̲ n s ̲ n 1 ( β h m + β h m + 1 + . . . + β h n 1 ) u ̲ 1 β h m ( 1 + β h + β h 2 + . . . + β h n m 1 ) u ̲ 1 β h m 1 β h n m 1 β h u ̲ 1 .
We choose the value of the parameter h ( 2 , 0 ) such that
0 < α < 1 | 1 + h | | h | .
Then β h = | 1 + h | + | h | α ( 0 , 1 ) and 1 β h n m < 1 . Therefore, s ̲ n s ̲ m < β h m 1 β h u ̲ 1
Since u ̲ 1 < then s ̲ n s ̲ m 0 as m ,and the sequence { s ̲ n } is a Cauchy sequence in E. Therefore, the series i = 1 u ̲ i ( s , t , r ) converges.
Similarly, we have { s ¯ n } is a Cauchy sequence. □
From the proof of Theorem 3 it follows an upper bound of the error:
Theorem 4.
Let the condition (i)–(iii) be fulfilled.
Then the error of the approximate solution can be estimated as follows
u ̲ s ̲ n β h n 1 β h | h | ( u ̲ 0 + g ̲ + M 1 ( d c ) ϕ 1 + M 2 ( b a ) ϕ 2 ) ,
where β h = | 1 + h | + | h | α and α is defined by (19).

6. Numerical Example

In this section, we will illustrate the obtained theoretical results on a numerical example.
Example 1.
Let A = [ 0 , 1 ] × [ 0 , 1 ] . Consider the Equation (1) in the partial case of
k 1 ( t , τ ) = 1 5 ( t + τ ) , k 2 ( s , ξ ) = 3 5 s ξ , G 1 ( u ) = 1 15 u 2 , G 2 ( u ) = 1 5 u
and
g ( s , t , r ) = ( ( 24 25 s t 7 900 s 2 t 4 ( 1 + r ) ) ( 1 + r ) , ( 24 25 s t 7 900 s 2 t 4 ( 3 r ) ) ( 3 r ) ) .
Then the exact solution of (1) in this partial case is given by
u e x a c t ( s , t , r ) = ( s t ( 1 + r ) , s t ( 3 r ) ) .
In this case the conditions (i)–(iii) are satisfied with M 1 = 2 5 , M 2 = 3 5 , L 1 = 2 5 , L 2 = 1 5 and α = 7 25 . We choose the value of parameter h : 7 25 < 1 | 1 + h | | h | or h ( 1.5625 , 0 ) . Numerically determined, the optimal value of the convergence control parameter h is equal to 1.0098 .
By using CAS “Wolfram Mathematica” and the proposed above method, we obtain
u ̲ 0 ( s , t , r ) = 0 , u ̲ 1 ( s , t , r ) = h ( u ̲ 0 ( s , t , r ) g ̲ ( s , t , r ) 0 t 1 75 ( t + τ ) u ̲ 0 2 ( s , τ , r ) d τ 0 1 3 25 s ξ u ̲ 0 ( ξ , t , r ) d ξ ) , u ̲ 1 ( s , t , r ) = h { 24 25 s t 7 900 s 2 t 4 ( 1 + r ) } ( 1 + r ) , u ̲ 2 ( s , t , r ) = ( 1 + h ) u ̲ 1 ( s , t , r ) h ( 0 t 1 75 ( t + τ ) 2 u ̲ 0 ( s , τ , r ) u ̲ 1 ( s , τ , r ) d τ + 0 1 3 25 s ξ u ̲ 1 ( ξ , t , r ) d ξ ) , u ̲ 2 ( s , t , r ) = { h ( 24 25 s t 7 900 s 2 t 4 ( 1 + r ) ) h 2 ( 24 2 25 2 s t + ( 7 3.10 4 s t 4 7 900 s 2 t 4 ) ( 1 + r ) ) } ( 1 + r ) , u ̲ 3 ( s , t , r ) = ( 1 + h ) u ̲ 2 ( s , t , r ) h ( 0 t 1 75 ( t + τ ) ( u ̲ 1 2 ( s , τ , r ) + 2 u ̲ 0 ( s , τ , r ) u ̲ 2 ( s , τ , r ) ) d τ + 0 1 3 25 s ξ u ̲ 2 ( ξ , t , r ) d ξ ) , u ̲ 3 ( s , t , r ) = { h ( 24 25 s t 7 900 s 2 t 4 ( 1 + r ) ) h 2 ( 1151 25 2 s t + 14 3.10 4 s t 4 ( 1 + r ) 7 450 s 2 t 4 ( 1 + r ) ) h 3 ( 13824 25 3 s t + 21 10 4 s t 4 ( 1 + r ) 343 36.25 3 s 2 t 4 ( 1 + r ) 26 27.25 3 s 3 t 7 ( 1 + r ) 2 + 931 675.900 2 s 4 t 10 ( 1 + r ) 3 ) } ( 1 + r ) , s ̲ 3 ( s , t , r ) = { 3 h ( 24 25 s t 7 900 s 2 t 4 ( 1 + r ) ) h 2 ( 1727 25 2 s t + 7 10 4 s t 4 ( 1 + r ) 7 300 s 2 t 4 ( 1 + r ) ) h 3 ( 13824 25 3 s t + 21 10 4 s t 4 ( 1 + r ) 343 36.25 3 s 2 t 4 ( 1 + r ) 26 27.25 3 s 3 t 7 ( 1 + r ) 2 + 931 675.900 2 s 4 t 10 ( 1 + r ) 3 ) } ( 1 + r ) .
Therefore, in this particular case we obtain for the error
Δ ̲ 3 ( s , t , 0.5 ) = | u ̲ e x a c t ( s , t , 0.5 ) s ̲ 3 ( s , t , 0.5 ) | .
The results are shown in the Table 1.
Remark 3.
In the standard case r = 1 we have E 1 = R and the Equation (1) is reduced to the exact equation. In this case
g ( s , t ) = g ̲ ( s , t , 1 ) = g ¯ ( s , t , 1 ) = 48 25 s t 14 450 s 2 t 4 ,
u e x a c t ( s , t ) = u ̲ e x a c t ( s , t , 1 ) = u ¯ e x a c t ( s , t , 1 ) = 2 s t ,
and differently than the “fuzy" case we obtain the following equalities:
u n ( s , t ) = u ̲ n ( s , t , 1 ) = u ¯ n ( s , t , 1 ) ,
s n ( s , t ) = s ̲ n ( s , t , 1 ) = s ¯ n ( s , t , 1 ) ,
Δ n ( s , t ) = Δ ̲ n ( s , t , 1 ) = Δ ¯ n ( s , t , 1 ) = | u e x a c t ( s , t ) s n ( s , t ) | , n = 0 , 1 , 2 , . . . .
These equalities lead to more simple calculations in the case of R comparatively with the “fuzy" case E 1 .

7. Conclusions

In this paper, HAM is applied for solving the two-dimensional nonlinear Volterra-Fredholm fuzzy integral equations, where the solution is found in the form of a series. It is shown that if this series is convergent, its sum is the solution of the considered equation. Sufficient conditions for the convergence of this series are given. Additionally, the error of the approximate solution, taken as the partial sum of generated series, is estimated. The presented example shows that the investigated method is effective in solving the equations of considered kind.

Author Contributions

Conceptualization, A.G.; Methodology, A.G., S.H.; Software, A.G.; Validation, A.G., S.H.; Formal Analysis, A.G., S.H.; Writing—Original Draft Preparation, A.G., S.H.; Writing—Review and Editing, A.G., S.H.; Funding Acquisition, S.H. All authors have read and agreed to the published version of the manuscript.

Funding

The research is supported by the Bulgarian National Science Fund under Project KP-06-N32/7.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Ahmad, M.Z.; Hasan, M.K. Modeling of Biological populations using fuzzy differential equations. In International Conference Mathematical and Computational Biology 2011; World Scientific: Singapore, 2012; pp. 354–363. [Google Scholar]
  2. Aleksandrov, V.M.; Kovalenko, E.V.; Mkhitarian, S.M. On a method of obtaining spectral relationships for integral operators of mixed problems of mechanics of continuous media. Appl. Math. Mech. 1983, 46, 825–832. [Google Scholar] [CrossRef]
  3. Biswas, S.; Roy, T.K. Fuzzy Linear Integral Equation and Its Application In Biomathematical Model. Adv. Fuzzy Math. 2017, 12, 1137–1157. [Google Scholar]
  4. Diamond, P. Theory and applications of fuzzy Volterra integral equations. IEEE Trans. Fuzzy Syst. 2002, 10, 97–102. [Google Scholar] [CrossRef]
  5. Wu, C.; Ma, M. On the integrals, series and integral equations of fuzzy set-valued functions. J. Harbin Inst. Tech. 1990, 21, 11–19. [Google Scholar]
  6. Jleli, M.; Karapinar, E.; Samet, B. On cyclic (psi,phi)-contractions in Kaleva-Seikkala’s type fuzzy metric spaces. J. Intell. Fuzzy Syst. 2014, 127, 2045–2053. [Google Scholar] [CrossRef]
  7. Roldan-Lopez-de-Hierro, A.; Karapınar, E.; Manro, S. Some new fixed point theorems in fuzzy metric spaces. J. Intell. Fuzzy Syst. Appl. Eng. Technol. 2014, 27, 2257–2264. [Google Scholar] [CrossRef]
  8. Wardowski, D. Fuzzy contractive mappings and fixed points in fuzzy metric spaces. Fuzzy Sets Syst. 2013, 222. [Google Scholar] [CrossRef]
  9. Bica, A.M.; Ziari, S. Open fuzzy cubature rule with application to nonlinear fuzzy Volterra integral equations in two dimensions. Fuzzy Sets Syst. 2019, 358, 108–131. [Google Scholar] [CrossRef]
  10. Enkov, S.; Georgieva, A.; Pavlova, A. Quadrature Rules and Iterative Numerical Method for Two-dimensional Nonlinear Fredholm Fuzzy Integral Equations. Commun. Appl. Anal. 2017, 21, 479–498. [Google Scholar]
  11. Georgieva, A.; Alidema, A. Convergence of homotopy perturbation method for solving of two-dimensional fuzzy Volterra functional integral equations. Adv. Comput. Ind. Math. Stud. Comput. Intell. 2019, 793, 129–145. [Google Scholar]
  12. Georgieva, A.; Pavlova, A.; Enkov, S. Iterative method for numerical solution of two-dimensional nonlinear Urysohn fuzzy integral equations. Adv. Comput. Ind. Math. Stud. Comput. Intell. 2019, 793, 147–161. [Google Scholar]
  13. Georgieva, A.; Pavlova, A.; Naydenova, I. Error estimate in the iterative numerical method for two-dimensional nonlinear Hammerstein-Fredholm fuzzy functional integral equations. Adv. Comput. Ind. Math. Stud. Comput. Intell. 2018, 728, 41–45. [Google Scholar]
  14. Behzadi, S.S. Solving Fuzzy Nonlinear Volterra-Fredholm Integral Equations by Using Homotopy Analysis and Adomian Decomposition Methods. Fuzzy Set Valued Anal. 2011, 35, 1–13. [Google Scholar] [CrossRef]
  15. Liao, S. Homotopy Analysis Method in Nonlinear Differential Equations; Springer: Berlin, Germany; Higher Education Press: Beijing, China, 2012. [Google Scholar]
  16. Vali, M.A.; Agheli, M.J.; Nezhad, S.G. Homotopy Analysis Method to Solve Two-Dimensional Fuzzy Fredholm Integral Equation. Gen. Math. Notes 2014, 22, 31–43. [Google Scholar]
  17. Liao, S. On the homotopy analysis method for nonlinear problem. Appl. Math. Comput. 2004, 147, 499–513. [Google Scholar] [CrossRef]
  18. Liao, S. Notes on the homotopy analysis method: Some definitions and theorems. Commun. Nonlinear Sci. Numer. Simul. 2009, 14, 983–997. [Google Scholar] [CrossRef]
  19. Abbasbandy, S.; Magyari, E.; Shivanian, E. The homotopy analysis method for multiple solutions of nonlinear boundary value problems. Commun. Nonlinear Sci. Numer. Simul. 2009, 14, 3530–3536. [Google Scholar] [CrossRef]
  20. Mirzaee, F.; Yari, M.K.; Paripour, M. Solving linear and nonlinear Abel fuzzy integral equations by homotopy analysis method. J. Taibah Univ. Sci. 2015, 9, 104–115. [Google Scholar] [CrossRef] [Green Version]
  21. Bataineh, A.S.; Noorani, M.S.M.; Hashim, I. Approximate analytical solutions of systems of PDEs by homotopy analysis method. Comput. Math. Appl. 2008, 55, 2913–2923. [Google Scholar] [CrossRef] [Green Version]
  22. Georgieva, A.; Naydenova, I. Application of homotopy analysis method for solving of two-dimensional linear Volterra fuzzy integral equations. AIP Conf. Proc. 2019, 2159, 030012. [Google Scholar]
  23. Ghanbari, M. Numerical Solution of Fuzzy Linear Volterra Integral Equations of the Second Kind by Homotopy Analysis Method. Int. Ind. Math. 2010, 2, 73–87. [Google Scholar]
  24. Muskhelishvili, N.I. Some Basic Problems of Mathematical Theory of Elasticity; Noordhoff: Groningen, The Netherlands, 1953. [Google Scholar]
  25. Georgieva, A. Solving two-dimensional nonlinear Volterra-Fredholm fuzzy integral equations by using Adomian decomposition method. Dyn. Syst. Appl. 2018, 24, 819–837. [Google Scholar]
  26. Goetschel, R.; Voxman, W. Elementary fuzzy calculus. Fuzzy Sets Syst. 1986, 18, 31–43. [Google Scholar] [CrossRef]
  27. Wu, C.; Gong, Z. On Henstock integral of fuzzy-number-valued functions. Fuzzy Sets Syst. 2001, 120, 523–532. [Google Scholar] [CrossRef]
  28. Gong, Z.; Wu, C. Bounded variation, absolute continuity and absolute integrability for fuzzy-number-valued functions. Fuzzy Sets Syst. 2002, 129, 83–94. [Google Scholar] [CrossRef]
  29. Sadatrasoul, S.; Ezzati, S. Quadrature rules and iterative method for numerical solution of two-dimensional fuzzy integral equations. Abstr. Appl. Anal. 2014, 2114, 1–18. [Google Scholar] [CrossRef]
  30. Russo, M.; Gorder, R.V. Control of error in the homotopy analysis of nonlinear Klein-Gordon initial value problems. Appl. Math. Comput. 2013, 219, 6494–6509. [Google Scholar] [CrossRef]
  31. Gorder, R.V.; Vajravelu, K. On the selection of auxiliary functions, operators, and convergence control parameters in the application of the homotopy analysis method to nonlinear differential equations: A general approach. Commun. Nonlinear Sci. Numer. Simul. 2009, 14, 4078–4089. [Google Scholar] [CrossRef]
  32. Odibat, Z. A study on the convergence of homotopy analysis method. Appl. Math. Comput. 2010, 217, 782–789. [Google Scholar] [CrossRef]
  33. Cherruault, Y. Convergence of Adomian’s method. Kybernetes 1989, 18, 31–38. [Google Scholar] [CrossRef]
  34. El-KallaI, L. Convergence of the Adomian method applied to a class of nonlinear integral equations. J. Appl. Math. Comput. 2008, 21, 327–376. [Google Scholar] [CrossRef] [Green Version]
Table 1. Values of errors Δ ̲ 3 ( s , t , 0.5 ) .
Table 1. Values of errors Δ ̲ 3 ( s , t , 0.5 ) .
h = 1 h = −1.0098
s t = 0.20.40.6 t = 0.20.40.6
0.12.0296 × 10 6 5.5943 × 10 6 1.4641 × 10 6 9.6097 × 10 7 3.3509 × 10 6 1.0844 × 10 5
0.36.2286 × 10 6 1.9018 × 10 5 5.5214 × 10 5 2.9769 × 10 6 1.1556 × 10 5 4.0119 × 10 5
0.51.0627 × 10 5 3.5632 × 10 5 1.1181 × 10 4 5.1318 × 10 6 2.1982 × 10 5 8.0508 × 10 5
0.71.5241 × 10 5 5.5688 × 10 5 1.8548 × 10 4 7.4422 × 10 6 3.4889 × 10 5 1.3308 × 10 4
h = 0.9 h = 1.4
s t = 0.20.40.6 t = 0.20.40.6
0.17.5640 × 10 5 1.5374 × 10 4 2.4064 × 10 4 1.2216 × 10 3 2.4483 × 10 3 3.6934 × 10 3
0.32.2747 × 10 4 4.7011 × 10 4 7.6690 × 10 4 3.6672 × 10 3 7.3840 × 10 3 1.1278 × 10 2
0.53.8006 × 10 4 7.9849 × 10 4 1.3538 × 10 3 6.1161 × 10 3 1.2371 × 10 2 1.9123 × 10 2
0.75.3341 × 10 4 1.1391 × 10 3 2.0022 × 10 3 8.5681 × 10 3 1.7409 × 10 2 2.7227 × 10 2

Share and Cite

MDPI and ACS Style

Georgieva, A.; Hristova, S. Homotopy Analysis Method to Solve Two-Dimensional Nonlinear Volterra-Fredholm Fuzzy Integral Equations. Fractal Fract. 2020, 4, 9. https://doi.org/10.3390/fractalfract4010009

AMA Style

Georgieva A, Hristova S. Homotopy Analysis Method to Solve Two-Dimensional Nonlinear Volterra-Fredholm Fuzzy Integral Equations. Fractal and Fractional. 2020; 4(1):9. https://doi.org/10.3390/fractalfract4010009

Chicago/Turabian Style

Georgieva, Atanaska, and Snezhana Hristova. 2020. "Homotopy Analysis Method to Solve Two-Dimensional Nonlinear Volterra-Fredholm Fuzzy Integral Equations" Fractal and Fractional 4, no. 1: 9. https://doi.org/10.3390/fractalfract4010009

APA Style

Georgieva, A., & Hristova, S. (2020). Homotopy Analysis Method to Solve Two-Dimensional Nonlinear Volterra-Fredholm Fuzzy Integral Equations. Fractal and Fractional, 4(1), 9. https://doi.org/10.3390/fractalfract4010009

Article Metrics

Back to TopTop