Research Article | Open Access

Volume 2015 |Article ID 346036 | https://doi.org/10.1155/2015/346036

A. A. Dahalan, J. Sulaiman, "Implementation of TAGE Method Using Seikkala Derivatives Applied to Two-Point Fuzzy Boundary Value Problems", International Journal of Differential Equations, vol. 2015, Article ID 346036, 7 pages, 2015. https://doi.org/10.1155/2015/346036

# Implementation of TAGE Method Using Seikkala Derivatives Applied to Two-Point Fuzzy Boundary Value Problems

Revised15 Apr 2015
Accepted19 Apr 2015
Published14 Jun 2015

#### Abstract

Iterative methods particularly the Two-Parameter Alternating Group Explicit (TAGE) methods are used to solve system of linear equations generated from the discretization of two-point fuzzy boundary value problems (FBVPs). The formulation and implementation of the TAGE method are also presented. Then numerical experiments are carried out onto two example problems to verify the effectiveness of the method. The results show that TAGE method is superior compared to GS method in the aspect of number of iterations, execution time, and Hausdorff distance.

#### 1. Introduction

Fuzzy boundary value problems (FBVPs) and treating fuzzy differential equations were one of the major applications for fuzzy number arithmetic . FBVPs can be approached by two types. For instance, the first approach addresses problems in which the boundary values are fuzzy where the solution is still in fuzzy function. Then the second approach is based on generating the fuzzy solution from the crisp solution . To solve these problems, numerical methods obtain their approximate solution. Consequently, in this paper, let two-point linear FBVPs be defined in general form as follows: where is a fuzzy function and , , and are continuous functions on , whereas, and are fuzzy numbers.

Based on the Seikkala derivative , (1) will be solved numerically by applying the second-order central finite difference scheme to discretize the two-point linear FBVPs into linear systems. Then the generated linear systems will be solved iteratively by using Two-Parameter Alternating Group Explicit (TAGE) method [4, 5]. By considering the Group Explicit (GE) method for the numerical solution of parabolic and elliptic problems, Evans [6, 7] discovered Alternating Group Explicit method. Later, Sukon and Evans  expanded this approach to initiate the TAGE method thus proving that this method is superior compared to AGE method. From previous studies, findings of the papers related to the TAGE iterative method and its variants  have shown that TAGE method has been widely used to solve the nonfuzzy problems. Due to the efficiency of the methods, this paper extends the application of TAGE iterative method in solving fuzzy problems. Since the fuzzy linear systems will be constructed, the iterative method becomes the natural option to get a fuzzy numerical solution of the problem.

The outline of the paper is organized as follows. Section 2 will discuss the finite difference method based on the second-order finite difference scheme in discretizing two-point FBVPs, while Section 3 presents the formulation and implementation of the TAGE methods in solving linear systems generated from the second-order finite difference scheme. Section 4 shows some numerical examples and conclusions are given in Section 5.

#### 2. Finite Difference Approximation Equations

To be clear, let be a fuzzy subset of real numbers. It is characterized by the corresponding membership function evaluated at , writing as a number in . -cut of , in which is denoted as a crisp number, can be written as in , for . The interval of the -cut of fuzzy numbers will be written as , for all , since they were always closed and bounded . Suppose is parametric form of fuzzy function . For arbitrary positive integer subdivide the interval , whereas for and .

Denote the value of and at the representative point by at . Thus, by using the second-order central finite difference scheme, problem (1) can be developed aswhich giveBy using parametric form of fuzzy function, (1) can be written asSuppose that and for . ThenBy applying (2a) and (3a), (6a) will be reduced tofor . Meanwhile, by substituting (2b) and (3b) into (6b), we will haveThen, (7a) and (7b) can be rewritten as follows:respectively, for . Since both of (8a) and (8b) have the same form in terms of the equation, except that, based on the interval of the -cuts, the differences are identified only in the upper and lower bounds, it can be rewritten asfor , where

Now, we can express the second-order central finite difference approximation (9) in a matrix form aswithSince this study will deal with an application of the method, the computational method of it will be diagonally dominant matrix and positive definite matrix .

#### 3. Two-Parameter Alternating Group Explicit Iterative Method

Based on previous study conducted by Evans, clearly we can see that they have discussed theoretically how to compute the value of parameter given by Mohanty et al. . In this paper, the optimum value of parameters and will be calculated by implementing several numerical experiments, so those optimum values will be found if the number of iterations is smaller.

Family of AGE can be considered efficient to two-step method to solve linear system. None of the researchers had been trying to apply this method in solving fuzzy problem generated from discretization of fuzzy partial difference equation. This paper will discuss the application of this iterative method which will solve the fuzzy linear system given by (1). Consider a class of methods mentioned in [4, 5] which is based on the splitting of the matrix into the sum of its constituent symmetric and positive definite matrices, as follows:where if is odd. Similarly, we define the following matrices:if is even, with . In this paper, we only consider that case is even.

Then (11) becomesThus, the explicit form of TAGE method can be written aswhere are the acceleration parameters, and a pair of and are invertible. From (17), therefore, the implementation of TAGE method is presented in Algorithm 1.

Algorithm 1 (TAGE method). (i)Initialize and .(ii)For , initialize parameters , , , , , , , and .(iii) First Sweep. For ,compute(iv) Second Sweep. For ,compute(v) Convergence Test. If the convergence criterion, that is, , is satisfied, go to Step (vi). Otherwise go back to Step (ii).(vi)Display approximate solutions.

#### 4. Numerical Experiments

Two examples of FBVPs are considered to verify the effectiveness of GS, AGE, and TAGE methods. For comparison purposes, three parameters were observed that are number of iterations, execution time (in seconds), and Hausdorff distance (as mentioned in Definition 2). Based on these two problems, numerical results for GS, AGE, and TAGE methods have been recorded in Tables 1 to 5.

 Methods n 512 1024 2048 4096 8192 Problem Number of iterations GS 681711 2431928 8548735 29480437 99066551 AGE 96747 354438 1279808 4549671 15883620 TAGE 77377 279463 876061 2879619 10383345 Execution time GS 48.94 211.19 989.91 5719.20 32465.10 AGE 8.00 39.00 202.00 1310.00 8125.00 TAGE 7.00 31.00 141.00 822.00 5342.00 Hausdorff distance GS AGE TAGE Problem Number of iterations GS 475487 1692329 5930853 20369573 68062962 AGE 67638 247434 891667 3161503 10997813 TAGE 53492 187245 671456 2122064 7505046 Execution time GS 35.27 155.77 764.09 4457.31 26063.40 AGE 6.00 27.00 141.00 912.00 5676.00 TAGE 5.00 20.00 107.00 608.00 3887.00 Hausdorff distance GS AGE TAGE
 Methods n 512 1024 2048 4096 8192 Problem Number of iterations GS 682475 2434982 8560953 29529307 99262033 AGE 96840 354815 1281323 4555751 15908020 TAGE 77449 279746 876948 2882382 10399116 Execution time GS 49.07 211.36 991.23 5874.81 32551.12 AGE 9.00 39.00 202.00 1301.00 8164.00 TAGE 7.00 31.00 141.00 827.00 5402.00 Hausdorff distance GS AGE TAGE Problem Number of iterations GS 476030 1694502 5939547 20404350 68202066 AGE 67704 247701 892745 3165828 11015151 TAGE 53543 187435 672208 2124610 7514448 Execution time GS 35.26 155.79 756.06 4465.35 25999.98 AGE 6.00 27.00 142.00 903.00 5652.00 TAGE 5.00 21.00 106.00 605.00 3893.00 Hausdorff distance GS AGE TAGE
 Methods n 512 1024 2048 4096 8192 Problem Number of iterations GS 683007 2437112 8569470 29563373 99398298 AGE 96905 355076 1282378 4559989 15925021 TAGE 77499 279944 877567 2884304 10408307 Execution time GS 49.25 210.43 988.93 5784.36 32665.34 AGE 9.00 39.00 203.00 1311.00 8152.00 TAGE 6.00 31.00 141.00 837.00 5397.00 Hausdorff distance GS AGE TAGE Problem Number of iterations GS 476410 1696018 5945607 20428592 68299033 AGE 67751 247888 893496 3168843 11027246 TAGE 53578 187569 672733 2126364 7520993 Execution time GS 35.40 155.80 757.38 4585.51 26078.03 AGE 6.00 27.00 141.00 912.00 5696.00 TAGE 5.00 21.00 107.00 620.00 3900.00 Hausdorff distance GS AGE TAGE
 Methods n 512 1024 2048 4096 8192 Problem Number of iterations GS 683321 2438369 8574499 29583490 99478766 AGE 96944 355232 1283001 4562489 15935054 TAGE 77528 280061 877932 2885438 10414635 Execution time GS 49.22 210.33 1026.58 5771.53 32617.94 AGE 8.00 39.00 203.00 1298.00 8186.00 TAGE 7.00 31.00 141.00 835.00 5395.00 Hausdorff distance GS AGE TAGE Problem Number of iterations GS 476633 1696912 5949186 20442908 68356295 AGE 67778 247998 893940 3170624 11034378 TAGE 53599 187647 673042 2127413 7524856 Execution time GS 35.42 155.72 757.27 4364.75 26127.43 AGE 6.00 27.00 141.00 914.00 5706.00 TAGE 4.00 20.00 107.00 612.00 3937.00 Hausdorff distance GS AGE TAGE
 Methods n 512 1024 2048 4096 8192 Problem Number of iterations GS 683426 2438784 8576162 29590144 99505380 AGE 96956 355282 1283208 4563320 15938400 TAGE 77538 280098 878054 2885812 10416768 Execution time GS 49.45 210.66 809.53 5758.67 32519.13 AGE 9.00 39.00 202.00 1313.00 8221.00 TAGE 7.00 31.00 141.00 817.00 5383.00 Hausdorff distance GS AGE TAGE Problem Number of iterations GS 476706 1697208 5950370 20447642 68375230 AGE 67786 248034 894086 3171216 11036748 TAGE 53606 187674 673146 2127768 7526132 Execution time GS 35.43 155.72 755.20 4615.31 25815.45 AGE 6.00 27.00 141.00 915.00 5662.00 TAGE 4.00 21.00 107.00 613.00 3941.00 Hausdorff distance GS AGE TAGE

Definition 2 (see ). Given two minimum bounding rectangles and , a lower bound of the Hausdorff distance from the elements confined by to the elements confined by is defined as

Problem 1. Consider where with the boundary conditions and . The exact solutions forarerespectively.

Problem 2 (see ). Consider where with the boundary conditions and . The exact solutions forarerespectively.

#### 5. Conclusions

In this paper, TAGE method was used to solve linear systems which arise from the discretization of two-point FBVPs using the second-order central finite difference scheme. The results show that TAGE method is more superior in terms of the number of iterations, execution time, and Hausdorff distance compared to the AGE and GS methods. Since TAGE is well suited for parallel computation, it can be considered as a main advantage because this method has groups of independent task which can be implemented simultaneously. It is hoped that the capa