Solving Directly Two Point Non Linear Boundary Value Problems Using Direct Adams Moulton Method

Problem statement: In this study, a direct method of Adams Moulton type was developed for solving non linear two point Boundary Value Problems (BVPs) directly. Most of the existence researches involving BVPs will reduced the problem to a system of first order Ordinary Differential Equations (ODEs). This approach is very well established but it obviously will enlarge the systems of first order equations. However, the direct method in this research will solved the second order BVPs directly without reducing it to first order ODEs. Approach: Lagrange interpolation polynomial was applied in the derivation of the proposed method. The method was implemented using constant step size via shooting technique in order to determine the approximated solutions. The shooting technique will employ the Newton’s method for checking the convergent of the guessing values for the next iteration. Results: Numerical results confirmed that the direct method gave better accuracy and converged faster compared to the existing method. Conclusion: The proposed direct method is suitable for solving two point non linear boundary value problems.


INTRODUCTION
This study is concerned with the numerical solution of solving directly the second order two-point boundary value problems of the form as follows Eq. 1 and 2: ( Many problems in applied sciences and engineering are modeled as two-point boundary value problems such as the boundary layer theory in fluid mechanic, heat power transmission theory, space technology and reaction kinetics. A recent application of the boundary value problem can be found in Adegbie and Alao (2007) Jamshidi and Rostami (2008); Hassan (2009). Since the boundary value problem has wide application in scientific research, therefore faster and accurate numerical solutions of boundary value problem are very importance.
There are several methods that can be used to solve the two point boundary value problems numerically. It had been proposed by Attili and Syam (2008); Ha (2001); Jafri et al. (2009) and Taiwo and Ogunlaran (2008). Ha (2001) had solved the two-point boundary value problem using fourth order Runge-Kutta method via shooting technique. The second order boundary value problem has been reduced to a system of first order equation. Jafri et al. (2009) has consider solving directly two-point boundary value problem for second order ordinary differential equations using multistep method in term of backward difference formulas.
In this study, we propose a direct method of Adams Moulton type via shooting technique to solve the boundary value problem directly. The Newton method will be implemented as the iterative method to estimate the guessing values. The given equations in (1) will be treated in their original second order form and therefore the requirement of the storage is lower.

METERIALS AND METHODS
Formulation of the method: The point, y n+1 at x n+1 can be obtained by integrating (1) The function f(x, y, y') in (4) and (6) will be approximated using Lagrange interpolating polynomial and the interpolation points involved are four points, five points and six points ie{x n−2 , x n−1 , x n+1 },{x n−3 , x n−1 , x n , x n+1 }and {x n−4 , x n−3 , x n−2 , x n−1 , x n , x n+2 }respectively. ) n 1 n n 1 n 1 n n 2 n 3 251f 646f 264f h y y 106f 19f 720 n 1 n n 1 n 1 n n n 2 n 3 135f 752f 264f h y y hy 96f 17f 1440

Direct method of order 6:
Predictor: n n 1 n 2 n 1 n n n 3 n 4 1427f 1596f 1446f h y y hy 692f 135f 1440

Corrector:
) n 1 n n 1 n 1 n n 2 n 3 n 4 475f 1427f 798f h y y 482f 173f 27f 1440 n 1 n n 1 n 1 n n n 2 n 3 n 4 863f 5674f 2542f h y y hy 1492f 52f 82f 10080 The Euler method will be used only once at the beginning of the code to find the additional points for the starting initial points. Then, the predictor and corrector direct method can be applied until the end of the interval. This direct method will be adapted for solving the boundary value problems via shooting techniques. Shooting technique will allow for new guessing and for each new guessing of the y', the Euler method will be used again to find the starting initial points.

Implementation of the method:
The values of y' n+1 and y n+1 will be implemented using the predictor and corrector schemes. The sequence of computation involved was PECE where P and C indicate the application of the predictor and corrector formula respectively and E indicate the evaluation of the function f. Shooting techniques is an analogy with the procedure of firing objects as a stationary target. It solves the problem with trial and error. To form an initial value problem out of boundary value problem (1), the initial value of y' need to be guessed. The solution depends on both x and s 0 . Starting with the initial guess, s 0 , the approximated solution of the derivative y'(a) gives Eq. 7-11: Equation 6 can be written as: For the first initial guessing, s 0 , we considered Richard et al. (1981) Eq. 9: The solution of (8) will coincide with the solution of (1-2) if we could find the value of s = s v such that: Newton method will be used to get a very rapidly converging iteration. We compute the {s v } defined as: Differentiate (8) Therefore, the new guess can be calculated base on the previous guess using: Both of Eq. 8 and 11 will be solved simultaneously using the direct method. The process is repeated over and over until the error |β-y(b,s n )|≤tolerance. The algorithms of the proposed method were developed in C language.
Tested problem: Three tested problems are presented and those problems are referred Ha (2001). The problems will be tested to the direct method of Adams Moulton type (DAM) mentioned earlier. However, Ha (2001) used the fourth-order Runge-Kutta method via shooting technique using generalized Newton's method and reduced the given problems to first order system of equations.

RESULTS
The following notations are used in the tables:

DISCUSSION
In Problem 1, we chose h = 0.01 and error bound = 10 −5 . Table 1 showed the value of iteration for calculating s v in DAM4, DAM5 and DAM6 methods when using approach in (9) and obtained s 0 =-1.3335. The final calculating s v is converged to the values of -14.000 for DAM4, DAM5 and DAM6. The iteration needed for calculating s v when s 0 =−1.3335 is five. The DAM methods manage to give good numerical results at each step. Using the same h and error bound 10 −5 we tried the initial guess s 0 = 0.25, 0.5 and 1.0 as Ha (2001). The comparison for iterating those initial guess with Ha (2001) is showed in Table 2 The maximum errors and number of iteration for DAM4 using s 0 = 0.25, 0.5, 1.0 are much better than HA as shown in Table 2-4. The DAM4, DAM5 and DAM6 methods have converging final s v = -14.0000 for each initial guess in Table 2-4. It is also observed that all the DAM method has the same number of iteration and the error at x last is very small. The DAM methods manage to obtain better accuracy as the order increased.
In Problem 2, we used the same approach as in (9) and obtained the initial guess s 0 = 0.5. Table 5 shows a good numerical result when using initial values s 0 = 0.5. The final calculating s v is converged to the values of 0000 . 0 − 0 for all DAM4, DAM5 and DAM6. The iteration needed for calculating s v when s 0 = 0.5 is fourteen. The error estimated at the last x is very accurate i.e., 3.129e-13, 5.118e-12 and 6.267e-11 for DAM4, DAM5 and DAM6 respectively. The DAM method manages to give good numerical results at each order. Then we chose h =0.05 and error bound = 10 −5 as Ha (2001); and tried the initial guess s 0 = 0.6975. The maximum error and number of iteration for DAM4 is better compared to HA in Table 6. The DAM4 only need 40 iterations in order to converge while HA took 9982 iterations. The DAM methods have better accuracy as the order increase and the converging final is s v = 0.00000. The DAM5 and DAM6 only need 41 and 42 iterations in order to converge. Table 7 show the numerical results of DAM4, DAM5 and DAM6 method when initial guess s 0 = 0.1667 using approach in (6) when solving Problem 3. This problem was tested using h = 0.05 and error bound 10 −5 . The final calculating of s v is converged to the values of -0.25000. The iteration needed for calculating s v when s 0 = 0.1667 is three. The DAM methods manage to give good numerical results at each order.
We tried the initial guess of s 0 = 3.99, 4.00, 4.10, 4.15 and 4.175 when solving Problem 3. The comparison for iterating those initial guess with Ha (2001) is showed in Table 8-12.          In Table 8-12, we could observe that the maximum error for DAM4 is better than HA. The approximating error at x last for DAM4 is much better than in HA. The iteration needed in DAM4, DAM5 and DAM6 for calculating s v is five. The converging final for all the DAM methods is -0.25000 but HA converge to -0.24935.

CONCLUSION
In this study, we have shown the proposed direct Adams Moulton method via shooting technique using Newton's method is suitable for solving two point second order non linear boundary value problems. This proposed method is simple, efficient and economically.