nextuppreviouscontents
Next:Part I Theoretical Foundation  Up: Abstract  Previous: Contents

Introduction

The survey paper [1] shows that the Richardson extrapolation technique is used to increase the accuracy of numerical solutions of elliptic problems with enough smoothness on the input data on corner domains. And the theoretical result also shows the existence of the asymptotic error expansion which carries out the success of the extrapolation technique. Although the above paper requires the smoothness on the input data, but the concentrated loading (will be defined in next section) or jump discontinuous input data has not been discussed, and the solutions for these two kinds of input data will have strong singularities. In this study, numerical experiments will be worked out to find the numerical solutions with these kinds of singularity.

In this report, both one-dimensional and two-dimensional problems will be considered as the numerical experiments to investigate the convergence of the numerical solutions. Moreover, by assuming the existence of the error asymptotic expansions at each node point, the convergence order can be evaluated by using the uniform refinement meshes. In one- dimensional problem, we will consider the following differential operator as the experiment

eqnarray714

where f(x) is the external forcing function. In the experiments, a continuous external forcing function will be worked out to verify the theoretical result in the survey paper [1], and then the concentrated loading and jump discontinuous input data will be taken as the examples in the experiment, trying to obtain the similar error asymptotic expansions. Afterward, a Poisson equation will be used for the two-dimensional experiment

eqnarray716

where f(x,y) is also the external forcing function and a square region will be chosen as the domain tex2html_wrap_inline6025 in this report. The same as the one-dimensional problem, the concentrated loading and jump discontinuous input data will also be taken as the examples in the experiment, however, we will try to obtain the error asymptotic expansions at different node points.

In addition, Green's function and Fourier series will also be used to find the exact solutions those will be compared with the numerical solutions. And the numerical solutions are obtained by the method of finite difference from the experiments in this study. Therefore, some definitions and basic results will be presented in the next section and will be used later.
 
 


nextuppreviouscontents
Next:Part I Theoretical Foundation  Up: Abstract   Previous: Contents
Cheung Sau Hung

Wed Sep 15 10:03:39 HKT 1999