Doctoral Degrees (Mathematics and Applied Mathematics)
Permanent URI for this collection
Browse
Browsing Doctoral Degrees (Mathematics and Applied Mathematics) by Subject "Algorithms"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
Item Open Access Analysis of numerical approximation algorithms for nonlinear differential equations using a discrete multiple scales technique(University of the Free State, 2002-12) Maré, Eben; Schoombie, S. W.English: Perturbation techniques for the solution of differential equations form an essential ingredient of the tools of mathematics as applied to physics, engineering, finance and other areas of applied mathematics. A natural extension would be to seek perturbation-type solutions for discrete approximations of differential equations. The main objective of the research project is to develop a perturbation technique for approximation. The spurious behavior, predicted theoretically, is shown to be present experimentally, independent of temporal discretization. We also detail some comparisons of central difference solutions of different orders of approximation to the KdV equation. The results show a clear benefit of higher order central differences relative to lower order methods. The benefit of the central difference methodology would also extend to more general regions over which we would solve partial differential equations. We also show that the method of multiple scales can provide an adequate explanation for spurious behavior in a difference scheme for the Van der Pal equation. the solution of discrete equations. We discuss the well-known method of multiple scales and show its use for the solution of the Korteweg-de Vries (KdV), Regularized Long Wave (RLW) and Van der Pol equations. In particular, for the KdV and RLW equations the analysis shows that the envelopes of modulated waves are governed by the nonlinear Schrödinger equation. We present a variation of the multiple scales technique which presents an ideal framework from which we devise a discrete multiple scales analysis methodology. We discuss a discrete multiple scales methodology derived by Schoombie [111], as applied to the Zabusky-Kruskal approximation of the KdV equation. This discrete multiple scales analysis methodology is generalized and applied to the solution of a generalized finite difference approximation of the KdV equation. We show the consistency of the method with the continuous analysis as the discretization parameters tend to zero. The discrete multiple scales technique is a powerful tool for the examination of modulational properties of the KdV equation. In the case of certain modes of the carrier wave, the discrete multiple scales analysis breaks down, indicating that the numerical solution deviates in behavior from that of the KdV equation. Several numerical experiments are performed to examine the spurious behavior for different orders of approximation. The spurious behavior, predicted theoretically, is shown to be present experimentally, independent of temporal discretization. We also detail some comparisons of central difference solutions of different orders of approximation to the KdV equation. The results show a clear benefit of higher order central differences relative to lower order methods. The benefit of the central difference methodology would also extend to more general regions over which we would solve partial differential equations. We also show that the method of multiple scales can provide an adequate explanation for spurious behavior in a difference scheme for the Van der Pal equation.