节点文献

Obrechkoff方法在求解常微分方程振荡、刚性问题中的应用研究

Research on Obrechkoff Numerical Methods for Oscillatory and Stiff Differential Equations

【作者】 戴勇鸣

【导师】 汪仲诚;

【作者基本信息】 上海大学 , 无线电物理, 2006, 博士

【摘要】 由于常微分方程本身的重要性以及在不同领域的广泛应用,贯穿整个20世纪,常微分方程的数值求解研究得到了巨大的发展。特别是,随着计算机性能的快速提高,一些著名数学软件的不断深化发展,更多的新思想得以实现,更多的复杂方法涌现出来,常微分方程数值求解以及数值方法发展研究的领域有不断深化扩大的趋势。 计算机的数值计算功能对物理学中常微分方程研究的用途不仅仅是可以得到数值结果,更为重要的是,它为物理学家提供了“计算机模拟实验”这个新的研究手段。有了计算机数值计算这个强有力的工具,我们将目光投向物理领域中一些较为复杂的常微分方程(非线性Duffing方程,周期性振荡方程以及刚性方程)的数值求解与相应数值方法的研究。 在物理领域中,常常可以遇到一些应用很是广泛的常微分方程,例如薛定锷方程、非线性Duffing方程、天体轨道方程以及刚性方程等。这些方程多为一阶或二阶的常微分方程,形式简单,却很少能得到解析解。即使数值求解也往往存在着求解精度不高或因方程本身性质特殊造成数值方法求解结果不尽理想。在这些问题中具有代表性的有两类问题:周期性振荡问题与刚性问题。 在本论文中,我们主要集中于这两类问题相应的数值方法研究做出探讨。 对于周期性振荡问题,我们主要关注二阶常微分方程 y″(x)+ω2y(x)=f(x,y),y′(0)=y′0,y(0)=y0 这类方程的近似解析解中常包含cos(ωx)、sin(ωx)或eiωx等周期性函数。鉴于其周期振荡性质,往往造成数值方法求解困难,结果出现不稳定,甚至发散。 我们研究发现对于具有周期振荡性质的问题必须有匹配的数值方法,即数值方法也需具有周期振荡性质。否则即使原本精度很高的方法,如果与所求解问题的性质不匹配,数值求解的结果也往往是不理想,甚至得到发散的结果。反之,如果数值方法与问题匹配但精度不够,同样也不能得到满意的结果。 为此,我们从两方面出发研究针对周期性振荡问题的数值方法。

【Abstract】 Because of the importance of ordinary differential equations and their wide application in diffirent fields, through the whole 20th century, the research on numerical solutions of ODEs made enormous progress. Particularly, with the fast development of electronic computer and some famous mathematics softwares, more and more new ideas and approaches which surpass classic methods could be realized. At the same time, the field of numerical methods has been expanded.The computational function of computer is not merely to get numerical results for physical research, the more important thing is, and it has offered this new research means of "computer simulation experiment" to physicist. With the powerful tool of electronic computation, we cast sight on numerical solutions to some comparatively complicated differential equations in the physics field (nonlinear Duffing equation, periodic oscillatory equation and stiff equation) and the research on the corresponding numerical methods.In the physics field, we often encounter some differential equation with wide application, for example, Schr6dinger equation, nonlinear Duffing equation, orbital equation and stiff equation. Most of these equations are first-order or second-order equations with simple form, but seldom of could be analytically solved. Because of low accurate numerical methods and the specific property of the equations, it is hard to get ideal results even in numerical way. There are two representative problems: periodic oscillatory problem and stiff problem.In this desertation, we focus mainly on numerical methods for oscillatory and stiff problems.For the oscillatory problem, we focus on second-order differential equationCommonly, there exist some periodic functions like cos(ωx)、 sin(ωx) or etωx in the approximately analytic solution to periodic oscillatory problem. Due to its oscillatory attribute, it is difficult to solve the problem numerically, and even if it is done, the numerical results often become unstable or even diffused.

  • 【网络出版投稿人】 上海大学
  • 【网络出版年期】2007年 02期
节点文献中: 

本文链接的文献网络图示:

本文的引文网络