Numerical cubature for fully symmetric regions and adaptive methods for PDEs

完全对称区域的数值体积和偏微分方程的自适应方法

• 批准号: 2699-2007
• 负责人: Keast, Patrick
• 金额: $ 1.17万
• 依托单位: Dalhousie University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2007
• 资助国家: 加拿大
• 起止时间: 2007-01-01 至 2008-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=366191
• 关键词: Numerical; cubature; fully; symmetric; regions

This research programme has two principal thrusts. One direction continues work carried out in thelast five years on the numerical solution of time dependent partial differential equations in onespace variable. Differential equations occur in many physical models, and often the solutionexhibits narrow layers with peaks or sharp variations. In many cases these peaks or areas of rapidchange move in time. For an approximate method of solution to be able to catch these features, thediscrete model must adapt to the solution. Software to approximate to the solutions with reliable errorestimates will be developed. We will investigate the use of high order interpolants in a method oflines context, in order to obtain inexpensive and reliable error estimates. Using anequi-distribution algorithm we will ensure that the error globally satisfies a user-defined errortolerance.The second thrust, which also is a continuation of a long standing research programme, isthe construction of methods for the numerical evaluation of integrals over multidimensionalregions having specified symmetries.  Examples of such regions are the n dimensional cube,sphere, simplex, octahedron, spherical surface, and the whole of n-space with a symmetricweight function. As shown in an earlier paper there is a strong connection between formulasfor the surface of the sphere and for the simplex. It is intended to investigate connectionsbetween formulas for the octahedron and the simplex. A long term plan is to set up softwareto accept as input the symmetries of the region, and a requested polynomial degree, and thento compute consistent structures for the formulas, set up the defining equations, and solvethem.

这项研究计划有两个主要目标。其中一个方向继续了过去五年来在单空间变量时变偏微分方程数值解方面所做的工作。微分方程出现在许多物理模型中，并且解通常呈现出具有峰值或急剧变化的窄层。在许多情况下，这些快速变化的峰值或区域随时间移动。对于能够捕捉这些特征的近似解方法，离散模型必须适应解。将开发具有可靠误差估计的近似解的软件。我们将研究使用高阶插值的方法oflines的背景下，为了获得廉价和可靠的误差估计。本文的第二个目标是建立具有特定对称性的多维区域上积分的数值计算方法，例如n维立方体、球面、单形、八面体、球面、椭圆形等。以及整个n-空间的权函数。正如在较早的文件有一个强大的连接formulasfor表面的领域和单纯形。它的目的是调查之间的联系公式的八面体和单纯形。一个长期的计划是建立软件接受输入的对称性的区域，和一个要求的多项式次数，然后计算一致的结构的公式，建立定义方程，并解决他们。