首页> 外文学位 >Variational Germano identity applied to the numerical simulation of multiscale partial differential equations.
【24h】

Variational Germano identity applied to the numerical simulation of multiscale partial differential equations.

机译:变分德耳曼恒等式应用于多尺度偏微分方程的数值模拟。

获取原文
获取原文并翻译 | 示例

摘要

The simulation of turbulence is a challenging problem due to the cost of resolving all length scales that arise in a flow. This problem is addressed in engineering applications using large eddy simulation (LES), a class of numerical algorithms designed to only resolve a coarse representation of the exact solution.; In the LES of a multiscale system the effect of the unresolved scales on the resolved scales is represented by a sub-grid model, which is a functional of only the resolved scales. Most popular subgrid models contain parameters which must be fine-tuned for a particular flow and numerical method. Furthermore, the performance of these models is sensitive to the precise value of the parameters. In this dissertation, the variational Germano identity (VGI) is derived to automatically determine the parameters of an arbitrary subgrid model. In addition, a new class of LES models based on a dynamic generalization of the variational multiscale method is developed.; The VGI derived in this dissertation is inspired by its filtered counterpart (derived by M. Germano in 1991). It is based on requiring the numerical solution to be optimal in a user-defined metric. Experiments, comparing filtered and variational forms of the Germano identity and a range of subgrid models, are performed on several systems including the incompressible Navier Stokes, Burgers, and compressible Navier-stokes equations. The results of these simulations are compared to solutions obtained using well-resolved, but computationally expensive, direct numerical simulations. It is concluded that the VGI is more robust than the filtered form and that it is able to adjust to a definition of an optimal solution.; The VGI is also applied to a new class of LES models developed in this study. These models, which are based on the variational multiscale concept, better represent the interscale transfer of energy within the resolved scales by allowing for two independent viscosities: one which acts on all resolved scales and another which acts only on the fine resolved scales. These models are tested for, Burgers equation and the three dimensional Navier Stokes equations, and are found to be more accurate than their single viscosity counterparts.
机译:由于解决流中出现的所有长度尺度的成本,湍流的模拟是一个具有挑战性的问题。在使用大型涡流仿真(LES)的工程应用中解决了这个问题,LES是一类数值算法,旨在仅解析精确解的粗略表示。在多尺度系统的LES中,未解析尺度对解析尺度的影响由子网格模型表示,该模型仅是解析尺度的功能。最流行的子网格模型包含必须针对特定流和数值方法进行微调的参数。此外,这些模型的性能对参数的精确值敏感。本文提出了变分Germano身份(VGI)来自动确定任意子网格模型的参数。另外,基于变分多尺度方法的动态概括,开发了一种新的LES模型。本文的VGI源自其过滤后的对应物(M. Germano于1991年得出)。它基于要求数值解在用户定义的量度中是最佳的。在包括不可压缩的Navier Stokes,Burgers和可压缩的Navier-stokes方程在内的多个系统上进行了实验,比较了Germano身份的过滤形式和变体形式以及一系列子网格模型。将这些模拟的结果与使用解析度高但计算量大的直接数值模拟获得的解进行比较。结论是,VGI比过滤后的表单更健壮,并且可以调整为最佳解决方案的定义。 VGI还应用于本研究中开发的新型LES模型。这些基于变分多尺度概念的模型通过允许两种独立的粘度更好地表示了分辨尺度内能量的尺度间转移:一种作用于所有分辨尺度,另一种仅作用于精细分辨尺度。测试了这些模型的Burgers方程和三维Navier Stokes方程,发现它们比单个粘度模型更准确。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号