本文研究了散度形式随机线性椭圆偏微分方程中均匀化矩阵的一些数值近似。我们关注的是底层随机场是参考周期张量的小扰动的情况。这种扰动的大小由一个实参数 eta 编码。在这种情况下,文献中已经理论上证明,对于本文考虑的两种模型,精确的均匀化矩阵具有关于参数 eta 的幂次展开,其系数是确定性的。在实践中,我们无法操作这种展开的精确项。所有对象都受限于离散化方法。因此,我们需要推导近似随机均匀化矩阵的类似展开。与精确均匀化矩阵的展开相反,近似均匀化矩阵的展开包含本质上的随机系数。特别地,二阶项本质上是随机的。本文的目的是根据近似过程的参数(使用的截断计算域的大小、有限元近似的网格尺寸)推导并研究这一展开。
英文摘要:
This article studies some numerical approximations of the homogenized matrix for stochastic linear elliptic partial differential equations in divergence form. We focus on the case when the underlying random field is a small perturbation of a reference periodic tensor. The size of such a perturbation is encoded by a real parameter eta. In this case, it has already been theoretically shown in the literature that the exact homogenized matrix possesses an expansion in powers of the parameter eta for both models considered in this article, the coefficients of which are deterministic. In practice, one cannot manipulate the exact terms of such an expansion. All objects are subjected to a discretization approach. Thus we need to derive a similar expansion for the approximated random homogenized matrix. In contrast to the expansion of the exact homogenized matrix, the expansion of the approximated homogenized matrix contains intrinsically random coefficients. In particular, the second order term is random in nature. The purpose of this work is to derive and study this expansion in function of the parameters of the approximation procedure (size of the truncated computational domain used, meshsize of the finite elements approximation).
Multi-Level Monte Carlo approaches for numerical homogenization
Texas A & M University(德克萨斯农工大学);Fraunhofer ITWM(弗劳恩霍夫工业数学研究所);University of Kaiserslautern(凯泽斯劳滕工业大学);École des Ponts ParisTech(巴黎路桥学院);Université Paris-Est(巴黎东大学);INRIA Rocquencourt(法国国家信息与自动化研究所罗康库尔中心)
Random walk in random environment, corrector equation, and homogenized coefficients: from theory to numerics, back and forth
INRIA Paris-Rocquencourt(法国国家信息与自动化研究所巴黎罗克安古尔分部);Université Libre de Bruxelles (ULB)(布鲁塞尔自由大学);INRIA Lille - Nord Europe(法国国家信息与自动化研究所里尔北欧分部);EPFL - Institut de mathématiques(洛桑联邦理工学院数学研究所);CMAP École polytechnique(巴黎综合理工学院计算与应用数学中心)