Multiscale methods for parabolic equations with continuum spatial scales (Q2475516)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Multiscale methods for parabolic equations with continuum spatial scales
scientific article

    Statements

    Multiscale methods for parabolic equations with continuum spatial scales (English)
    0 references
    0 references
    0 references
    0 references
    11 March 2008
    0 references
    Parabolic equation with rough in space diffusion coefficient together with initial condition is investigated in this paper. These types of problems are now of great interest because they arise in many porous media applications, e.g., in compressible flow, Richards equation and many others. The aim of this paper is to develop new finite element methods for the proposed problem. Authors present two numerical multiscale methods using global information. First method-Galerkin multiscale finite element method (MsFEM) with limited global information is based on construction of special basis function for coarse grid. Such basis functions are constructed from the solution of the elliptic equation with the same boundary condition as the original problem poses. This problem is solved on the fine-scale grid by using standard finite element method and then this solution is used for construction of the special basis functions. Second type of discretization-Partition of Unity Method (PUM) is based on finding an accurate approximation of the solution in each ``patch'' and then use partition of unity functions to collect those ``patch'' approximation together. For both discretization schemes convergence to the continuous solution independent of small scales is proved. Comparing the convergence rate of obtained results authors derived that convergence rate by PUM method is slightly less than that given by Galerkin MsFEM. Time discretization is based on backward Euler scheme. Numerical experiments for fully disretization based on Galerkin MsFEM conclude the paper. Obtained numerical results confirm that the solution can be captured more accurately on coarse grid when some type of limited global information is used.
    0 references
    global information
    0 references
    continuum scale
    0 references
    backward Euler scheme
    0 references
    Galerkin MsFEM
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references