An efficient hierarchical multiscale finite element method for stokes equations in slowly varying media
Direct numerical simulation (DNS) of fluid flow in porous media with many scales is often not feasible, and an effective or homogenized description is more desirable. To construct the homogenized equations, effective properties must be computed. Computation of effective properties for nonperiodic...
Saved in:
Main Authors: | , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2014
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/100579 http://hdl.handle.net/10220/18461 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | Direct numerical simulation (DNS) of fluid flow in porous media with many scales is
often not feasible, and an effective or homogenized description is more desirable. To construct the
homogenized equations, effective properties must be computed. Computation of effective properties
for nonperiodic microstructures can be prohibitively expensive, as many local cell problems must be
solved for different macroscopic points. In addition, the local problems may also be computationally
expensive. When the microstructure varies slowly, we develop an efficient numerical method for
two scales that achieves essentially the same accuracy as that for the full resolution solve of every
local cell problem. In this method, we build a dense hierarchy of macroscopic grid points and a
corresponding nested sequence of approximation spaces. Essentially, solutions computed in high
accuracy approximation spaces at select points in the the hierarchy are used as corrections for the
error of the lower accuracy approximation spaces at nearby macroscopic points. We give a brief
overview of slowly varying media and formal Stokes homogenization in such domains. We present a
general outline of the algorithm and list reasonable and easily verifiable assumptions on the PDEs,
geometry, and approximation spaces. With these assumptions, we achieve the same accuracy as
the full solve. To demonstrate the elements of the proof of the error estimate, we use a hierarchy
of macro–grid points in [0, 1]2 and finite element (FE) approximation spaces in [0, 1]2. We apply
this algorithm to Stokes equations in a slowly porous medium where the microstructure is obtained
from a reference periodic domain by a known smooth map. Using the arbitrary Lagrange–Eulerian
(ALE) formulation of the Stokes equations (cf. [G. P. Galdi and R. Rannacher, Fundamental Trends
in Fluid-Structure Interaction, Contemporary Challenges in Mathematical Fluid Dynamics and Its
Applications 1, World Scientific, Singapore, 2010]), we obtain modified Stokes equations with varying
coefficients in the periodic domain. We show that the algorithm can be utilized in this setting. Finally,
we implement the algorithm on the modified Stokes equations, using a simple stretch deformation
mapping, and compute the effective permeability. We show that our efficient computation is of the
same order as the full solve. |
---|