| 1. |
- Asadzadeh, Mohammad, 1952, et al.
(author)
-
A new approach to Richardson extrapolation in the finite element method for second order elliptic problems
- 2009
-
In: Mathematics of Computation. - 0025-5718 .- 1088-6842. ; 78:4, s. 1951-1973
-
Journal article (peer-reviewed)abstract
- This paper presents a nonstandard local approach to Richardson extrapolation, when it is used to increase the accuracy of the standard finite element approximation of solutions of second order elliptic boundary value problems in $ \mathbb{R}^N$, $ N \ge 2$. The main feature of the approach is that it does not rely on a traditional asymptotic error expansion, but rather depends on a more easily proved weaker a priori estimate, derived in [19], called an asymptotic error expansion inequality. In order to use this inequality to verify that the Richardson procedure works at a point, we require a local condition which links the different subspaces used for extrapolation. Roughly speaking, this condition says that the subspaces are similar about a point, i.e., any one of them can be made to locally coincide with another by a simple scaling of the independent variable about that point. Examples of finite element subspaces that occur in practice and satisfy this condition are given.
|
|
| 2. |
- Asadzadeh, Mohammad, 1952, et al.
(author)
-
Asymptotic error expansions for the finite element method for second order elliptic problems in R_N, N>=2, I: Local interior expansions
- 2010
-
In: SIAM Journal on Numerical Analysis. - : Society for Industrial & Applied Mathematics (SIAM). - 0036-1429 .- 1095-7170. ; 48:5, s. 2000-2017
-
Journal article (peer-reviewed)abstract
- Our aim here is to give sufficient conditions on the finite element spaces near a point so that the error in the finite element method for the function and its derivatives at the point have exact asymptotic expansions in terms of the mesh parameter h, valid for h sufficiently small. Such expansions are obtained from the so-called asymptotic expansion inequalities valid in RN for N ≥ 2, studies by Schatz in [Math. Comp., 67 (1998), pp. 877-899] and [SIAM J. Numer. Anal., 38 (2000), pp. 1269-1293].
|
|
