New recovery based a posteriori error estimators

Keywords

Advanced Numerical Methods for Scientific Computing
Code:
MOX 52
Title:
New recovery based a posteriori error estimators
Date:
Thursday 25th November 2004
Author(s):
Bottasso, Carlo L.; Maisano, Giorgio; Micheletti, Stefano; Perotto, Simona
Download link:
Abstract:
In this paper we formulate some new a posteriori recovery-based error estimators. The first one provides us with an improved approximation for the solution gradient. The other two furnish and estimate for the $L^2$-norm of the error on the solution itself. In more detail, the first estimator is a variant of the well-known Zienckiewicz-Zhu method and it turns out to be exact 1D for quadratic solution on non-uniform grids. The second one is based on a solution enrichment relying upon the Zienckiewicz-zhu recovered gradient. Finally the third estimator consists of a roughening of the solution followed by a Zienckiewicz-Zhu-like recovery applied to the solution itself. The three new proposed methods are compared in terms of their effectivity indices and solution accuracy on two and three dimensional problems
This report, or a modified version of it, has been also submitted to, or published on
Bottasso, C. L.; Maisano, G.; Micheletti, S.; Perotto, S., On some new recovery based a posteriori error estimators, Comput. Methods Appl. Mech. Engrg., 195 (2006), no. 37-40, 4794-4815