Adaptive Finite Elements with Algebraic Stabilization for Convection-Dominated Transport

TL;DR

This paper investigates residual-based a posteriori error estimation for stabilized finite element methods applied to convection-diffusion equations, analyzing their performance on adaptive meshes for complex flow problems.

Contribution

It provides a comprehensive numerical study of various algebraic stabilization techniques and error estimators, revealing their interactions and practical behaviors in adaptive settings.

Findings

Residual estimators are reliable for linear and nonlinear problems.

Limiters' effectiveness depends on mesh alignment and convection field.

Strongly upwind-biased limiters yield more accurate solutions.

Abstract

We present a numerical investigation of residual-based a posteriori error estimation for finite element discretizations of convection--diffusion equations stabilized by algebraic flux correction and related algebraic stabilization techniques. In particular, we consider AFC schemes employing the BJK and Monolithic Convex (MC) limiters and algebraically stabilized methods including MUAS, SMUAS, and the BBK approach. The performance of the estimators and limiters are studied on adaptively refined meshes for several two-dimensional test problems, including boundary layers, interior layers, and a nonlinear convection problem with solution-dependent transport. The experiments assess accuracy, preservation of the discrete maximum principle, adaptive mesh behaviour, and computational efficiency. The results show that the interaction between stabilization and a posteriori error estimation depends strongly on mesh alignment and on the character of the convection field. In particular, for problems with moving or curved layers, the behaviour of the limiters differs significantly: strongly upwind-biased limiters provide the most accurate solutions, while smoother algebraic stabilizations lead to more efficient nonlinear iterations. The study also indicates that residual-based estimators remain reliable for both linear and nonlinear problems but may react to changes in limiter activation during adaptive refinement. Overall, the numerical results clarify the practical behaviour of several widely used stabilization techniques within an adaptive framework and highlight aspects that are not yet fully explained by the current theory, particularly for nonlinear transport problems.