An exponential-fitting finite element method for convection-diffusion problems (Q628954): Difference between revisions

In this paper, the two-dimensional convection diffusion equation: (1) \( - \varepsilon \Delta u + \beta . \nabla u = f\) in \( \Omega\), \(u=g\) on \(\partial \Omega\) is considered. Here \(\Omega \in \mathbb{R}^2\) is a bounded domain with boundary \(\partial \Omega\), \(\varepsilon\) is the diffusion coefficient, \(\beta = (\beta_1, \beta_2)\) is the convection velocity, \(f\) is the source function, and \(g\) represents the value of \(u\) on the boundary. An exponential fitting (EF) finite element method (FEM) for the discretization of the model problem (1) is considered. The authors review two versions of (FEM)s for the convection diffusion problem: streamline upwind Petrov-Galerkin, and two variants of streamline-crosswind diffusion methods. All proposed schemes are based on piecewise bilinear elements. Moreover the Hemker's function is reviewed and EFFEM is presented. The exponential term of the basis function is associated with the diffusion coefficients, convection velocity and the mesh size \(h\) in order to get the solution exactly on the grid. The error of this scheme has order \(h^{1/2}\) in the energy norm and this scheme on the benchmark problems is examined. Main result: The EFFEM scheme by using exponential basic functions to fit the solution is introduced. An error estimate for the EFF has been provided. The authors have shown that the error bound for the EFFEM in the energy norm is in order of \(O(\sqrt{h/\varepsilon}e^{-ch/(2 \varepsilon)})\) which is much smaller than tradition Galerkin method or various Petrov-Galerkin methods. The numerical results also show the robustness of the EFFEM that it is much more accurate than SUPG method and SCD method, and has the similar performance to that of the MSD which uses the bilinear elements with optimal selections in parameterizing the quantity of the artificial diffusion in both streamline and crosswind directions. The numerical results presented here demonstrate the efficiency and accuracy of the method of the authors.