A high order conservative flux optimization finite element method for steady convection-diffusion equations

This article presents a high order conservative flux optimization (CFO) finite element method that preserve the important mass conservation property locally for the convection-diffusion equations. The numerical scheme is an extension of the first order CFO scheme proposed in [25], it is based on the classical Galerkin finite element method enhanced by a flux approximation on the boundary of a prescribed set of arbitrary control volumes (either the finite element partition itself or dual voronoi mesh, etc). The numerical approximations can be characterized as the solution of a constrained-minimization problem with constraints given by the flux conservation equations on each control volume. The discrete linear system is a typical saddle-point problem, but with less number of degrees of freedom than the Raviart-Thomas (for k > 1) and the Brezzi-Douglas-Marini (for k > 3) mixed finite element methods, where kdenotes the degree of polynomial approximation spaces. Moreover, the numerical solution of the proposed scheme is of super-closeness with the finite element solution. Error estimates of optimal order are established for the numerical flux as well as the primary variable approximations. We present several numerical studies in order to verify convergence of the CFO schemes. Test cases under the presence of boundary layers, internal discontinuities, and a simplified two-phase flow in highly heterogeneous porous media model problem will also be presented. The numerical results show obvious advantages of applying high order CFO schemes. (C) 2020 Elsevier Inc. All rights reserved.