A posteriori finite element error estimation for second-order hyperbolic problems. (Q1867612): Difference between revisions

The author develops a posteriori finite element discretization error estimates for the second-order wave equation in one and two dimensions. A posteriori error estimates are an integral part of adaptive procedures for solving partial differential equations. They are used to steer the adaptive process by indicating regions where more or less resolution is needed and to control solution accuracy and stop the adaptive process. Herein a posteriori estimates for the spatial bi-\(p\) finite element approximations are presented and it is proved that they are asymptotically correct under mesh refinement. A method of construction of elliptic and hyperbolic error estimations is given. In one dimension it is shown that significant part of the spatial finite element error is proportional to a Lobatto polynomial and an error estimate is obtained by solving a set of either local elliptic or hyperbolic problems. For even-degree approximations error estimates are computed by solving a set of either local elliptic or hyperbolic problems and for odd degree approximations an error estimate is computed using jumps of solution gradients across element boundaries. Although the theory is not developed for non-rectangular meshes, the author expects good results with bi-\(p\) and hierarchical approximations when the solution is smooth and the mesh is not severely distorted. Indeed, effectivity indices for the bi-\(p\) and hierarchical bases are in ascess of 0.90 for virtually all mesh-order combinations. The effectivity indices of both these approximations appear to converge to unity under mesh refinement. The data provided, in the tabular form, indicate a better performance for even-degree polynomials than for odd. Furthermore, the even degree estimator does not involve communication with neighbouring elements and thus is ideal for parallelization. This paper extends the superconvergence results for elliptic and parabolic problems worked out by \textit{L. B. Wahlbin} [Superconvergence in Galerkin finite element methods, Springer, New York (1995; Zbl 0826.65092)], to second-order hyperbolic problems, and the semi-discrete finite element solution is superconvergent at Lobatto points. The author expects the error estimate to apply to more efficient higher-order finite element approximatoins with lumped mass matrix. Computational results indicate that the hyperbolic error estimates are slightly more accurate than their elliptic counterparts with some additional computational effort.