On the Convergence of a General Class of Finite Volume Methods

DOI10.1137/040612993MaRDI QIDQ5700303zbMATH OpenOpenAlexFDO

Publication date 28 October 2005

Published in SIAM Journal on Numerical Analysis (Search for Journal in Brave)

Full work available at URL https://semanticscholar.org/paper/77ef242f1e940772175e94c9fd9c7a71f4c33da4, https://figshare.com/articles/journal_contribution/On_the_convergence_of_a_general_class_of_finite_volume_methods/23347931

zbMATH Keywords

approximation orders convergence error analysis essentially nonoscillatory schemes finite volume methods higher order schemes multivariate polynomials numerical examples optimal recovery positive definite kernels reconstruction scheme thin-plate spline approximation

Mathematics Subject Classification ID

Hyperbolic conservation laws (35L65) Finite difference methods for initial value and initial-boundary value problems involving PDEs (65M06) Stability and convergence of numerical methods for initial value and initial-boundary value problems involving PDEs (65M12) Error bounds for initial value and initial-boundary value problems involving PDEs (65M15)

Review text

For discretizing in space, finite volume methods use cell average information. To be more precise, for a fixed time such cell averages are employed to reconstruct the unknown function approximately. For a good reconstruction in regions where the solution is known or expected to be smooth, a higher order reconstruction scheme is desirable. The first higher order reconstruction schemes employed were based on polynomials and suffered from the typical behavior of multivariate polynomials, such as oscillation and ill-conditioning. It was proposed to employ optimal recovery based on conditionally positive definite kernels instead. Some numerical examples indicate that these recovery processes indeed lead to higher order schemes. It is the goal of this paper to show that the recovery process can lead to arbitrary high orders, provided the target function is sufficiently smooth and the correct (conditionally) positive definite kernel is employed. However, since the analysis is based upon approximation properties of polynomials, the proof will need slightly larger stencils than those proposed earlier by \textit{T. Sonar} [IMA J. Numer. Anal. 16, No.~4, 549--581 (1996; Zbl 0860.65096); Comput. Methods Appl. Mech. Eng. 140, No.1--2, 157--181 (1997; Zbl 0898.76086); SIAM J. Numer. Anal. 35, No.~6, 2350--2369 (1998; Zbl 0921.65069)]. On the other hand, since the ``correct selection of stencils is still under investigation, the author's results might also contribute to this problem. Moreover, these results are not restricted to (conditionally) positive definite kernels at all. On the contrary, they will work for every interpolatory and stable reconstruction process. Finally, the results are established for an arbitrary space dimension. This paper is organized as follows. In the Section 1 some general notations which are necessary to state the convergence results are introduced. Section 2 is devoted to a short review on finite volume and ENO (essentially nonoscillatory) schemes. Section 3 describes how such schemes can be derived using optimal recovery. Sections 4 and 5 are the main sections where the error analysis is provided. In section 6 the author takes a special look at thin-plate spline approximation, which is one of the most popular reconstruction methods. For numerical examples the reader is refered to the papers by T. Sonar [loc.cit.].

Recommendations

Cited in

(16)

This page was built for publication: On the Convergence of a General Class of Finite Volume Methods

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5700303)