Bernstein-Bézier weight-adjusted discontinuous Galerkin methods for wave propagation in heterogeneous media

•A DG scheme for wave propagation in arbitrary heterogeneous media is presented.•Sparse matrix-based implementation of the Bernstein polynomial multiplication.•Representation of the Bernstein polynomial projection matrix in terms of sparse matrices.•Computational complexity is reduced from O(N2d) to...

Full description

Saved in:
Bibliographic Details
Published in:Journal of computational physics Vol. 400; p. 108971
Main Authors: Guo, Kaihang, Chan, Jesse
Format: Journal Article
Language:English
Published: Cambridge Elsevier Inc 01.01.2020
Elsevier Science Ltd
Subjects:
Algorithms
Bernstein
Complexity
Computational physics
Computer simulation
Discontinuous Galerkin
Galerkin method
GPU
Heterogeneous media
High order
Mathematical analysis
Multiplication
Polynomials
Propagation
Wave propagation
Weight
Discontinuous Galerkin
Bernstein
GPU
Heterogeneous media
High order
ISSN:0021-9991, 1090-2716
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•A DG scheme for wave propagation in arbitrary heterogeneous media is presented.•Sparse matrix-based implementation of the Bernstein polynomial multiplication.•Representation of the Bernstein polynomial projection matrix in terms of sparse matrices.•Computational complexity is reduced from O(N2d) to O(Nd+1) in d dimensions.•GPU implementation achieves significant speedups with respect to quadrature-base ap-proach. This paper presents an efficient discontinuous Galerkin method to simulate wave propagation in heterogeneous media with sub-cell variations. This method is based on a weight-adjusted discontinuous Galerkin method (WADG), which achieves high order accuracy for arbitrary heterogeneous media [1]. However, the computational cost of WADG grows rapidly with the order of approximation. In this work, we propose a Bernstein-Bézier weight-adjusted discontinuous Galerkin method (BBWADG) to address this cost. By approximating sub-cell heterogeneities by a fixed degree polynomial, the main steps of WADG can be expressed as polynomial multiplication and L2 projection, which we carry out using fast Bernstein algorithms. The proposed approach reduces the overall computational complexity from O(N2d) to O(Nd+1) in d dimensions. Numerical experiments illustrate the accuracy of the proposed approach, and computational experiments for a GPU implementation of BBWADG verify that this theoretical complexity is achieved in practice.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0021-9991
1090-2716
DOI:10.1016/j.jcp.2019.108971