An algorithm for stabilizing hybridizable discontinuous Galerkin methods for nonlinear elasticity

Bernardo Cockburn, Jiguang Shen

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

It is now a well known fact that hybridizable discontinuous Galerkin for nonlinear elasticity may not converge to the exact solution if their inter-element jumps are not properly penalized or, equivalently, if the values of their stabilization function are not suitably chosen. For example, if their stabilization function is of order one, the method generates spurious oscillations in the region in which the elastic moduli tensor is indefinite. This phenomenon disappears when the values of the stabilization function are increased. Mixed methods display the same problem, as their stabilization function is identically zero. Here, we find explicit formulas for the lower bound of the values of the stabilization function which allow us to avoid this unpleasant phenomenon. Our numerical experiments show that, when we use polynomials of degree k>0 for the approximate gradient, first-order Piola-Kirchhoff and displacement, all the approximations converge with the optimal order, k+1. They also show that, if we increase the polynomial degree of the local spaces by one and update the values of the stabilization function only for the elements for which the elasticity tensor is indefinite, we obtain that all approximate solutions converge with order k+1 and that a local post processing of the displacement converges with order k+2.

Original languageEnglish (US)
Article number100001
JournalResults in Applied Mathematics
Volume1
DOIs
StatePublished - Jun 1 2019

Fingerprint

Nonlinear Elasticity
Discontinuous Galerkin Method
Galerkin methods
Elasticity
Stabilization
Converge
Tensors
Tensor
Polynomials
Polynomial
Discontinuous Galerkin
Mixed Methods
Elastic Modulus
Post-processing
Explicit Formula
Jump
Approximate Solution
Exact Solution
Elastic moduli
Update

Cite this

An algorithm for stabilizing hybridizable discontinuous Galerkin methods for nonlinear elasticity. / Cockburn, Bernardo; Shen, Jiguang.

In: Results in Applied Mathematics, Vol. 1, 100001, 01.06.2019.

Research output: Contribution to journalArticle

@article{9b9c99fe5c7b4ba28bdb456265d599b3,
title = "An algorithm for stabilizing hybridizable discontinuous Galerkin methods for nonlinear elasticity",
abstract = "It is now a well known fact that hybridizable discontinuous Galerkin for nonlinear elasticity may not converge to the exact solution if their inter-element jumps are not properly penalized or, equivalently, if the values of their stabilization function are not suitably chosen. For example, if their stabilization function is of order one, the method generates spurious oscillations in the region in which the elastic moduli tensor is indefinite. This phenomenon disappears when the values of the stabilization function are increased. Mixed methods display the same problem, as their stabilization function is identically zero. Here, we find explicit formulas for the lower bound of the values of the stabilization function which allow us to avoid this unpleasant phenomenon. Our numerical experiments show that, when we use polynomials of degree k>0 for the approximate gradient, first-order Piola-Kirchhoff and displacement, all the approximations converge with the optimal order, k+1. They also show that, if we increase the polynomial degree of the local spaces by one and update the values of the stabilization function only for the elements for which the elasticity tensor is indefinite, we obtain that all approximate solutions converge with order k+1 and that a local post processing of the displacement converges with order k+2.",
author = "Bernardo Cockburn and Jiguang Shen",
year = "2019",
month = "6",
day = "1",
doi = "10.1016/j.rinam.2019.01.001",
language = "English (US)",
volume = "1",
journal = "Results in Applied Mathematics",
issn = "2590-0374",

}

TY - JOUR

T1 - An algorithm for stabilizing hybridizable discontinuous Galerkin methods for nonlinear elasticity

AU - Cockburn, Bernardo

AU - Shen, Jiguang

PY - 2019/6/1

Y1 - 2019/6/1

N2 - It is now a well known fact that hybridizable discontinuous Galerkin for nonlinear elasticity may not converge to the exact solution if their inter-element jumps are not properly penalized or, equivalently, if the values of their stabilization function are not suitably chosen. For example, if their stabilization function is of order one, the method generates spurious oscillations in the region in which the elastic moduli tensor is indefinite. This phenomenon disappears when the values of the stabilization function are increased. Mixed methods display the same problem, as their stabilization function is identically zero. Here, we find explicit formulas for the lower bound of the values of the stabilization function which allow us to avoid this unpleasant phenomenon. Our numerical experiments show that, when we use polynomials of degree k>0 for the approximate gradient, first-order Piola-Kirchhoff and displacement, all the approximations converge with the optimal order, k+1. They also show that, if we increase the polynomial degree of the local spaces by one and update the values of the stabilization function only for the elements for which the elasticity tensor is indefinite, we obtain that all approximate solutions converge with order k+1 and that a local post processing of the displacement converges with order k+2.

AB - It is now a well known fact that hybridizable discontinuous Galerkin for nonlinear elasticity may not converge to the exact solution if their inter-element jumps are not properly penalized or, equivalently, if the values of their stabilization function are not suitably chosen. For example, if their stabilization function is of order one, the method generates spurious oscillations in the region in which the elastic moduli tensor is indefinite. This phenomenon disappears when the values of the stabilization function are increased. Mixed methods display the same problem, as their stabilization function is identically zero. Here, we find explicit formulas for the lower bound of the values of the stabilization function which allow us to avoid this unpleasant phenomenon. Our numerical experiments show that, when we use polynomials of degree k>0 for the approximate gradient, first-order Piola-Kirchhoff and displacement, all the approximations converge with the optimal order, k+1. They also show that, if we increase the polynomial degree of the local spaces by one and update the values of the stabilization function only for the elements for which the elasticity tensor is indefinite, we obtain that all approximate solutions converge with order k+1 and that a local post processing of the displacement converges with order k+2.

UR - http://www.scopus.com/inward/record.url?scp=85065210513&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85065210513&partnerID=8YFLogxK

U2 - 10.1016/j.rinam.2019.01.001

DO - 10.1016/j.rinam.2019.01.001

M3 - Article

VL - 1

JO - Results in Applied Mathematics

JF - Results in Applied Mathematics

SN - 2590-0374

M1 - 100001

ER -