Table of Contents
Fetching ...

A Semismooth Newton Solver and its application to an $hp$-FE Discretization in Elastoplasticity

Patrick Bammer, Lothar Banz, Miriam Schönauer, Andreas Schröder

TL;DR

This work develops a semismooth Newton solver for a broad class of nonsmooth nonlinear systems arising from $hp$-finite element discretizations in elastoplasticity. By exploiting a structured Clarke subdifferential and the notion of eigencomplementarity among block submatrices, the authors establish well-definedness and local convergence, with a $1+\alpha$-order rate when the nonsmoothities are $\alpha$-order semismooth. They apply the framework to a mixed variational formulation of elastoplasticity with linear kinematic hardening, using biorthogonal bases to decouple constraints and obtain a decoupled nonlinear system that fits the theoretical setting. Numerical experiments on uniform, p-, and hp-adaptive meshes confirm robustness with respect to $h$ and $p$ and corroborate superlinear convergence, with observed rates around $4/3$ in practice.

Abstract

In this paper, we consider a class of systems of nonlinear equations, which arise in discretized mixed formulations of problems in solid mechanics by $hp$-finite elements. We introduce a semismooth Newton solver for this specific class and prove its well-definedness and local convergence. Thereby, the analysis heavily relies on a special eigenvalue interplay of two matrices involved in the considered nonlinear system. Next, we apply the general results to an $hp$-finite element discretization of a problem in elastoplasticity, which can be formulated as a system of nonlinear equations of the above type by using biorthogonal basis functions. Finally, numerical examples demonstrate the applicability and robustness of the proposed semismooth Newton solver with respect to $h$ and $p$.

A Semismooth Newton Solver and its application to an $hp$-FE Discretization in Elastoplasticity

TL;DR

This work develops a semismooth Newton solver for a broad class of nonsmooth nonlinear systems arising from -finite element discretizations in elastoplasticity. By exploiting a structured Clarke subdifferential and the notion of eigencomplementarity among block submatrices, the authors establish well-definedness and local convergence, with a -order rate when the nonsmoothities are -order semismooth. They apply the framework to a mixed variational formulation of elastoplasticity with linear kinematic hardening, using biorthogonal bases to decouple constraints and obtain a decoupled nonlinear system that fits the theoretical setting. Numerical experiments on uniform, p-, and hp-adaptive meshes confirm robustness with respect to and and corroborate superlinear convergence, with observed rates around in practice.

Abstract

In this paper, we consider a class of systems of nonlinear equations, which arise in discretized mixed formulations of problems in solid mechanics by -finite elements. We introduce a semismooth Newton solver for this specific class and prove its well-definedness and local convergence. Thereby, the analysis heavily relies on a special eigenvalue interplay of two matrices involved in the considered nonlinear system. Next, we apply the general results to an -finite element discretization of a problem in elastoplasticity, which can be formulated as a system of nonlinear equations of the above type by using biorthogonal basis functions. Finally, numerical examples demonstrate the applicability and robustness of the proposed semismooth Newton solver with respect to and .

Paper Structure

This paper contains 12 sections, 8 theorems, 88 equations, 1 figure, 1 algorithm.

Key Result

Lemma 1

The function $\mathfrak{F}$ is Lipschitz-continuous and ($\alpha$-order) semismooth. Furthermore, any element $\boldsymbol{H}\in\partial\mathfrak{F}(\mathfrak{a},\mathfrak{b},\mathfrak{c})$ for arbitrary $(\mathfrak{a},\mathfrak{b},\mathfrak{c})^\top\in\mathbb R^{dM}\times \mathbb R^{LN}\times\mathb with block-diagonal matrices $\boldsymbol{X} = \operatorname{diag}(\boldsymbol{X}_1,\ldots,\boldsym

Figures (1)

  • Figure 1: Iteration number $k$ vs. $q_\text{SSN}(\rho)$.

Theorems & Definitions (19)

  • Lemma 1
  • proof
  • Remark 1
  • Definition 1
  • Remark 2
  • Lemma 2
  • proof
  • Lemma 3
  • proof
  • Lemma 4
  • ...and 9 more