Numerical solutions of strain localization with nonlocal softening plasticity

Xilin Lu, Jean-Pierre Bardet, Maosong Huang

Research output: Contribution to journalArticle

17 Citations (Scopus)

Abstract

Softening plasticity, when it is deprived of a characteristic length, often leads to boundary value problems that are ill-posed and produce unreliable numerical solutions. As meshes are refined, plastic strains localize into bands that become narrower and narrower, and generate load-displacement responses with excessive softening, which may even degenerate into unrealistic snapbacks. In the case of discrete systems, the underlying reasons for this poor numerical performance can be examined using a spectral analysis of the tangential stiffness matrix derived from incremental equilibrium equations. In one-dimensional boundary-valued problems, the introduction of a weak softening element into a mesh of elasto-plastic elements produces a dominant eigenvector that clearly exhibits a strain localization width directly related to the weak element size. The dominant eigenvector that controls the incremental solution varies spuriously when the mesh is refined, which results in mesh dependency. Over-nonlocal softening plasticity introduces a length scale, forces more elements to become plastic, and preserves the width of the plastic zone when the mesh is refined. When one-dimensional meshes are refined, spectral analysis shows that the dominant eigenvector does not vary erratically, and that its deformation patterns are smooth and display a constant localization width. This explains why the spatial distribution of plastic strains and the global load-displacement responses converge to analytical solutions in one dimension. The generalization to higher dimensions will be the object of future studies.

Original languageEnglish (US)
Pages (from-to)3702-3711
Number of pages10
JournalComputer Methods in Applied Mechanics and Engineering
Volume198
Issue number47-48
DOIs
StatePublished - Oct 1 2009
Externally publishedYes

Fingerprint

plastic properties
Eigenvalues and eigenfunctions
softening
Plasticity
mesh
Plastics
plastics
Spectrum analysis
Plastic deformation
eigenvectors
Stiffness matrix
Spatial distribution
Boundary value problems
spectrum analysis
stiffness matrix
equilibrium equations
boundary value problems
spatial distribution

Keywords

  • Characteristic length
  • Nonlocal
  • Softening plasticity
  • Spectral analysis
  • Strain localization

ASJC Scopus subject areas

  • Computer Science Applications
  • Computational Mechanics
  • Mechanics of Materials
  • Mechanical Engineering
  • Physics and Astronomy(all)

Cite this

Numerical solutions of strain localization with nonlocal softening plasticity. / Lu, Xilin; Bardet, Jean-Pierre; Huang, Maosong.

In: Computer Methods in Applied Mechanics and Engineering, Vol. 198, No. 47-48, 01.10.2009, p. 3702-3711.

Research output: Contribution to journalArticle

@article{a81a3b53c8774c3693e6989d7e4155c1,
title = "Numerical solutions of strain localization with nonlocal softening plasticity",
abstract = "Softening plasticity, when it is deprived of a characteristic length, often leads to boundary value problems that are ill-posed and produce unreliable numerical solutions. As meshes are refined, plastic strains localize into bands that become narrower and narrower, and generate load-displacement responses with excessive softening, which may even degenerate into unrealistic snapbacks. In the case of discrete systems, the underlying reasons for this poor numerical performance can be examined using a spectral analysis of the tangential stiffness matrix derived from incremental equilibrium equations. In one-dimensional boundary-valued problems, the introduction of a weak softening element into a mesh of elasto-plastic elements produces a dominant eigenvector that clearly exhibits a strain localization width directly related to the weak element size. The dominant eigenvector that controls the incremental solution varies spuriously when the mesh is refined, which results in mesh dependency. Over-nonlocal softening plasticity introduces a length scale, forces more elements to become plastic, and preserves the width of the plastic zone when the mesh is refined. When one-dimensional meshes are refined, spectral analysis shows that the dominant eigenvector does not vary erratically, and that its deformation patterns are smooth and display a constant localization width. This explains why the spatial distribution of plastic strains and the global load-displacement responses converge to analytical solutions in one dimension. The generalization to higher dimensions will be the object of future studies.",
keywords = "Characteristic length, Nonlocal, Softening plasticity, Spectral analysis, Strain localization",
author = "Xilin Lu and Jean-Pierre Bardet and Maosong Huang",
year = "2009",
month = "10",
day = "1",
doi = "10.1016/j.cma.2009.08.002",
language = "English (US)",
volume = "198",
pages = "3702--3711",
journal = "Computer Methods in Applied Mechanics and Engineering",
issn = "0374-2830",
publisher = "Elsevier",
number = "47-48",

}

TY - JOUR

T1 - Numerical solutions of strain localization with nonlocal softening plasticity

AU - Lu, Xilin

AU - Bardet, Jean-Pierre

AU - Huang, Maosong

PY - 2009/10/1

Y1 - 2009/10/1

N2 - Softening plasticity, when it is deprived of a characteristic length, often leads to boundary value problems that are ill-posed and produce unreliable numerical solutions. As meshes are refined, plastic strains localize into bands that become narrower and narrower, and generate load-displacement responses with excessive softening, which may even degenerate into unrealistic snapbacks. In the case of discrete systems, the underlying reasons for this poor numerical performance can be examined using a spectral analysis of the tangential stiffness matrix derived from incremental equilibrium equations. In one-dimensional boundary-valued problems, the introduction of a weak softening element into a mesh of elasto-plastic elements produces a dominant eigenvector that clearly exhibits a strain localization width directly related to the weak element size. The dominant eigenvector that controls the incremental solution varies spuriously when the mesh is refined, which results in mesh dependency. Over-nonlocal softening plasticity introduces a length scale, forces more elements to become plastic, and preserves the width of the plastic zone when the mesh is refined. When one-dimensional meshes are refined, spectral analysis shows that the dominant eigenvector does not vary erratically, and that its deformation patterns are smooth and display a constant localization width. This explains why the spatial distribution of plastic strains and the global load-displacement responses converge to analytical solutions in one dimension. The generalization to higher dimensions will be the object of future studies.

AB - Softening plasticity, when it is deprived of a characteristic length, often leads to boundary value problems that are ill-posed and produce unreliable numerical solutions. As meshes are refined, plastic strains localize into bands that become narrower and narrower, and generate load-displacement responses with excessive softening, which may even degenerate into unrealistic snapbacks. In the case of discrete systems, the underlying reasons for this poor numerical performance can be examined using a spectral analysis of the tangential stiffness matrix derived from incremental equilibrium equations. In one-dimensional boundary-valued problems, the introduction of a weak softening element into a mesh of elasto-plastic elements produces a dominant eigenvector that clearly exhibits a strain localization width directly related to the weak element size. The dominant eigenvector that controls the incremental solution varies spuriously when the mesh is refined, which results in mesh dependency. Over-nonlocal softening plasticity introduces a length scale, forces more elements to become plastic, and preserves the width of the plastic zone when the mesh is refined. When one-dimensional meshes are refined, spectral analysis shows that the dominant eigenvector does not vary erratically, and that its deformation patterns are smooth and display a constant localization width. This explains why the spatial distribution of plastic strains and the global load-displacement responses converge to analytical solutions in one dimension. The generalization to higher dimensions will be the object of future studies.

KW - Characteristic length

KW - Nonlocal

KW - Softening plasticity

KW - Spectral analysis

KW - Strain localization

UR - http://www.scopus.com/inward/record.url?scp=70349425798&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70349425798&partnerID=8YFLogxK

U2 - 10.1016/j.cma.2009.08.002

DO - 10.1016/j.cma.2009.08.002

M3 - Article

AN - SCOPUS:70349425798

VL - 198

SP - 3702

EP - 3711

JO - Computer Methods in Applied Mechanics and Engineering

JF - Computer Methods in Applied Mechanics and Engineering

SN - 0374-2830

IS - 47-48

ER -