Error assessment and mesh adaptivity for regularized continuous failure models

This paper deals with the adaptive finite element analysis of structural failure. A gradient-enhanced damage model has been chosen to simulate material degradation. Since this model is regularized in the post-peak regime, the finite element solution does not suffer from pathological mesh dependence...

Descripción completa

Detalles Bibliográficos
Autores: Pannachet, T., Díez, Pedro|||0000-0001-6464-6407, Askes, Harm, Sluys, L. J.
Tipo de recurso: artículo
Fecha de publicación:2009
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/80363
Acceso en línea:https://hdl.handle.net/2117/80363
https://dx.doi.org/10.1016/j.cma.2009.11.010
Access Level:acceso abierto
Palabra clave:Numerical methods and algorithms
Gradient-enhanced damage model
Error estimation
Quantity of interest
Goal-oriented error estimation
Models, Teoria dels
Classificació AMS::74 Mechanics of deformable solids::74S Numerical methods
Àrees temàtiques de la UPC::Matemàtiques i estadística::Anàlisi numèrica::Mètodes numèrics
Descripción
Sumario:This paper deals with the adaptive finite element analysis of structural failure. A gradient-enhanced damage model has been chosen to simulate material degradation. Since this model is regularized in the post-peak regime, the finite element solution does not suffer from pathological mesh dependence and thus converges to an objective solution upon mesh refinement. However, the error analyses have shown that the error in the nonlocal equivalent strain field becomes dominant during the post-peak loading stages. The accuracy of the nonlocal equivalent strain field (and the corresponding damage quantity) also greatly influences the accuracy of the quantity of interest. Two error measures have been proposed. The goal-oriented error estimates have provided similar error distributions, although some small differences have been found in the softening regime. Objective error estimates, together with adaptive criteria, have been used to perform automated h-adaptivity during computation.