Extensions of goal-oriented error estimation methods to simulations of highly-nonlinear response of shock-loaded elastomer-reinforced structures

Abstract

This paper describes extensions of goal-oriented methods for a posteriori error estimation and control of numerical approximation to a class of highly-nonlinear problems in computational solid mechanics. An updated Lagrangian formulation of the dynamical, large-deformation response of structures composed of strain-rate-sensitive elastomers and elastoplastic materials is developed. To apply the theory of goal-oriented error estimation, a backward-in-time dual formulation of these problems is derived, and residual error estimators for meaningful quantities of interest are established. The target problem class is that of axisymmetric deformations of layered elastomer-reinforced shells-of-revolution subjected to shock loading. Extensive numerical results on solutions of representative problems are given. It is shown that extensions of the theory of goal-oriented error estimation can be developed and applied effectively to a class of highly-nonlinear, multi-physics problems in solid and structural mechanics.