Mesh‐ and model adaptivity for NTFA and full‐field elasto‐plastic homogenization based on downwind and upwind approximations

Abstract To increase the quality of computational results for heterogeneous materials like fiber‐reinforced composites with Prandtl–Reuss‐type material laws, goal‐oriented measures of the adaptive finite element method coupled to model adaptivity is established. The former is an adaptive mesh refinement on the macroscale, which allows to control the spatial discretization errors. The latter is an efficient combination of a numerically low cost nonuniform transformation field analysis (NTFA) and numerically high cost full‐field elasto‐plastic homogenization methods on the microscale. The present contribution deals with the application of the concept of downwind and upwind approximations to a goal‐oriented a posteriori error estimator based on duality techniques by means of reduced order homogenization schemes like NTFA, and with accuracy and numerical efficiency of the proposed goal‐oriented adaptive framework. NTFA consists of an offline phase and an online phase. During the offline phase, some relevant information of the micro system under consideration is precomputed allowing a reduced set of equations to be solved in the online phase. Thus, NTFA leads to a quite efficient homogenization method but less accurate compared to the full‐field homogenization method which is characterized with a high computational demand for accounting nonlinear microstructural mechanisms. Due to nonlinearities and time‐dependency of plasticity, the estimation of error transport and error generation are obtained with a backward‐in‐time dual method despite a high demand on memory capacity. In this contribution, the dual problem is solved with a forward‐in‐time dual method that allows estimating the full error during the resolution of the primal problem without the need for extra memory capacity. Several numerical examples illustrate the effectiveness of the proposed adaptive approach based on downwind and upwind approximations.