Robustness of specimen design criteria for identification of anisotropic mechanical behaviour from heterogeneous mechanical fields

Abstract

To reduce the number of tests required to characterize anisotropic elastoplastic constitutive models, an approach is to design specimen geometries to diversify the stress states generated in a single test. The experiments are then processed using an inverse identification method based on full-field measurements to achieve the full potential of that specific test. Recent optimization methods were able to design complex specimens in which highly heterogeneous stress fields were generated. However, the specimen design is only assessed based on numerical simulations and does not consider the effect of the biases introduced by the full-field measurement method. The goal of this work is therefore to take into account some of the most frequently observed measurement biases in the specimen selection process. The proposed approach uses synthetic test images generated with numerical simulations. Four specimen geometries have been ranked based on two selection criteria. The first one is an indicator of the heterogeneity of the stress fields obtained by finite element simulations (unbiased data). The second one quantifies error for the identification procedure due to measurement biases. The two criteria provide different rankings for the set of specimens. It is concluded that the design with the most heterogeneous stress fields (first criterion) is not necessarily the more robust design in terms of measurement noise (second criterion), so the optimized geometry should be selected based on a compromise between these two criteria.