Design and validation of a heterogeneous interior notched specimen for inverse material parameter identification

Nowadays, virtual manufacturing, digitalisation and simulation predictions are essential for the design and the development of engineering parts and components, as well as companies’ productivity. The implemented constitutive model and the accuracy of the determined material parameters influence the reliability of these predictions. It is necessary to calibrate multiple parameters of a complex and robust constitutive model. This is a very time-consuming task and involves high costs when using a classical approach, in which several different homogeneous mechanical experiments are performed. The non-homogeneous mechanical tests can provide a larger variety of mechanical information in just one experiment, reducing the number of required tests for material characterisation. The aim of this work is to numerically design a specimen with an interior notch for a uniaxial loading test that presents strain heterogeneity. The shape of the interior notch is optimised to maximise the strain heterogeneity. The optimisation procedure is guided by a cost function established by several strain heterogeneity indicators. The best specimen shape exhibits numerically uniaxial tension and compression, pure shear and plane strain tension in the plastic region. The identifiability of Swift’s hardening law parameters was quantitatively investigated based on the partial derivative of the strain fields with respect to the sought parameters. Results show that the proposed design strategy enables to increase in strain heterogeneity. It is shown that the optimised specimen has significantly better identifiability compared to a tensile specimen with a circular hole. Finally, a comparison between a classical Finite Element Model Updating (FEMU) methodology and a FEMU-based approach that uses a DIC-levelling strategy was conducted in order to inversely identify the Swift hardening law parameters and validate the designed specimen. It was confirmed that the DIC-levelling method finds more accurate results than the classical approach, with the compromise of the computational time. It was also shown that the optimum designed specimen is adequate for inverse identification strategies.