Experimental characterization of material strain-rate dependence based on full-field Data-Driven Identification

Mechanical characterization usually relies on standardized sample geometries where homogeneous state of strain and stress are prescribed. Hence, many tests are required to capture the material response over various loading conditions. Using complex geometry allows for exploring wider domain in a single test but would require to have access to local strains and stresses to feed models. In that context, digital image correlation and clustering technique can be used to formulate an inverse problem able to identify fields of stress tensors without a priori constitutive modelling. This study explores the performances of a rate-dependent formulation of such a data-driven stress identification method, for capturing using a single test, the monotonic high strain-rate dependent response of a mild steel alloy. After presenting the problem formulation and resolution framework, a digital twin of a high speed tensile test performed on a notched sample geometry is used to explore identification performances. It allows defining confidence intervals depending on multiple indicators (stress magnitude, multiaxiality) and evaluate the range of strain-rate levels simultaneously captured. The method is eventually applied to a real experiment instrumented with high spatial resolution ultra high speed camera. Stress tensor fields are identified, within a 10 $\text{\%}$ confidence over the major part of the sample, and its material rate-dependence is retrieved from 20 to 300 s⁻¹ and found in very good agreement with literature. This is the first experimental application of the DDI in a high strain-rate context. The proposed framework may substantially widen the sample design space for mechanical characterization but also allow for probing local stresses during dynamic localization processes where in-situ quantitative data are still missing.