| Abstract Scope |
Conventional mechanical testing protocols produce high quality, but low quantity, data on the material constitutive response. Generally, they are low-throughput and need large material volumes of consistent material state. Moreover, significant uncertainties continue to persist in the constitutive model forms and parameters that cannot be resolved in a cost-effective manner. Our research aims to develop novel protocols that fundamentally change the traditional paradigm by leveraging non-standard high-throughput test protocols that utilize very small material volumes, while being amenable to fully autonomous explorations of vast materials design spaces (includes variations in both materials compositions and process histories). These new protocols are also designed to leverage data produced by established physics-based simulations (e.g., finite element models incorporating a variety of material constitutive laws) and emergent machine learning approach (e.g., Bayesian inference on unknown model forms and parameter values in candidate material constitutive laws, active learning). |