[SY-L1] Objective fusion of multiscale experiments and multiscale models using Bayesian inference
There is currently no formal framework for fusing the information gathered from multiscale materials modeling and measurement efforts in ways that optimally inform each other. This is mainly because the space of governing physics in any selected multiscale materials phenomenon is extremely large (this includes all potential model forms and the ranges of parameter values needed to identify the governing physics as accurately as possible), and the amount of the relevant experimental data is typically limited, incomplete, and uncertain. Consequently, a direct calibration of the governing physics based on the available measurements using standard regression techniques usually does not produce reliable results. In this paper, we explore the benefits of applying Bayesian inference techniques combined with reduced-order models and higher-throughput experimental assays in establishing a mathematically rigorous framework for addressing the challenge identified above. More specifically, the new framework will be demonstrated with a very simple case study - the identification of the intrinsic single crystal material properties from spherical indentation stress-strain measured on a polycrystalline sample.