Work

Informatics-driven mechanical characterization toward autonomous materials research

Year: 2026

Type: article

Abstract: Mechanical characterization remains one of the principal efficiency-limiting steps in the materials development and deployment pipeline. The traditional paradigm, based on standardized tests using geometrically simple specimens followed by calibration and validation of pre-assumed constitutive laws, has proven remarkably effective for ensuring reproducibility and inter-laboratory comparability. However, it was not conceived for the combinatorial design spaces emerging from modern materials discovery frameworks. Each conventional test probes only a limited region of the admissible stress–strain space, whereas characterizing complex or anisotropic materials requires multiple specimens and loading configurations, increasing inter-specimen variability and compounding time and resource costs. Consequently, existing workflows remain structurally incompatible with the throughput, adaptability, and autonomy demanded by modern manufacturing and digital design systems. Recent advances in experimental and computational mechanics are reshaping what is achievable. Full-field measurement techniques now provide high-resolution full-field kinematic information per experiment, while developments in scientific computing and machine learning enable data-driven constitutive model discovery and strategies for autonomous systems. Despite progress, these components remain largely fragmented, and coherent frameworks linking experimental design, data integration, model discovery, and validation remain underdeveloped. This review identifies four interdependent pillars enabling autonomous mechanics-based materials characterization: informative experimental design using heterogeneous specimen geometries; multi-fidelity data integration for denoising, sparse reconstruction, and dimensionality reduction of full-field measurements; physics-informed constitutive model discovery enforcing thermodynamic admissibility; and closed-loop validation with adaptive feedback driven by residual model uncertainty. We outline a path toward autonomous characterization systems operating at the pace and scale required for materials engineering.

Source: Extreme Mechanics Letters

DOI: 10.1016/j.eml.2026.102511