Transforming Mechanical Test Labs into Autonomous Knowledge Discovery Hubs

Abstract

The integration of machine learning, large language models, intelligent robotics, quantum computing, and advanced data acquisition is transforming traditional mechanical test labs into autonomous hubs for knowledge discovery. As we face pressing challenges like climate change and resource scarcity, leveraging these technologies in laboratory settings becomes essential for faster and more insightful scientific progress. This talk will explore an automated, data-driven approach to knowledge discovery through fatigue crack growth experiments. In this setup, intelligent robotic systems continuously track the crack tip of a fatigue crack, capturing high-resolution digital image correlation (DIC) data. A machine learning model then processes these datasets to automatically detect and assess crack tip positions and associated crack tip loadings. Feature extraction is enhanced by a combination of classical algorithms and artificial intelligence, creating a rich dataset of analyzed results like the evolution of the plastic zone or the fracture surface characteristics. To ensure the coherence and interoperability of diverse data sources, graph databases embedded with ontologies, semantics, and provenance data are employed. This structured approach enables the automated recognition of cause-and-effect relationships, consolidating knowledge within a unified knowledge graph. By streamlining data capture, analysis, and integration, autonomous labs hold the potential to significantly shorten development cycles, accelerating the path to market for new materials and products. This paradigm shift from manual testing to autonomous knowledge discovery not only advances the scientific method but also strengthens our capacity to address complex, global challenges with speed and precision.