Sprecher
Beschreibung
Traditional mechanical testing often relies on manual observation and fragmented data storage, creating bottlenecks in scientific progress. To reduce development times and make mechanical testing more sustainable, we must transition from manual logging to high-throughput, standardized data acquisition. This presentation demonstrates a paradigm shift in human-machine collaboration within fatigue crack growth experiments.
We present an autonomous data acquisition framework for fatigue crack growth experiments in which intelligent robotics, Digital Image Correlation (DIC), and machine learning (ML) operate as closed-loop sensing agents. High-resolution DIC continuously tracks crack tip position and deformation fields, while ML models extract higher-order descriptors, including plastic zone evolution and fracture-relevant damage features, in real time.
Central to the framework is a semantic orchestration layer based on graph databases, domain ontologies, and explicit provenance models. Experimental parameters, sensor states, derived features, and processing steps are represented as first-class entities in a unified knowledge graph. This enables automated metadata capture, cross-modal data alignment, and machine-driven reasoning over experimental context, rather than post-hoc annotation.
By decoupling experimental execution from semantic interpretation, the framework transforms mechanical testing into a self-describing, machine-navigable process. The result is an autonomous experimental pipeline that supports scalable data generation, reproducible analysis, and seamless integration into self-driving laboratory workflows.
