In various sectors, such as manufacturing and engineering, data generation for data-driven models is expensive and time-consuming. For instance, material sciences often resort to destructive testing, while resource-intensive simulations, though an alternative to physical testing, pose computational demands. This project aims to address these challenges by developing machine learning methods that enhance accessibility and cost-effectiveness in data generation for resource-constrained settings.

As the range of data platforms is rapidly evolving, the goal of this research project is to investigate and prototype upcoming architectures and technologies and to assess their applicability and potentials for industrial enterprises.

While there are many concepts, techniques and tools for metadata management, most focus on sub-aspects, e.g., metadata management with semantic technologies. There is no common understanding of what comprehensive metadata management in an enterprise entails and how it can be implemented. It is the goal of this project to design concepts and techniques for comprehensive metadata management across the entire enterprise data landscape.