In the era of digital transformation, data-driven tools are vital for enhancing materials development and component design. Central to these data-driven tools are dataspaces, which streamline data management and data sharing, fostering integration and collaboration among stakeholders. By leveraging dataspaces, organizations can homogenize materials and process data from different sources and adhere to the FAIR (Findable, Accessible, Interoperable, and Reusable) principles of data handling, effectively utilizing data to drive innovation. This contribution showcases the effective usage of a dataspace platform with integrated machine learning applications to train models and apply them to predict creep performance parameters based on accelerated creep tests. Key challenges in developing such machine learning models using data from numerical simulations are addressed including the definition of a suitable parameter space, efficient dataset generation, and the representation of creep curves in a lower-dimensional feature space.

Creep is an industrially significant material behavior that describes the time-dependent deformation of materials under constant stress, particularly at elevated temperatures. Understanding creep performance is crucial for the design and longevity of components in high-temperature applications, such as in power plants or aerospace. Standard creep tests are time-consuming and cumbersome, posing significant challenges in conducting tests and slowing down materials and component design processes. Therefore, this contribution focuses on accelerated creep tests that offer faster results on the one hand, while on the other, they do not provide direct identification of creep performance parameters as with standard tests.