The research group "Microstructure - Data Science" deals with the data-driven characterization and optimization of microstructures. The main focus of the research work is on the development of characterization methods for microstructures as well as on the advancement and application of data-driven analysis tools that make the interaction between microstructural properties and macroscopic material behavior visible. For the generation of digital twins of microstructures and process chains, besides large-scale simulations with the phase-field method, generation algorithms for the creation of 3D microstructures with tailored properties are used. For this purpose, tools are developed on the basis of or with the integration of real microstructure images, which can realistically generate, for example, open-pored membrane structures, grain structures or rock fills. In cooperation with the research group "Research Data Management'' concepts for the highly automated handling and efficient evaluation of large data sets are developed and applied. The overall goal of the research questions includes bridging length scales by identifying macroscopic laws and developing predictive models as supporting component for an accelerated material design process.
Feature extraction in porous microstructures by a data science approach using a principal component analysis.