Materials data mining through microstructure graph embeddings – Towards thorough exploration of microstructural driving forces for degradation processes

• Applicant: Dr.-Ing. Ali Riza Durmaz
• Subject Area: Computer-Aided Design of Materials and Simulation of Materials Behaviour from Atomic to Microscopic Scale
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 550126120

The proposed work aims to assess the possibility of encoding the microstructure of a polycrystalline alloy through graphs in low-dimensional, yet expressive vector representations, so-called embeddings. Graph neuronal networks are used to produce such representations, which in turn can be used in downstream tasks. For instance, to predict the material's response to a specific mechanical loading or to perform materials design. While such vector representations are nowadays heavily used in natural language modeling or computer vision, the proposal aims to investigate whether the high-dimensional full-field microtexture can be treated and encoded similarly. Within the work, we combine unique competencies and data availability in the field of 3D materials characterization and data processing in the Pollock group of the University of California, Santa Barbara with competencies in graph machine learning of the applicant.

• DFG Programme: WBP Fellowship
• International Connection: USA
• Participating Institution: University of California, Santa Barbara
• Host: Professorin Tresa M. Pollock, Ph.D.