Accelerating Microstructure-to-Property Computations of Metallic Materials through Physics-Enhanced Deep Learning Techniques - with Special Application to Nickel-Based Superalloys

• Applicants: Dr.-Ing. Tim Brepols; Shahed Rezaei
• Subject Area: Mechanics ; Applied Mechanics, Statics and Dynamics
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 561202254

The goal of this project is to reduce the huge computational effort in multiscale mechanical modeling of materials with complex microstructures. This will be realized by a novel approach for creating and training fast surrogates for material models and solvers. The suggested new approach is based on advanced deep learning techniques, a specific branch of machine learning. At all stages of the model development process, great emphasis is placed on the integration of physical constraints as well as data from already established and trusted methods. This involves e.g. performing multi-component, multi-phase field simulations and physics-based full-field finite element simulations to generate a valuable database for the relationships between the underlying microstructure and overall effective properties. In this way, and by leveraging and advancing concepts from the field of physics-informed deep operator learning, accurate and realistic "structure-property relationships" will become feasible. Utilizing rapid surrogates significantly enhances the overall future design process of materials, making it more economical and efficient. The latter is achieved by substantially reducing the computational cost for predictions and minimizing the required memory for storing and solving equations in the traditional approach. The material focus of this project is on a nickel-based superalloy for high temperature applications. Such Ni-based superalloys are widely used in various industries and are therefore relevant examples. However, the suggested methodology itself is general in nature and will be applicable to any other heterogeneous material system.

• DFG Programme: Research Grants
• Co-Investigator: Professorin Dr.-Ing. Stefanie Reese