Integrating machine learning with advanced processing and characterization for polycrystalline materials: a...

Area of Science:

• Materials Science
• Computational Materials Science
• Machine Learning

Background:

• Polycrystalline materials are crucial in many applications.
• Understanding their microstructure (grains, grain boundaries) is key to controlling properties.
• Traditional materials research is often slow and labor-intensive.

Purpose of the Study:

• To present novel machine learning (ML)-based methodologies for polycrystalline materials research.
• To integrate experimental and computational approaches for accelerated materials discovery.
• To demonstrate data-driven design for optimizing material properties.

Main Methods:

• High-energy milling and in situ electron microscopy for synthesis and observation.
• Phase-field modeling with Bayesian data assimilation for process simulation.
• Scanning precession electron diffraction and neural networks for microstructural analysis.
• Bayesian optimization (BOXVIA) for data-driven process design.

Main Results:

• Demonstrated ML methodologies for analyzing and predicting properties of polycrystalline materials.
• Successfully applied a data-driven design approach to an iron-based superconductor.
• Evaluated bulk magnetic properties of the designed superconductor.

Conclusions:

• Machine learning offers powerful tools for accelerating materials research and development.
• Integrated data-driven approaches can optimize the design of complex materials.
• Future work should focus on further challenges in data-driven material development and iron-based superconductors.