LEARNING CURVES: Schooling Machine Learning in Materials Science

Описание: About Dr. Rémi Dingreville presentation:

Materials science has always been about mastering the lessons inscribed in the intricate interplay between processing, structure, and properties. Yet this process–structure–property paradigm is really complex, high-dimensional, nonlinear, and often beyond the reach of human intuition.

Machine learning (ML) is reshaping how we confront this complexity, not only shortening our learning curves but at times rewriting the very syllabus of discovery. In this talk, I will introduce advanced ML workflows that are “schooled” to detect microstructural fingerprints in large, multimodal datasets and to link them directly to material performance. Beyond accelerating simulations, these approaches create feedback loops that adapt processing routes, improving properties by design rather than trial and error.

I will highlight case studies ranging from data-driven pattern recognition to experiment–simulation hybrids and physics-informed models, showing how each approach teaches us something different about material behavior. Together, these strategies illustrate how ML can both accelerate discovery and deepen understanding—transforming materials science from observation-driven to intelligence-driven innovation. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525. SAND2025-10343A.