Harley Uses Deep Learning to Understand Abnormal Grain Growth

Dr. Joel Harley is co-PI on a project that combines artificial intelligence and deep learning with materials and chemical science. The project, “Elucidating Grain Growth in Thermomagnetic Processed Materials with Transfer learning and Reinforcement Learning,” is led by Amanda Krause (UF MSE), and also includes Michael Tonks (UF MSE) and Micheal Kesler (Oakridge National Laboratory). The goal of the project is to combine deep learning with physics-driven models to elucidate one of the most fundamental, yet poorly understood, mechanisms in materials science: abnormal grain growth. A grain is small region of a material with the same crystal structure where all of the molecules are effectively aligned in the same direction.

The work will utilize machine learning to imitate the physical process of grain growth in the material based on observations from data. This can then be used to simulate and predict the grain growth. This advance aims to revolutionize our understanding about fundamental materials science concepts.

The $1.2M project is part of a $27.6M DOE funding initiative for targeted research in data science to accelerate discovery in chemistry and material sciences.