Project Description
Materials Morphology Prediction with Deep Learning
Faculty Advisor: Dr. Mahmood Mamivand
In this project, we aim to accelerate the process of morphology prediction for materials. Classically, these morphologies are predicted via physics-based models, but they are computationally expensive. We aspire to make these predictions live. In this project, we explore the applications of generative adversarial networks (GAN), a class of emerging machine learning frameworks, to accelerate the prediction of microstructural evolution.
Role of Participant(s):
In this project, students learn high-fidelity models such as the phase-field method to generate databases for machine learning algorithms. Then students learn to quantify the morphology of the images with deep learning algorithms and finally use the ML techniques to correlate the microstructures to material parameters. Ideally, we would use the developed algorithms to read experimental transmission electron microscopy images to predict material parameters.