Grain-Boundary Graph Convolutional Recurrent Neural Network Model
Spatio-temporal model predicting abnormal grain growth in material microstructures from Monte Carlo simulations.
May 2021 – May 2022 · Nano-Human Interfaces Laboratory, Lehigh University
In my research at the Nano-Human Interfaces Lab, I explored material microstructures while studying challenges in grain-boundary analysis. My focus was on predicting abnormal grain growth, which can significantly alter the way we understand material properties.
I used a combination of spatial and temporal neural networks to predict when these abnormalities would occur based on previous sequences. Specifically, I developed a pipelined Graph Convolutional Network (GCN) and Recurrent Neural Network (RNN) to analyze Monte Carlo simulations of a Grain Boundary Potts model. This allowed me to predict when grain abnormality would occur in the material.
