Computational Modeling of a Solid Electrolyte using Convolutional Neural Network (CNN) Machine Learning

Project Description

Project Overview:  Solid electrolytes are the next frontier in battery technology however, due to their complex design, their fabrication is costly. This provides a hurdle in the study of this class of materials, as even measuring conductivity of a novel material requires fabrication which is slow and expensive. Computational Materials modeling is an emerging field of materials science that seeks to address these issues by creating tools to accurately predict the properties of novel new age materials such as solid electrolytes and allow researchers to screen material before allocating capital to their fabrication.  The project team will be generating data via physical material modeling from known solid electrolytes to train convolutional neural networks to predict the conductivity of these materials, with the objective of creating a neural network the can accurately and robustly predict the conductivity of solid electrolytes.  The project will run phase field simulations using UW-Madison’s Center for High Throughput Computing (CHTC) and the CNN Python code (with Google Colaboratory tool)  to help disseminate machine learning.

 Client:  Dr. Jiamian Hu, Department of Materials Science and Engineering University of Wisconsin-Madison

 Student Team:  Patrick Stiles, Josh Emory, Alexis Tousignant, and Michael Orth