Using Machine Learning to Relate Grain Boundary Properties to SOAP ... | Aidan Gesch | RRS Live 2025 скачать в хорошем качестве

Using Machine Learning to Relate Grain Boundary Properties to SOAP ... | Aidan Gesch | RRS Live 2025

Aidan Gesch is a Senior Aerospace Engineering and Mechanics, General Business major in the Randall Research Scholars Program (RRSP). Their research project presentation, "Using Machine Learning to Relate Grain Boundary Properties to SOAP Descriptors," was completed under the advisement of Chongze Hu from the Aerospace Engineering and Mechanics Department. Project Description: A grain boundary is the interfacial region between misaligned crystals in a ceramic or metallic material, which has a large impact on the properties of the material at large. Many computational methods exist to analyze the properties of grain boundaries, but they are often slow and expensive. This project aims to characterize the advantage of using machine learning to predict the properties of a grain boundary based on its structure. Recent works have used the geometric definition of the grain boundary to predict its properties, but performance is improved by using the Smooth Overlap of Atomic Positions (SOAP) descriptor instead. The results of this project can be applied to any grain boundary material system, but the methods can be implemented to any project that traditionally uses atomistic simulations, so long as a sufficient training database exists.