Prediction of Reduced Glass Transition Temperature using Machine Learning

TL;DR

This paper explores machine learning algorithms to predict the reduced glass transition temperature (Trg) in materials, aiming to improve accuracy over empirical methods and aid in developing advanced materials.

Contribution

It evaluates various machine learning models for predicting Trg and finds ensemble models outperform others, offering a new approach for material property prediction.

Findings

Ensemble models outperform other algorithms in predicting Trg.

Machine learning can enhance the prediction of glass transition properties.

The study supports data-driven methods in materials science.

Abstract

The advent of computational material sciences has paved the way for data-driven approaches for modeling and fabrication of materials. The prediction of properties like the glass-forming ability (GFA) by using the variation in alloy composition remains to be a challenging problem in the field of material sciences. It also results in significant financial concerns for the manufacturing industry. Despite the existence of various empirical guides for the prediction of GFA, a comprehensive prediction model is still highly desirable. This work focuses on studying some of the popular machine learning algorithms for the prediction of the reduced glass transition temperature (Trg) of material compositions. From the experimentation, we conclude that the ensemble model performs better for predicting Trg. This result can prove instrumental in the branch of material sciences by helping us to develop materials having remarkable properties.