Using Machine Learning to Improve Material Properties Prediction in Glass Production
Event Type
DescriptionThe conventional way of designing new glasses has been an empirical scientific endeavor via expensive, inefficient trial-and-error for discovery; These constraints limit the speed that new compositional properties can be explored in the glass industry. However, machine learning can be leveraged to predict material properties prior to manufacture. In addition to percent composition, properties from molecular dynamics simulations can also be provided to machine learning algorithms to improve material property prediction. In this work, we implemented two learning models to explore the possibility of estimating the compositional values for the kinetic properties of glass. The macroscopic properties of glass include glass transition temperature (Tg) and density (p). Both properties have a significant impact on the eventual product and define the functionality of the glass. The objective of this study is to predict the density of the glass based on the composition. The results of ridge linear regression predict with a relatively low error of 0.05 (g·cm-3) root mean square for the multicomponent oxide glasses. The procedure used in this study can be extended to predict other macroscopic properties as a function of the glass composition. Such an approach can help accelerate the design of novel glasses with optimized properties.