Unveiling glass forming ability patterns in bulk metallic glasses via advanced machine learning approaches

Abstract

In the present study machine learning (ML) models were used to predict the Glass Forming Ability (GFA) of various bulk metallic glass (BMG) systems. A careful feature selection procedure was employed to identify important parameters that exert influence on glass formation. Linear Regression (LR), Random Forest model (RF), Support Vector Regression (SVR), and K-Nearest Neighbours (KNN) were used to estimate the compositional dependencies through numerous physical, thermodynamic, and topological characteristics on the GFA. Important parameters affecting the glass formation were selected utilising the rigorously curated feature subset RF model yields in Maximum R-squared value amongst all the models showing good predictability of Dmax. Further, three key features, Mismatch entropy (ΔSσ/kB), electronegativity difference (∆χ), and PHSS were identified which shows the highest R2 value (0.749) of Dmax prediction. Hence, this model has pinpointed the ideal glass-forming characteristics. The results obtained through the study will help to provide a better understanding of BMGs and their future development.