Data-driven glass-forming ability criterion for bulk amorphous metals with data augmentation

Abstract

• A data augmentation technique is employed for regression problems in the current work, and the machine learning model for predicting glass-forming ability of bulk metallic glasses are developed on the augmented training dataset. The data augmentation technique enhances the predictive accuracy greatly. • A new feature T g T x ( T l − T g ) ( T l − T x ) is proposed based on domain knowledge, and this feature has significant importance in predicting glass-forming ability of bulk metallic glasses. • The novel ML model gives the glass forming ability rules: the average atomic radius ranging from 140 pm to 165 pm, the value of T g T x ( T l − T g ) ( T l − T x ) higher than 2.5, the entropy of mixing higher than 10 J/K/mol, and the enthalpy of mixing ranging from −32 kJ/mol to −26 kJ/mol. A data augmentation technique is employed in the current work on a training dataset of 610 bulk metallic glasses (BMGs), which are randomly selected from 762 collected data. An ensemble machine learning (ML) model is developed on augmented training dataset and tested by the rest 152 data. The result shows that ML model has the ability to predict the maximal diameter D max of BMGs more accurate than all reported ML models. In addition, the novel ML model gives the glass forming ability (GFA) rules: average atomic radius ranging from 140 pm to 165 pm, the value of T g T x ( T l − T g ) ( T l − T x ) being higher than 2.5, the entropy of mixing being higher than 10 J/K/mol, and the enthalpy of mixing ranging from -32 kJ/mol to -26 kJ/mol. ML model is interpretative, thereby deepening the understanding of GFA.