Machine learning-based prediction of elastic properties of amorphous metal alloys

Abstract

The Young's modulus $E$ is the key mechanical property that determines the resistance of solids to tension/compression. In the present work, the correlation of the quantity $E$ with such characteristics as the total molar mass $M$ of alloy components, the number of components $n$ forming an alloy, the yield stress $\sigma_{y}$ and the glass transition temperature $T_{g}$ has been studied in detail based on a large set of empirical data for the Young's modulus of different amorphous metal alloys. It has been established that the values of the Young's modulus of metal alloys under normal conditions correlate with such a mechanical characteristic as the yield stress as well as with the glass transition temperature. As found, the specificity of the ``chemical formula'' of alloy, which is determined by molar mass $M$ and number of components $n$, does not affect on elasticity of the material. The machine learning algorithm identified both the quantities $M$ and $n$ as insignificant factors in determining $E$. A simple non-linear regression model is obtained that relates the Young's modulus with $T_{g}$ and $\sigma_{y}$, and this model correctly reproduces the experimental data for metal alloys of different types. This obtained regression model generalizes the previously presented empirical relation $E\simeq49.8\sigma_{y}$ for amorphous metal alloys.