A machine learning-driven framework for the property prediction and generative design of multiple principal element alloys

Abstract

Multi-principal element alloys (MPEAs), inclusive of so-called high entropy alloys (HEAs), represent an innovative class of metallic materials that reveal unique properties and potentially broad applicability. However, the compositional complexity of MPEAs presents challenges in discerning physical mechanisms that control properties, and in harnessing such mechanisms to drive the design of new alloys. An ability to design metallic alloys that possess user-defined requisite properties has emerged as a critical area of interest within the field of materials science and engineering. This research illustrates how the integration of data science, machine learning (ML), and generative design strategies can evolve alloy design to a predictive, data-oriented approach. A comprehensive workflow utilising machine learning for feature analysis, property prediction, and generative design of novel MPEA generation is proposed. This workflow facilitates the examination and comparison of the predictive capabilities of ML models in determining the mechanical properties of MPEAs, providing insights into the influence of different design parameters. Additionally, by integrating a generative adversarial model, the prediction of novel MPEAs and anticipate their mechanical behaviours is revealed.