Recent applications of machine learning in alloy design: A review

Abstract

The history of machine learning (ML) can be traced back to the 1950 s, and its application in alloy design has recently begun to flourish and expand rapidly. The driving force behind this is partially due to the inefficiency of traditional methods in designing better-performing alloys, partially due to the success of ML in other areas and alloy data becoming more accessible. ML methods can quickly predict the properties of the alloy from the data and suggest compositions for particularly required properties, thereby minimizing the need for resource-intensive experiments or simulations. The present work provides a critical review of this domain starting with an introduction to ML components, followed by an overview of the forward prediction of alloy properties, and an elaboration of the inverse design of alloys. This paper aims to summarize crucial findings, reveal key trends, and provide guidance for future directions.