Prospective Revisions on the Use of Machine Learning in the Aluminum Alloy Development: A Review on Techniques and Future Applications

Alloy design is the major driving force of next-generation materials technology. Traditionally alloy design has relied on empirical rules and iterative trial-and-error experimentation, with the process of identifying novel compositions being time-consuming, costly, and inefficient. The landscape has recently been revolutionized by advances in machine learning (ML) that enable data-driven methods to improve the efficiency of sophisticated alloy design, selection, and property prediction. ML algorithms can learn effectively the relationships between composition, processing, structure, and properties from existing data, and thus guide the discovery of novel alloys with target properties. In this review, a survey of ML approaches employed in alloy design is provided, including supervised and unsupervised learning, feature engineering, and combination with physical modeling frameworks, such as CALPHAD.