The Use of Machine Learning in Aluminum and Magnesium Matrix Composites: A Review

Abstract Aluminum (Al) and magnesium (Mg) matrix composites are attractive materials for automotive, aerospace, and energy applications due to their low density, high specific strength, and favorable wear performance. However, despite extensive experimental efforts, predicting their mechanical, tribological, and corrosion behaviors remains challenging because of the complex and nonlinear interactions among composition, processing parameters, and reinforcement characteristics. To address this need, this review provides a structured synthesis of recent machine learning (ML) applications to Al and Mg matrix composites, focusing on studies published between 2021 and 2025. A systematic multidatabase search was conducted, and the selected works were examined within a unified ML pipeline encompassing data acquisition, preprocessing practices, feature handling, learning strategies, training protocols, and performance evaluation. The analysis reveals trends in dataset size, target properties, descriptor types, and modeling choices. It also highlights recurring challenges related to data scarcity, heterogeneous reporting of preprocessing and evaluation practices, and uneven coverage of different property classes. Informed by these findings, the review provides practical recommendations for data curation, preprocessing, model selection, and training to facilitate more reliable ML-based prediction and data driven guidance for improving the properties of Al and Mg matrix composites.