Prediction of mechanical properties of magnesium alloys through machine learning: a review

The present work reviews recent research on the mechanical properties of magnesium (Mg)-based materials using machine learning (ML) and deep learning methods. Magnesium-based materials are one of the important lightweight and biodegradable materials due to their favourable characteristics. However, preparation of magnesium-based materials is a challenging task owing to their inflammable nature and consumes time and energy as well. The mechanical and chemical properties are highly dependent upon material composition, purity, and process parameters. The purpose of this review is to enhance the process of understanding the material properties from traditional techniques to artificial intelligence with higher accuracy. Both supervised and unsupervised ML approaches are used to predict properties of the model. Support vector machine, regression, and decision tree are commonly used supervised learning algorithms with higher accuracy and minimum error (MAE, RMSE). Similarly, principal component analysis, and clustering are commonly used unsupervised learning algorithms. In addition, deep learning methods such as artificial neural network, convolutional neural network are paid significant contributions to material characterisation. Hence, this paper aims to review magnesium alloy-based works with respect to four categories, namely microstructure, mechanical properties, corrosion and wear characterisation.