State-of-the-Art Review of Machine Learning Applications in Additive Manufacturing; from Design to Manufacturing and Property Control

In this review, some of the latest applicable methods of machine learning (ML) in additive manufacturing (AM) have been presented and the classification of the most common ML techniques and designs for AM have been evaluated. Generally, AM methods are capable of creating complex designs and have shown great efficiency in the customization of intricate products. AM is also a multi-physical process and many parameters affect the quality in the development. As a result, ML has been considered as a competent modeling tool for further understanding and predicting the process of AM. In this work, most commonly implemented AM methods and practices that have been paired with ML methods along with their specific algorithms for optimization are considered. First, an overview of AM and ML techniques is provided. Then, the main steps in AM processes and commonly applied ML methods, as well as their applications, are discussed in further detail, and an outlook of the future of AM in the fourth industrial revolution is given. Ultimately, it was inferred from the previous papers that the most widely applied AM techniques are powder bed fusion, direct energy deposition, and fused deposition modeling. Also, there are other AM methods which are mentioned. The application of ML in each of the renowned techniques are reviewed more explicitly. It was found that, the lack of training data due to the novelty of AM, limitations of available materials to be applied in AM methods, non-standardization in AM data and process, and computational capability were some of the constraints of the application of ML in AM methods.