Applications and challenges of machine learning in metal additive manufacturing

Metal Additive manufacturing (MAM) enables the fabrication of geometrically complex, high-performance components but suffers from nonlinear process-structure-property relationships, defect formation, and costly optimization. This review systematically evaluates machine learning applications across the MAM lifecycle, including process parameter optimization, defect detection and monitoring, material property prediction, and intelligent design. It further explores frontier advances in physics-informed learning, explainable artificial intelligence, and digital twins for closed-loop control. By identifying critical hurdles such as data scarcity, limited interpretability and real-time implementation, this paper provides a strategic roadmap toward a reliable and intelligent MAM process.