Current application status and prospects of machine learning in friction stir welding and additive manufacturing

Friction stir welding and its derived solid-state additive manufacturing technologies stand as one of the effective approaches to avoid melting defects and achieve the fabrication of high-performance lightweight alloy components. However，the friction stir welding and solid-state additive manufacturing processes involve complex thermo-mechanical-fluid-microstructure couplings，posing significant challenges to traditional trial-and-error methods for process optimization. The emergence of machine learning provides a transformative solution for process understanding and intelligent control in this field. This paper presents a systematic review of machine learning applications in friction stir welding and additive friction stir deposition. It categorizes and elaborates on the current research status and data processing strategies in aspects such as performance prediction，defect detection，and in-situ control. Addressing the limitations of purely data-driven models，it focuses on investigating different fusion paradigms of physics-informed machine learning and their cutting-edge applications. Finally， it points out that future research should concentrate on further developing generalizable prediction models，achieving real-time closed-loop intelligent control，and integrating active learning for autonomous process exploration，aiming to provide references for advancing these technologies toward intelligent and high-performance development.