Machine Learning in DED-Arc: A Critical Review

Directed Energy Deposition (DED) utilizing an electric arc (DED-Arc) for additive manufacturing of metallic components has emerged as a promising manufacturing technology across aerospace, automotive and construction industries, due to its inherently high deposition rates, large build volume capabilities and low equipment acquisition and material costs. However, its high process complexity and susceptibility to geometric and metallurgical defects have motivated AI and Machine Learning technologies as viable alternatives for process modelling, complementing analytical and numerical approaches whose use in process control is limited by computational cost. This study presents a critical review of prominent publications applying ML to DED-Arc processes, organized in five distinct application domains: offline process modelling and parameter optimization, thermal prediction and surrogate modelling, in situ monitoring and defect detection, post-process quality assessment and pre-process design and path planning. Through explicitly highlighting key limitations of the identified literature clusters, a strong basis for future research directions is established, setting the foundation for mainstream industrial adoption.