Toward closed-loop quality assurance in powder bed fusion additive manufacturing: Defect detection, machine learning, and computational modeling

Powder Bed Fusion (PBF) is a critical enabling technology in metal additive manufacturing (AM) to produce high-performance components deployed in extreme environments, including aerospace, energy, and biomedical applications. However, internal defects, such as lack of fusion porosity, gas-entrapped pores, and keyhole-induced voids continue to impose limitations on structural integrity, fatigue life, and process reliability. Achieving defect-free manufacturing under demanding performance requirements necessitates advanced detection and control strategies. Several recent reviews have addressed either in-situ monitoring techniques or machine-learning-based quality analytics in PBF, but typically in isolation. By contrast, this work provides an integrated review of defect detection across the PBF process chain, with emphasis on in-situ sensing, ex-situ characterization, machine learning-based classification, and mechanistic, numerical, and simulation approaches across micro-, meso-, and macro-scales. Particular focus is placed on computational models that capture critical physical phenomena at different scales, providing insight into defect formation and mitigation in PBF processes. The article also identifies current research gaps and outlines future directions for developing robust defect detection frameworks that support the qualification of AM components for mission-critical and extreme applications. These insights contribute to advancing the state-of-the-art in high-reliability additive manufacturing and accelerating its industrial adoption.