Artificial intelligence for quality control and monitoring in directed energy deposition metal additive manufacturing: applications, challenges, and opportunities

Directed energy deposition (DED) is a versatile metal additive manufacturing (MAM) technique that enables the fabrication and repair of high-value components across diverse industries. However, ensuring consistent quality in DED remains a persistent challenge due to the complex interplay of thermal gradients, material behavior, and process parameters. In recent years, artificial intelligence (AI) methodologies, particularly machine learning (ML) and deep learning (DL), have shown significant promise in enhancing quality control and real-time monitoring in DED processes. This study presents a systematic literature review of AI-enabled quality control and monitoring in DED, with emphasis on defect detection, process-state monitoring, physics-informed modeling, digital twins, industrial readiness, and the comparison between AI-based and conventional inspection approaches. Using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methodology, 145 peer-reviewed articles published between 2015 and 2026 were analyzed. Quality issues in DED are classified into geometrical, morphological, and microstructural categories, and the use of AI is analyzed for each category. The review shows that convolutional neural networks (CNNs), random forests, autoencoders, and related ML/DL models have been widely applied for melt-pool monitoring, thermal and visual signal interpretation, defect classification, geometric prediction, and process-condition assessment. The comparative analysis indicates that conventional methods remain essential for direct measurement, validation, and qualification, whereas AI-based methods are mainly positioned for in-process prediction, early warning, screening, and potential closed-loop control. Emerging physics-informed ML and digital twin frameworks further extend AI capabilities by linking process data with thermal, geometric, and defect-related quality states. Despite these advances, key challenges remain in data acquisition, sensor reliability, the interpretability of AI models, and generalization across geometries and materials. This review identifies future research needs in data-efficient learning, physics-integrated sensor fusion, robust and interpretable modeling, standardized quality metrics, and industrially deployable closed-loop control. By synthesizing current research, the study provides a consolidated foundation for advancing AI-enabled DED quality assurance from research purpose demonstrations toward scalable, traceable, and qualification-ready manufacturing systems.