Abstract

Metal additive manufacturing (AM), particularly laser powder bed fusion (LPBF) and directed energy deposition (DED), has become an important route for producing lightweight, customized, and high-value metallic components. However, industrial adoption is still limited by process instability, layer-to-layer variability, porosity, lack-of-fusion defects, keyhole formation, distortion, and the high cost of post-build inspection. This paper presents a digital twin-enabled framework for real-time quality monitoring in metal AM processes. The proposed framework links the physical AM machine, in-situ sensing, thermal-process modeling, machine-learning-based defect prediction, and a decision layer for quality risk assessment. A hybrid methodology is formulated using volumetric energy density, transient heat-transfer balance, normalized sensor-feature extraction, probabilistic defect classification, and statistical anomaly monitoring. An illustrative LPBF case is used to demonstrate how melt-pool temperature, exposure time above melting threshold, spatter intensity, acoustic-emission energy, and layerwise geometric deviation can be converted into a real-time quality index. The framework is intended to reduce dependence on purely post-process inspection and support earlier detection of defects during fabrication. The results indicate that a digital twin structure can organize multi-sensor data into actionable quality indicators and can provide a practical basis for process correction, traceability, and certification-oriented quality assurance. The study contributes a publication-ready conceptual and mathematical framework that can be expanded with experimental data for journal submission.