Digital Twin-Based Process Optimization and Defect Prediction in Metal Additive Manufacturing for Critical Mechanical Components

Abstract

Digital twin technology has emerged as a promising route for improving process stability, traceability, and quality assurance in metal additive manufacturing, particularly for safety-critical and high-value mechanical components. This study presents a digital twin-based framework for process optimization and defect prediction in laser powder bed fusion (LPBF) of critical mechanical parts. The proposed framework integrates a process digital thread, in-situ sensing, a simplified thermal–physics model, and a machine-learning-based defect prediction module within a closed-loop optimization architecture. A representative Ti-6Al-4V load-bearing bracket is considered as a case study to demonstrate how the digital twin can monitor layer-wise thermal behavior, estimate defect probability, and recommend improved combinations of laser power, scan speed, hatch spacing, and layer thickness. The study further examines recent developments in digital twin implementation for metal additive manufacturing and shows that current research has progressed from static simulation toward real-time synchronization, sensor fusion, and quality-oriented architectures. However, important challenges remain in model transferability, standards integration, and component-level qualification. The proposed methodology addresses these issues by explicitly linking process parameters to key defect mechanisms, including lack of fusion, keyhole porosity, thermal distortion, and powder-bed irregularity. Illustrative calculations and representative numerical results indicate that the optimized parameter set can reduce predicted porosity, improve density and geometric stability, and maintain practical productivity. Overall, the study demonstrates the potential of digital twins to provide a more reliable and data-driven pathway for manufacturing critical components in aerospace, biomedical, and energy applications.