Deep learning-based monitoring of laser beam direct energy deposition to detect deviations in process conditions

Laser Beam Direct Energy Deposition (DED-LB) is an additive manufacturing technique that requires precise process control to ensure defect-free parts. This study proposes an AI-driven monitoring framework utilizing deep learning techniques to evaluate process conditions and identify anomalies that may compromise the integrity of the manufactured part. During deposition, an off-axis high-resolution camera properly configured was used to acquire frames that were subsequently processed to extract brightness values and deposition head motion. To test the proposed methodology, controlled variations in key process parameters, including laser power, powder flow rate, and deposition speed, were introduced to induce anomalies. Laser shutdown and powder flow interruption produced appreciable reductions in brightness while the variation in speed had no detectable effect on laser-particle interaction. The results were statistically evaluated and compared with metallographic analysis.