Enhancing Additive Manufacturing Quality through Vibration Analysis and Machine Learning

Ensuring defect-free additive manufacturing (AM) parts is essential for reliable performance in safety-critical applications. To minimize defects, process-parameter optimization is required. In this study, we first develop a machine-learning model that predicts optimal L-PBF settings—namely laser power, hatch distance, and scan speed—using relative-density results from a structured experimental design. Recognizing that conventional non-destructive evaluation via computed tomography is both time-consuming and expensive, we next propose a rapid, cost-effective inspection method based on vibration and modal analysis. Frequency-response features are extracted from printed specimens and used to train a second machine-learning classifier, enhancing defect-detection accuracy. The combined framework—parameter-optimization modelling followed by vibration-based NDE—demonstrates significant reductions in both build defects and inspection time, offering a scalable pathway toward real-time quality assurance in AM processes.