Aeroacoustic Process Monitoring and Anomaly Detection in Cold Spray Additive Manufacturing

Abstract Cold spray (CS) is an emerging additive manufacturing method used to deposit a wide range of materials by spraying solid particles at supersonic velocities using high-pressure millimeter scale de Laval nozzles. As CS technology finds applications in diverse areas, including 3D printing, the need for in situ process monitoring becomes increasingly apparent. The CS process is influenced by various process parameters, including nozzle gas inlet pressure, temperature, and powder feed rate. Accurately detecting variations in these parameters, as well as identifying process anomalies (e.g., nozzle wear, clogging), is crucial for the broader implementation of the technology. In situ detection of anomalous events and process health monitoring is paramount for identification of inconsistencies, ensuring product quality, enhancing cost efficiency, and reducing waste by early detection of faults. To this end, in this study, airborne acoustic emission was monitored during CS processes to discern acoustically detectable process parameters. Characteristics of aeroacoustic waves emitted under both free jet and deposition conditions were analyzed. Results indicate that changes in nozzle gas inlet pressure and temperature, powder feed rate, and nozzle wear status are discernible through acoustic power spectrum analysis. Time-domain analysis further facilitated the identification of anomalies associated with powder injection termination, deposit/substrate delamination, and nozzle geometry changes. Notably, the sliding window first order backward differentiation of total power and the power band in the time domain proved effective in detecting gradual anomalies, such as nozzle throat wear, whereas the second-order differentiation highlighted abrupt process changes, like delamination. This study demonstrates that airborne acoustic signals offer valuable insights pertaining to process faults in CS, establishing aeroacoustic signal monitoring as a promising component of stand-alone or multi-modal process monitoring for CS operations. Furthermore, the study offers invaluable insights for aeroacoustic signal feature engineering for the development of machine learning models for process monitoring in CS.