Defect detection in selective laser melting technology by acoustic signals with deep belief networks

Selective laser melting (SLM) is one of the most important and successfully additive manufacturing processes in 3D metal printing technologies. Critical quality issues such as porosity, surface roughness, crack, and delamination continue to present challenges within SLM-manufactured parts. Monitoring and in-process defect diagnosis are the key to improving the final part quality. Currently, it greatly hinders the adaptability and the development within the defect detection system since the setup restricts the vision and photo diode applications in the SLM process monitoring. Additionally, defect detection with traditional classification approaches makes the system rather complex due to introducing a series of steps. To meet these needs, this study proposes a novel method for the defect detection within the SLM parts. The setup was flexibly conducted using a microphone, and the defect detection was obtained by the framework of deep belief network (DBN). It is implemented by a simplified classification structure without signal preprocessing and feature extraction. The experimental results showed that the utilization of acoustic signals was workable for quality monitoring, and the DBN approach could reach high defect detection rate among five melted states without signal preprocessing.