Defect Detection from In-situ Monitoring of LPBF Additive Manufacturing

Additive Manufacturing provides many potential advantages, relative to traditional manufacturing methods, for the Navy and other organizations in the aerospace community.  Although flight critical aerospace quality metal alloy components have been produced and flight tested, confidently expanding the use of AM in these applications requiring stringent quality control and repeatability. The vendor and user community has been continually investigating multiple in-situ process sensor technologies to enable advancements in process monitoring and control. Machine Learning (ML) methods and systematic, intelligent fusion of sensor data provides an attractive route to more confidently warn the user of the presence, location, and type of defects in AM parts. In this Phase I effort, CFD Research and our partners from the Advanced Research Laboratory Penn State will implement state-of-the-art ML methods with data fusion strategies. ML training and application will demonstrate the feasibility of advancing defect detection and defect location prediction accuracy, and of predicting defect types, from multiple in-situ sensor modes. The selected ML model structures will enable efficient, intelligent identification of the most important sensor data for future model improvements. In Phase II, extensive testing and training will be used to validate the models and extend the methodology for estimation of critical mechanical properties.