In-Situ Monitoring Approach for Real-Time Prediction of Humping Defect in Direct Energy Deposition Process

Humping is a common defect in direct energy deposition (DED) additive manufacturing that impacts the geometric accuracy and mechanical properties of printed parts. The various mechanisms contributing to its onset and the intricate interdependence among process parameters constrained its offline prediction. Therefore, this study presents a realtime monitoring approach for early humping detection, laying the groundwork for process control to improve DED productivity, reliability and sustainability. A machine-learning approach is developed to detect humping formation based on realtime processing of high-speed multi-spectral imaging. Experimental tests demonstrated the system sensitivity to key-features associated with humping, illustrating its suitability for in-situ applications.