Porosity Detection of Laser Based Additive Manufacturing Using Melt Pool Morphology Clustering

The microstructural and mechanical properties of Laser Based Additive Manufacturing (LBAM) are still inconsistent and unreliable, which is a major barrier that prevents Additive Manufacturing (AM) from entering main stream production. The key challenge is the lack of understanding for the underlying process-properties relationship. We monitor Laser Engineered Net Shaped (LENS) process using a state-of-art thermal image system, and the resulting highspeed Melt Pool (MP) data stream is used to characterize the complex thermo-physical process. We propose a novel method based on Self-Organizing Map to cluster the MPs based on their morphology and link MPs clusters’ characteristics to the porosity of fabricated parts, which is crucial to mechanical properties of parts. The results are validated using X-Ray tomography of Ti64 thin-wall. Our approach identifies various patterns of MP morphology, which corresponds to different types of porosities. The proposed method can potentially be used to certify the part quality in a real-time and non-destructive manner.