Hierarchical spatial-temporal modeling and monitoring of melt pool evolution in laser-based additive manufacturing

Abstract

Melt pool dynamics reflect the formulation of microstructural defects in parts during Laser-Based Additive Manufacturing (LBAM). The thermal images of the melt pool collected during the LBAM process provide unique opportunities for modeling and monitoring its evolution. The recognized anomalies are evidence of part defects that are to be eliminated for higher product quality. A unique concern in analyzing thermal images is spatial-temporal correlations – the heat transfer within the melt pool causes spatial correlations among pixels in an image, and the evolution of the melt pool causes temporal correlations across images. The objective of this study is to develop a LBAM modeling-monitoring framework that incorporates spatial-temporal effects in characterizing and monitoring melt pool behavior. Spatial-Temporal Conditional Autoregressive (STCAR) models are explored. STCAR-AR is identified as the best candidate among the numerous STCAR variants. A novel two-level control chart is constructed on top of the STCAR-AR model to monitor the melt pool dynamics. A hierarchical structure underlies the two-level control chart in the sense that global anomalies recognized in Level II can be traced in Level I for further inspection. A comparison with other recently developed in-situ monitoring approaches shows that the proposed framework achieves the best detection power and false positive rate.