Abstract
Laser Powder Bed Fusion (LPBF) is one of the prominent additive manufacturing methods developed in the last decades to fabricate metal parts with high geometric resolution. To fully leverage the advantages of LPBF, it is generally desired to produce low-roughness surfaces to minimize the efforts in post-process machining. However, understanding the formation of the surface roughness remains obscure due to the complexity of the fabrication process. This paper presents a machine-learning approach to monitor the surface roughness in-situ using a high-speed infrared (IR) camera observing the printed tracks during LPBF process. Thermal signatures such as temperature gradient surrounding the melt pool in different orientations, maximum temperature, and maximum temperature standard deviations were extracted from the IR images and compared with the local surface roughness. Experimental studies have been conducted to train and validate the model. The results have shown good correlations between the temperature gradient and the roughness.
