High-speed imaging investigation and process mapping of the plume behavior in laser powder bed fusion

In this work, high-speed imaging of experiments was conducted with varied processing conditions and materials to parametrically understand the plume’s severity, size, and trajectory on a commercial laser powder bed fusion (LPBF) machine. In this context, the plume refers to the bright melt-pool-scale metal vapor and condensate. A U-Net convolutional neural network (CNN) was trained to segment the plume from experimental images. A comparison of the plume generated from different metals shows that Ti-6Al-4V has a significantly brighter and larger plume than Inconel 718 and 316L SS, which may be attributed to increased emission from the melt pool, increased scattering of the emission from the melt pool, or elemental emission lines. The effect of powder was studied for both Ti-6Al-4V and 316L SS and shows a reduction in the size and brightness of the plume in most overhead images, suggesting that the plume visibility and/or the plume itself is suppressed by the powder and spatter. Process mapping the plume in power and scanning velocity space shows that the plume transitions from ejecting toward the rear of the melt pool in the transitional regime to ejecting directly above the melt pool in both the severe keyholing and conduction-dominated regimes, consistent with the vapor depression geometry under the laser. The temporal variability of the plume increases with increasing power-to-velocity ratio, which is attributed to melt pool and vapor depression instability, but can also be large at very low power-to-velocity ratios. This experimental study aims to increase our understanding of the plume’s behavior at the melt pool scale in LPBF and can be used to validate multi-physics models of the plume and inform parameter selection for both minimal melt-pool-scale laser-plume interaction and avoiding plume interference in melt pool imaging.