[Data] Self-Supervised Bayesian Representation Learning of Acoustic Emissions from Laser Powder Bed Fusion Process for In-situ Monitoring

Different Laser Powder Bed Fusion (LPBF) process spaces were deliberately introduced by employing two distinct 316L stainless steel powder distributions (with particle sizes >45 μm and < 45 μm) and processing them with two sets of laser parameters, resulting in the creation of four datasets [D1, D2, D3, and D4]. These datasets encompass LoF pores, conduction mode, and keyhole formations, each associated with three LPBF regimes denoted as D1, D2, D3, and D4. The experiments utilized a Sisma MYSINT 100 commercial LPBF printer and an airborne AE sensor system with a flat frequency response ranging from 0 to 150 kHz. Validation of the ground truths for the three laser regimes across the four datasets, representing distinct process spaces, was accomplished through the confirmation of cross-sectional images. In the course of fabricating a cube using a powder bed and laser, data acquisition from an AE sensor was triggered when the optical intensity reached a threshold of 0.5 V for each scan length. The photodiode trigger gain was adjusted to saturate at 5 V, and the ensuing continuous-time window, where the optical signal remained at 5 V for 12.5 ms, was calculated and segmented to generate the dataset. Irrespective of the specific regime (Lack of Fusion, Conduction, and Keyhole) or the cube being fabricated (with two powder distributions), the signals obtained during this process were then segmented into a 12.5 ms window comprising 5000 data points. To eliminate any noise, an offline application of a low-pass Butterworth filter with a 150 kHz cut-off frequency was employed, aligned with the frequency response specification of the AE sensor. Each dataset has two files against it [raw/groundtruth label].