[Data] Monitoring Of Laser Powder Bed Fusion Process By Bridging Dissimilar Process Maps Using Deep Learning-based Domain Adaptation on Acoustic Emissions

The primary objective in creating this dataset is to map dissimilar process parameter spaces that yield the same Laser Powder Bed Fusion (LPBF) process regimes – Lack of Fusion (LoF), conduction mode, and keyhole pores – using the acoustic signatures derived from LPBF, all without supervision. To achieve this goal, two cuboid samples (width: 20 mm, length: 20 mm) were manufactured from two powder batches, namely D1 and D2. The powder underwent sieving using a mesh 325 vibrating siever (with a 45 μm gap between wires) to obtain two distributions, D1 (upper range of the sieved distribution) and D2 (powder sieved with a diameter < 45 μm). The creation of two powder size distributions aimed to investigate how changes in particle size distribution influence the characteristics of acoustic spectra when interacting with a laser source. The intentional induction of two process spaces involved using two 316L stainless steel powder distributions (> 45 µm and < 45 µm) and processing them with two sets of laser parameters. In addition to altering particle size distribution, laser power parameters were slightly offset across each regime between the two cubes to introduce a shift in the distribution of acoustic signature datasets. The three distinct processing regimes were achieved in each cuboid (D1 and D2) using different combinations of laser power and scanning speed. Verification of the occurrence of these three regimes was conducted by examining workpiece cross-sections through optical microscopy. These cross-sections were cut perpendicular to the scan tracks and further ground and polished in accordance with metallographic preparation standards. The datasets within the .zip file are labeled as: [a]_classspace_5000, [a]_rawspace_5000.npy, were 'a' can be either D1 or D2, representing two datasets with domain shifts. Each row in rawspace.npy corresponds to a ground truth in classpace.npy.