PBF-LB/M Processing Parameters and Micrographs of Ti-6Al-4V for Machine Learning

Description This dataset was especially designed and created for machine learning approaches focusing on the relationship between process parameters and defect and contains 400 light microscopy images of metallic micrographs obtained from additively manufactured Ti-6Al-4V (Grade 5) alloy. Each image was manufactured using different processing parameter combinations by varying the laser power, the hatching distance, the laser scanning velocity and the powder layer thickness. All processing parameter combinations and corresponding images are given in the attached CSV file. The selection of processing parameter combination was done randomly in-between given borders to obtain an equal distribution for all four processing parameters. The laser power was varied in-between 150 W and 350 W, the laser scanning velocity in-between 800 mm s‑1 and 1600 mm s‑1 and the hatching distance in-between 70 µm and 150 µm. For the powder layer thickness four steps were selected: 25 µm, 50 µm, 75 µm and 100 µm. Sample Manufacturing For each process parameter combination one cuboid was printed. To speed up metallographic preparation and microscopy three cuboids were combined in one specimen geometry. The cuboids are 4 x 4 x 4.5 mm3 (w x d x h) in size and were connected by a 1 x 4 x 1.5 mm3 foot. The small foot and the gap above it separate the volumes manufactured with different processing parameter combinations leading in negligible interaction of the segments. For manufacturing a Nikon SLM 125 HL equipped with a Grade 5 titanium substrate plate preheated to 200 °C was used. Each build job was printed using the skip layer function to combine all four selected powder layer thicknesses in each build job. The used powder was provided by Concept Laser with a particle size distribution of +20 µm to -63 µm. After additive manufacturing the specimens were removed from the substrate plate via electronic discharge machining (EDM). EDM was also used cutting the specimens vertically in half. Subsequently the specimens were embedded in cold-curing resin, grinded and polished. For light microscopy a Leica M205A was used. Keywords Ti-6Al-4V, titanium alloys, Machine Learning, Image Data, Light Microscopy, Laser Additive Manufacturing, PBF-LB/M, Porosity, Process Defects Funding This work was funded by the University of Bremen Research Alliance (UBRA) AI Center for Healthcare within the project PORTAL.