Quantitative metallographic characterization of process-dependent pores and cracks in selective laser-melted GH3230 alloy using a TWS-assisted workflow

Abstract

An independently validated Trainable Weka Segmentation (TWS)-assisted workflow was developed for quantitative metallographic characterization of pores and cracks in selective laser-melted GH3230. Fifty manually annotated micrographs, excluded from classifier development, were used for validation. TWS achieved the highest recall (0.940 ± 0.044) and the lowest relative error in the equivalent crack-width descriptor (6.2 ± 3.9 %), while Otsu thresholding gave a lower area-fraction error. Applied to process and post-processing conditions, the workflow resolved distinct pore/ crack responses to laser power, scanning speed, hatch spacing, scanning strategy, heat treatment, and hot isostatic pressing. Hot isostatic pressing reduced the crack area fraction from 2.328 % to 0.532 %. The workflow provides a reproducible two-dimensional basis for metallographic process screening of GH3230.