Avoiding surface-related defects is increasingly important as laser-based powder bed fusion of metals (PBF-LB/M) moves toward larger and shaped beam profiles to raise productivity. Although such beams can mitigate certain surface-related instability mechanisms, they frequently introduce characteristic top-surface features such as bulging or wavy topographies. Despite the relevance of these surface features, the top-surface topography is usually classified only qualitatively, and systematic quantitative characterization is largely missing. The aim of the work is to quantitatively characterize and classify the top-surface topographies produced by Gaussian, ring-shaped, and rectangular beams using varying laser power, scan speed and hatch distance. A machine learning based classification algorithm for the top-surface topography identifies four distinct surface classes, Flat, Bulging, Lack of Fusion and Wavy , using a small set of measurable areal surface parameters. The classes can be distinguished by the arithmetic mean height S a , autocorrelation length S al , developed interfacial ratio S dr , and reduced peak height S pk , and the resulting linear rules enable accurate and interpretable classification of the occurring top-surface topographies. These key findings demonstrate that complex surface variations can be reduced to compact, physically meaningful decision criteria, providing a quantitative tool for mapping process windows and identifying parameter combinations that promote Flat or defective surface states.
Prudlik et al. (Fri,) studied this question.