A melt pool geometry-based approach is developed to predict surface roughness in metal additively manufactured parts for a range of processing parameters. It is shown that surface roughness on a particular facet can be estimated by stacking melt pools along the facet, extracting their outer contour and applying the necessary transformations. To be able to predict surface quality of various processing parameters in a reasonable time frame, a machine learning framework is developed. This framework is trained over melt pool data generated by high-fidelity FE simulations.
EXTENDED ABSTRACT