An artificial intelligence classifier for electron beam powder bed fusion as-built surface topographies

Additive Manufacturing, a pillar of Industry 4.0, to enable automatic real-time process control, relies on in -situ measurements, some of which -currently under development -exploit surface topography. Topographic characterisation requires a large set of parameters, loosely linked to visual appearance upon which related in-situ measurements are mostly based. A supervised machine learning classifier of as-built surfaces based on topographical characterisation is proposed and applied to tool steel test pieces fabricated by electron beam powder bed fusion. The methodology is developed to provide process engineers with the visual appearance of the topographical parameters set, and enable multi-scale, information-rich quality control.(c) 2023 CIRP.