The development of an augmented machine learning approach for the additive manufacturing of thermoelectric materials

Through the integration of machine learning (ML) techniques alongside additive manufacturing (AM) experimentation, we demonstrate an iterative process to rapidly predict laser-material interactions and melt pool geometries throughout the build parameter space for a bismuth telluride thermoelectric (TE) material. In doing so, we determined process parameters that created crack-free, highly dense (>99 %) n-type bismuth telluride (Bi2Te2.7Se0.3) parts through laser powder bed fusion (LPBF). Further, the ML-assisted understanding of the processing space allowed for the identification of build parameters that successfully yielded geometrically enhanced Bi2Te2.7Se0.3 parts with reduced build times and no increase in experimental effort.