| Abstract Scope |
Despite using constant processing parameters in laser powder bed fusion (PBF-LB) additive manufacturing (AM), properties may deviate from anticipated due to local variations in thermal history due to the local geometry, laser scanning pattern, spatter, and overall layout of samples on the build plate. This work aims to link in process signals, namely those collected by photodiodes, to the mechanical properties of samples made using PBF-LB AM. Tensile Ti-6Al-4V and AlSi10Mg samples were fabricated using a range of processing conditions to determine if in process signals can be linked to the mechanical properties that result from differences in thermal history. Machine learning was used to link the signals to properties. The model is able to predict the ultimate tensile strength and elongation to failure of additively manufactured samples, pointing to a potential for real-time process monitoring and correction. |