Abstract
Additive manufacturing has found its niche in critical component applications in the aerospace and nuclear industries. For these industries, a need for a cost-effective quality assurance method has also increased to incorporate these components. Within laser powder bed fusion (LPBF), in-situ sensing has shown promise with various forms of defect detection, but has only shown limited success in microstructural characterization. Utilizing concurrent in-situ data collection from a complementary metal oxide semiconductor (CMOS) and photodiode sensor, this work establishes a relationship between in-situ sensor monitoring, crystallographic texture and mechanical properties through machine learning (ML). By combining the in-situ data, ML and a dataset of over 100 samples, including X-ray diffraction and tensile testing results, the model successfully predicts textures of 718 Ni alloy up to 90% accuracy. The analysis demonstrates an important progress in predicting whether sample textures are dominated by (111) or (200) crystal orientations, or a mixture of both and how this affects the mechanical properties. Furthermore, three key characteristic samples were investigated via electron backscatter diffraction to delve deeper into mechanical property differences brought by microstructural features. While the model requires future datasets to improve reliability, it opens a pathway to use in-situ processing data to predict the microstructure and mechanical properties of LPBF materials.
