An in situ crack detection approach in additive manufacturing based on acoustic emission and machine learning

The use of additive manufacturing (AM) and its particular realization - laser powder bed fusion (L-PBF) - is on the rise. However, the method is not free from flaws, mainly represented by structural defects of the printed specimen, such as cracks and pores, requiring processing monitoring. In this work, we propose a concept of the in situ crack detection system for AM fabricated parts based on acoustic emission (AE) signal and machine learning (ML) methods. The detection implies the differentiation of crack AE events from background noise sound. We construct classification ML models and show that they reach the highest classification accuracy, up to 99%, for events represented in the space of spectra principal components. The presented in situ crack detection approach can be easily implemented or used as a basis for a more sophisticated detection procedure.