Process quality assessment with imaging and acoustic monitoring during Laser Powder Bed Fusion

Abstract

Acoustic monitoring of laser powder bed fusion (LPBF) has shown a high sensitivity to stochastic defects, e.g., cracks, pores and lack of fusion (LOF), and melting instability. The advantage of this method is the possibility to filter raw data and extract acoustic signal features for the data analysis, thus minimizing data and computing time. In this research during the build of components from hot work tool steel powder, acoustic signals and powder bed images were acquired for post-process data analysis and search for correlations with LOF. Different densities caused by LOF were obtained by changing the shielding gas velocity. In the analysis, selected combinations of features with the relationship between the build phases and the final properties such as density and surface roughness, were investigated. For the current dataset prediction of the optimal state showed an accuracy of 98%. This investigation suggests the applicability of the smart data-centric machine learning method to predict the relationship of process parameters, monitoring signals, and material properties.