Towards data driven quality monitoring: alignment and correlation of photodiode-based co-axial melt pool monitoring signals to part quality in laser powder bed fusion

Abstract

Online quality control of advanced manufacturing processes often utilize process monitoring as a qualification method to reduce or eliminate time consuming and costly destructive and non-destructive tests. The need for process monitoring based qualification becomes more apparent with the advent of laser based additive manufacturing for complex geometries and lot-size-one production. However, the usability of the process monitoring solution heavily relies on the accuracy of the established alignment and assignment of process signals to specific part and defects herein. This research proposes a data fusion strategy for data pre-processing to structure and align multi-source sensor data for improving correlation accuracy in process monitoring. A laser powder bed fusion system equipped with a photodiode-based co-axial melt pool monitoring is used for the study. First, the monitoring system collects photon emission data from the melt pool using photodiodes with high frequency (>100 kHz) which are aligned spatially and temporally using position data (xy-coordinates) of the galvanometer scanner and the laser ON/OFF signals. Next, a clustering approach is used to assign each photodiode signal with the individual parts on the build platform. Finally, correlation between melt pool monitoring signals and part level porosity is studied using micro-computed X-ray tomography (μCT) from a build job containing parts produced under varying process conditions.