Infrared-based Detection of Porosity Formation in Laser-based Powder Bed Fusion of Ti6Al4V

The prospects of a broad industrial adoption of additive manufacturing technologies are strongly dependent on the capability of manufacturing systems to consistently deliver high quality parts. While closed loop control is still in its infancy, the in-situ identification of defects is currently of scientific and industrial interest. Radiative measurements have been gaining ground in the monitoring of laser-based powder bed fusion of Metals due to key features that can be extracted from the melt pool and surrounding regions. This paper presents an investigation of the use of high-speed short-wave infrared imaging and photodiodes for the detection of porosity formation in Ti6Al4V parts during processing. The formation of sparks and the material cooldown information are geometrically registered to micro computed tomography data of each sample, building a labelled database for the development of a logistic regression-based porosity prediction model. The model performance is assessed for low, mid, and high energy processing conditions, resulting in F1-scores between 0.51 and 0.81 for the classification of layers, depending on the dataset used for training and testing as well as on the defect threshold selected for binary labelling.