Prediction of microstructural defects in additive manufacturing from powder bed quality using digital image correlation

Additive manufacturing (AM) of metal components offers tremendous promise to create complex parts with limited supervision. However, process quality and widespread acceptance are currently stymied by the formation of microstructural defects and residual stresses. In particular, lack-of-fusion (LoF) and keyhole defects frequently occur during AM production dependent on process conditions and can severely limit mechanical performance if not corrected. The effects of powder layer quality, where variations from the ideal have the potential to significantly alter defect formation, have not been thoroughly investigated in the literature. In this work, we employ a novel method incorporating in situ three-dimensional digital image correlation (3D-DIC) imaging of the powder bed in-process to identify and quantify the severity of powder bed irregularities during processing. Anomalies in the powder bed were detected, and their geometries quantified, layer by layer using the 3D-DIC analysis for parts produced at multiple energy density levels. Ex situ characterization via scanning electron microscopy identified the locations of physical defects for comparison to DIC data. The quantified powder bed 3D-DIC data, alongside ex situ identification of physical defect locations, was then fed into a Naive-Bayes clas-sification algorithm to predict the likelihood of physical defect formation based on the severity of in-process powder bed errors. This methodology has the potential to be used to predict physical defect formation based on detected powder spreading errors in-process, prior to the formation of many microstructural defects.