摘要
This study presents an automated thresholding method for analyzing and quantifying the internal composition of additive manufacturing (AM) parts using computed tomography (CT) data. A mixed skewed-Gaussian distribution (MSGD) algorithm, derived from a statistical image analysis technique called Mixed Gaussian Distribution (MGD) clustering, integrates a mixture of skewed-Gaussian distributions to model the internal phases from CT data. The parameters of the MSGD algorithm (i.e. probability, mean, standard deviation, and skew) are inferred from the measured grayscale histogram using least-squares fitting and are assigned to phases present in the CT data. Upon fitting, the MSGD technique guides the thresholding of phases in CT data. From the MSGD fitted and thresholded CT data, phase volume percentages and spatial variations of density of the phases are quantified. The MSGD algorithm was validated using previously reported CT analysis and experimental porosity measurements of two Cobalt Chrome (CoCr) specimens (∼1 % and ∼13 % porosity) fabricated by powder bed fusion (PBF). Compared with the 1.1 % and 13.7 % porosity of the specimens measured by the Archimedes method, the MSGD method predicted a porosity of 1.6 % +/− 0.7 % and 14.5 % +/− 1.9 %, a measured increase of 0.5 % and 0.8 %, respectively. These results show a similarity in predicted porosity between Archimedes and MSGD method indicating that CT and the MSGD method may provide a reasonable estimate for part porosity.