A statistics of extremes-based method to predict the upper bound of geometrical defects in powder bed fusion

Abstract

Geometrical defects (e.g., pores and lack of fusion) induced during powder bed fusion (PBF) pose a great challenge to fatigue performance for load-bearing structures. The random porosity may further cause uncertainty of fatigue behavior. Fatigue life and scattering bands are largely determined by the maximum size of a geometrical defect. Therefore, the prediction of the upper bound size of geometrical defects is critical. This work focuses on quantifying defect size and distribution to identify the extreme, life-limiting defect size that drives fatigue failure. A comprehensive analysis of the geometrical defects in PBF parts is presented. A statistics of extremes-based method is presented to predict the maximum defect. Defect size is measured using the Feret caliper (FC) diameter and then fitted to a cumulative distribution function (CDF) for linear regression modeling. The maximum defect size is then extrapolated from the regression line for a given probability. It is shown that multiple CDFs can be used for PBF defect size predictions, allowing for flexibility and optimization of prediction models. Furthermore, the FC diameter metric is proposed as the standard defect size metric due to its equivalent representation of 3D and 2D defects.