Linking Geometric Features to Distortion Risk in Laser Powder Bed Fusion via Clustering

Abstract This article investigates the relationship between geometric features of 3D geometries and distortion in laser-based powder bed fusion (LPBF) additive manufacturing processes. We used 97 shape descriptors, including several unique descriptors created specifically for this application, to extract critical geometric attributes from a set of 510 3D models. Simulations of the LPBF process predicted distributions of distortion after printing. Clustering analysis was applied to group diverse geometries based on the distribution of distortion magnitude across the part, identifying three distinct clusters of low, medium, and high distortion risk for our evaluated 3D models. Our findings reveal specific geometric characteristics that are strongly associated with high process-induced distortion from LPBF, including overhanging length, part size, skewness and kurtosis of the material distribution, and the ratio of the width of the parts’ base to its maximum width. Our approach can help identify patterns in simulation data for diverse sets of 3D models and extract meaningful geometric characteristics that lead to differences among groups of models.