Curvature-Based Segmentation of Powder Bed Point Clouds for In-Process Monitoring

Abstract

This paper presents a curvature-based analysis of point clouds collected in-process with fringe projection in a polymer powder bed fusion process. The three-dimensional point clouds were obtained from outside of the build chamber with a fringe projection measurement system which was provided with access through an observation window. The curvature-based thresholding of powder bed point clouds demonstrates the ability to separate consolidated areas from the powder bed effectively. This segmentation of the point clouds with masks enables the detection of changes in the outline of consolidated areas between layers, computation of average drop due to the consolidation of the powder bed and separate analysis of both powder bed and consolidated areas. The high-level insights extracted from the analysis of the point clouds could improve process control strategies, such as in-line defect detection during an additive manufacturing build as well as an in-process feedback system for tuning the optimal values of additive process parameters. In summary, we show curvature-based thresholding as an effective segmentation for fringe projection point clouds, which can be further applied to detect defects, such as geometric defects and dimensional inaccuracy.