Classifying the Dimensional Variation in Additive Manufactured Parts from Laser-Scanned 3D Point Cloud Data using Machine Learning Approaches

The objective of this work is to differentiate (classify) additive manufactured (AM) parts from one another with respect to the severity of their dimensional variations based on sparse sampling from a large (~ 2 million data points for each part) laser-scanned dataset of three-dimensional part coordinates (3D point cloud). The outcome is a method to classify the dimensional variation of AM parts with minimal measurements (< 5% of points sampled). This is practically important result, because, it reduces the measurement burden for post-process quality assurance in AM - parts are laser scanned and their dimensional variation is quickly assessed on the shop floor. To realize the objective, test parts are made using the fused filament fabrication (FFF) polymer AM process. The FFF process conditions are varied in a phased designed experiments manner to pro-duce parts with different severities of dimensional variation. Subsequently, each of the test part is laser-scanned and 3D point cloud data acquired. To classify the dimensional variation amongst parts, the authors invoke spectral graph Laplacian eigenvalues (?*) as an extracted feature from the laser scanned 3D point cloud data in conjunction with various machine learning techniques. Six machine learning approaches are juxtaposed: sparse representation, k-nearest neighbors, neu-ral network, naïve Bayes, support vector machine, and decision tree. Of these, the sparse repre-sentation technique provides the highest classification accuracy (F-score > 95%).