Predictive Modeling of Out-of-Plane Deviation for the Quality Improvement of Additive Manufacturing

Abstract

Additive manufacturing (AM) is a new technology for fabricating products straight from a 3D digital model, which can lower costs, minimize waste, and increase building speed while maintaining acceptable quality. However, it still suffers from low dimensional accuracy and a lack of geometrical quality standards. Moreover, there is a need for a robust AM configuration to perform in-situ inspections during the fabrication. This work established a 3D printing-scanning setup to collect 3D point cloud data of printed parts and then compare them with nominal 3D point cloud data to quantify the deviation in all X, Y, and Z directions. Specifically, this work aims at predicting the anticipated deviation along the Z direction by applying a deep learning-based prediction model. An experiment with regard to a human “Knee” prototype fabricated by Fused Deposition Modeling (FDM) is conducted to show the effectiveness of the proposed methods.