Metal additive manufacturing (AM), particularly, laser powder-bed fusion (LPBF) has transformative potential to achieve geometric design freedom at low production volumes. However, porosity and localized defects remain a significant challenge to implementation in mission critical aerospace applications. While the quality of LPBF is now competitive with or surpasses castings for established materials such as Stainless Steel 304 or 316, increased demands based on enhanced geometry require stringent certification before the printed components can be used in applications experiencing cyclic loading. Researchers at Notre Dame have demonstrated machine learning techniques for analyzing the integrity of printed parts from data stored during the manufacturing process. The main objective of this proposal is to complete the initial design and testing of an Automated Manufacturing Process Analysis System. This device will 1) to eliminate the need to store large amounts of data for post-processing, 2) aggregate and synchronize data from multiple sensors, 3) synthesize data into features and interpolate onto a voxelized space, 4) allow the results of off-line trained machine learning algorithms to be applied forward and 5) enable future layer-to-layer process control.
Benefit: An effective AMPAS could prove to be a key enabling technology in additive manufacturing by ensuring part integrity for mission critical applications. This could take the form of making process adjustments based on measurements or rejecting parts likely to fail (and re-fabricating successful parts). Likely initial users of the AMPAS include manufacturers using LPBF for mission critical parts. This includes installations supporting US Navy operations. We anticipate that a successful AMPAS system will penetrate further into the commercial/educational LPBF space and beyond LPBF to DED and other AM technologies. Finally, the same general architecture of the AMPAS can be used for other Industry 4.0 digital twin applications.
Keywords: additive manufacturing, additive manufacturing, defects, Machine Learning, Nondestructive Evaluation, discontinuities, laser powder bed fusion, Artificial Intelligence
