Process-Informed Segmentation of Dense Point Clouds for Layer Quality Assessment in Large-Scale Metal Additive Manufacturing

Abstract

Large-scale metal additive manufacturing has become increasingly popular in key industrial sectors such as aerospace and petroleum for sustainable manufacturing at the point-of-need. For example, wire-arc additive manufacturing (WAAM) offers high-deposition rates on large printing areas by robotic-assisted welding of thick-layers of material. However, WAAM technologies suffer from low geometric accuracy due to unstable high temperature deposition processes. Physics-based and machine learning (ML) methods have been developed to predict and control WAAM geometric accuracy. These methods rely on accurate layer identification or segmentation from dense point clouds. However, layer segmentation can be challenging and time-consuming due to high surface roughness, excessive layer remelting, and severe out-of-plane layer displacement. To enable fast and accurate layer segmentation for geometric qualification of fabricated parts, we propose an efficient ML algorithm that exploits knowledge of the idealize layer geometry to identify the boundaries between deposited layers. Simulation studies demonstrate the accuracy of the procedure under limited surface roughness conditions. Experimental studies illustrate the applicability of the methodology in practice.