Beyond parabolic weld bead models: AI-based 3D reconstruction of weld beads under transient conditions in wire-arc additive manufacturing

The ability to predict the geometry of the weld bead plays a key role in accurate path planning and determination of welding parameters in wire arc additive manufacturing. However, little attention has been paid to the weld bead geometry and its prediction when the deposition path is not straight. Thus, this work focuses on the 3D reconstruction of the weld bead based on the deposition path. One of the main findings of this paper is that the weld bead shape changes from a symmetrical cross-section in straight portions of the path to an asymmetrical shape in non-straight regions. To predict the 3D geometry of the weld bead, an AI-based architecture called AIBead was developed. A suitable parametrization of the deposition path is proposed that is a key to train the AIBead properly and to outperform currently used parabolic models.