Metal additive manufacturing is an emerging industrial process enabling the fabrication of complex part geometries layer-by-layer using highly accurate metal deposition trajectories. In contrast with conventional manufacturing, metal additive manufacturing has potential for sustainable production by minimizing material waste or enabling part repair. In wire arc additive manufacturing, a traveling electric arc melts a metallic wire. The molten metal is deposited droplet-by-droplet and solidifies as a bead. Rapid solidification, inherent to the process, leads to inhomogeneous coarse microstructures and anisotropic properties. Currently, poor understanding of the relationships between process parameters, microstructure, material properties and part performance hinders the industrial adoption of wire arc additive manufacturing. Understanding the effect of the deposition parameters on the solidification microstructure is of prime importance to qualify additively produced parts for industrial applications.

A decimetric hexahedral stainless-steel sample was made using wire arc additive manufacturing. The sample was characterized at multiple scales using light optical microscopy (LOM), scanning electron microscopy (SEM), and electron back-scatter diffraction (EBSD). The optical micrographs featured periodic millimeter-scaled bead geometry inside the block. SEM revealed the microstructure inside the beads, demonstrating elongated grains with various growth directions. The crystal orientation of the grains was observed by EBSD. An unsupervised learning algorithm was used to segment the bead into several microstructural regions and study the grain continuity at the bead interfaces.

A two-dimensional model of droplet solidification predicts the evolution of the liquid-solid interface during cooling and the crystal orientation. A phase-field approach implemented in an open-source finite element program is used. The bead contours were extracted from the LOM observations to implement adequate boundary conditions. The model results are compared to the EBSD observations as well as to the clusters determined using unsupervised learning on SEM images. This work is a step towards the accurate prediction of microstructures resulting from wire arc additive manufacturing.