Abstract
Wire Arc Additive Manufacturing (WAAM) is a manufacturing process that deposits weld beads layer-by-layer in a planar fashion, leading to a final part. Metallic parts fabricated using WAAM can meet demands from the aerospace and automotive industries, as well as promote automation from design to fabrication. The advantage of WAAMcompared to subtractive manufacturing techniques make it ideal for producing medium to large scale components with low to medium geometrical complexity. Due to the layer-by-layer nature of weld bead deposition in WAAM, the accuracy of the printed geometry is largely dependent on the knowledge of the bead profile employed, which by itself is dependent on a variety of process parameters, such as wire feed rate and torch speed. For multi-bead printing, process parameters extend to stepover ratios as well. Existing models for modelling bead profile are based on its width and height, which do not necessarily capture the geometry of the weld bead accurately. This could affect the stepover increment strategy, which dictates the geometry of the resulting overlapping valley. In this paper, the performance of a variety of machine learning frameworks for predicting the bead cross-sectional profiles is formulated and evaluated. To model the geometry of a bead, different representations describing the single bead geometry are explored, such as direct cartesian representations using polynomials and vertical coordinates, as well as a higher dimensional representation using planar quaternions for supervised learning. Experiments are conducted on single bead SS316L and bronze materials to compare the performance of various frameworks. It was found that among these, the planar quaternions representation with a non-linear neural network framework captures and retains the curvature characteristics of the bead during the learning and prediction process most accurately with a mean Chi-Square goodness of fit of 0.026 mm. A new methodology, the Varying-Ratios Flat-Top Overlapping Model (VFOM), for the derivation of an optimal multi-bead stepover ratio is also introduced. Based on the single bead investigation, the single bead profiles vary across different process parameters. Hence, eliminating the assumption of parabolic single bead profiles, the stepover ratios of each bead are calculated based on their input process parameters while assuming a flat-top surface for an ideal multi-bead. Experiments are conducted with bronze material using both proposed and literature models. Results show that the proposed methodology is better in producing a smooth surface, but does not perform as well in maintaining the multi-bead thickness. Further exploration is conducted on the feasibility of combining both the proposed planar quaternions learning framework for single bead and the VFOM. Results of the experiment show that the combined VFOM does not perform well compared to the other models, and further optimisation is needed to improve the quality of multi-beads printed with this approach.