Prediction and Control of Product Shape Quality for Wire and Arc Additive Manufacturing

Wire and arc additive manufacturing (WAAM) has become an economically viable option for fast fabrication of large near-net shape parts using high-value materials in the aerospace and petroleum industries. However, wide adoption of WAAM technologies has been limited by low shape accuracy, high surface roughness, and poor reproducibility. Since WAAM part quality is affected by a multitude of factors related to part geometries, materials, and process parameters, experimental characterization or physics-based simulation for WAAM process optimization can be cost prohibitive, particularly for new part designs. As an effective alternative, data-analytical approaches have been developed for prescriptive modeling and compensation of shape deviations in 3D printed parts. However, WAAM faces a unique challenge of large shape deviation and high surface roughness at the same time. Accurate prediction and control of WAAM part quality require process-meaningful error decomposition under geometric measurement uncertainties. We propose a generalized additive modeling approach to separate global geometric shape deformation from surface roughness. Under this statistical framework, tensor product basis expansion is adopted to learn both the low-order shape deformation and high-order roughness patterns. The established predictive model enables optimal geometric compensation for product redesign to reduce shape deformation from the target geometry without altering process parameters. Experimental validation on WAAM manufactured cylindrical walls of various radii shows the effectiveness of the proposed framework.