Optimizing process stability and energy efficiency in wire arc additive manufacturing of nickelbased superalloys using machine learning

Wire arc additive manufacturing (WAAM) is rising as a key technique in the context of sustainability. In WAAM, single weld beads (WB) are the fundamental entity in toolpath planning. Its characteristics (e.g., height -h, width -w, aspect ratio -h/w, and surface roughness -Sa) play a crucial role in the process stability and the final part quality. In this research, the prediction models for WB's characteristics in cold metal transfer (CMT)-WAAM of nickel-based superalloy are developed, using gaussian process regression (GPR) models. The non-sorted grey wolf optimization (NSGWO) algorithm combined with techniques for order preferences by similarity to ideal solution (TOPSIS) is then applied to identify the proper process parameters with the goal of improving the process stability and minimizing energy consumption. To achieve these goals, the experiment is designed with the two key variables -the wire feed speed (wfs) and the travel speed (v). Subsequently, the WB's characteristics are collected, and the energy consumption (E) is recorded during the deposition of each WB. The results demonstrate that the GPR models feature high accuracy with the R 2 values of 0.98, 0.95, 0.90, 0.97, and 0.99 for w, h, h/w, Sa and E, respectively. The optimized process parameters (wfs = 8.20 m/min and v = 60 cm/min) enable a reduction of 66% in Sa and 170% in E compared to the worst case. The WB fabricated with the optimal variables also features regular width with Sa and h/w values very close to the predicted values, confirming the accuracy and efficacy of the proposed approach.