Data-driven parameter optimization for bead geometry in wire arc additive manufacturing of 17-4 PH stainless steel

Abstract

Due to its high strength, corrosion resistance, and toughness, 17-4 Precipitation Hardening (PH) stainless steel is widely used in aerospace, petrochemical, and marine industries. Additive manufacturing (AM) technologies enable the fabrication of complex and/or customized components while offering superior material efficiency and shorter lead times. Because of its high deposition rate, Wire Arc Additive Manufacturing (WAAM) can produce large metal structures. However, consistent bead profiles remain challenging because the process is highly sensitive to variations in thermal input and deposition conditions. Achieving uniform bead geometry during additive manufacturing is essential to avoid defects such as humming, spattering, and distortion, which can compromise the structural integrity of 3D components.

To achieve a uniform bead profile in WAAM, in this study, a full-factorial design of experiments is implemented to optimize the process parameters such as Wire Feed Rate (WFR), Torch Travel Speed (TTS), and Gas Flow Rate (GFR) for 17-4PH stainless steel. A backpropagation neural network (BPNN) is trained to model a non-linear relationship between these process parameters and bead geometry. Moreover, a genetic algorithm (GA) optimizes for bead uniformity and deposition efficiency. With a Pearson Correlation Coefficient (PCC) of 0.85, the optimized parameters exhibited significantly improved uniformity and higher deposition efficiency.