Prediction and multi-response optimization in wire-arc additive manufacturing of high-strength-alloyed steel using ANFIS models and NSGA-II algorithm

Wire-arc additive manufacturing is widely used for producing large-scale metal components. In wire-arc additive manufacturing, the prediction and optimization of process parameters to control the weld seam geometry is crucial for the deposition path planning and ensuring the part quality. This study develops predictive models for the weld seam width ( w ), height ( h ), and contact angle ( θ ) using an Adaptive Neuro-Fuzzy Inference System, followed by multi-objective optimization through NSGA-II and TOPSIS. The Adaptive Neuro-Fuzzy Inference System models achieved high prediction accuracy with R ² values above 0.99 for training and above 0.90 for testing. The ANOVA results revealed that voltage ( U _a ) had the strongest influence on the weld seam width and contact angle, contributing 62.92% and 74.31%, respectively, while travel speed ( S _w ) most strongly affected the weld seam height (42.91%). The optimal parameters ( I _a  = 150.6 A, U _a  = 19.9 V, and S _w  = 32 cm/minute) produced smooth and uniform weld seams and a regular thin wall without major defects. The proposed modeling-optimization framework provides a practical tool for engineers to accurately predict the weld seam geometry and select appropriate wire-arc additive manufacturing processing parameters for improved manufacturing quality.