NiTi wire welding using Fe, Cu and Ni fillers: multi-objective modelling and AHP–entropy approach to sustainability evaluation

Purpose This research introduces a comprehensive framework for optimising fibre laser butt welding of NiTi shape memory alloy (SMA) wires, integrating experimental studies, predictive modelling and sustainability evaluation, with particular focus on the tensile strength–corrosion trade-off. Design/methodology/approach A 43-run central composite design (CCD) experiment was conducted to assess how laser power, pulse duration, frequency, wire diameter and filler powder types (Fe/Cu/Ni) influence tensile strength and corrosion rate. The resulting response patterns were examined using ANOVA and Pareto charts, and three modelling methods – adaptive neuro-fuzzy inference system (ANFIS), particle swarm optimised ANFIS (PSO-ANFIS) and Gaussian process regression (GPR) – were compared. A hybrid AHP–entropy approach was used to benchmark the sustainability of different welding techniques. Findings ANOVA and Pareto analysis confirmed pulse time, frequency and power as dominant factors for both responses. Fe filler produced tensile strengths over 400 MPa, while Ni filler had the lowest corrosion rate (∼0.001298 mm/year) and more uniform weld microstructures, showing a strength–corrosion trade-off. GPR model outperformed others with high accuracy for tensile strength (R2 = 0.8752; RMSE = 23.9706 MPa) and corrosion rate (R2 = 0.9565; RMSE = 0.0001 mm/year). Confirmatory tests supported predictions, and fibre laser welding was ranked most sustainable. Practical implications The integration of predictive modelling with sustainability evaluation offers a strong tool for optimising technical performance and environmental impact in NiTi SMA welding. This study's insights are vital for industrial engineering, where high-performance materials and sustainable manufacturing improve reliability, cut costs and ensure environmental compliance. Originality/value The proposed methodology offers a novel comparative perspective on modelling strategies, supporting informed decision-making in advanced manufacturing using a relatively small dataset.