Prediction and optimization of WEDM parameters for machining of NiTi-shape memory alloy using ANFIS-PSO approach

Abstract This present work deals with improving the WEDM machining performance through soft computing techniques. In this study, a new strategy was implemented on NiTi-SMA (Nitinol-SMA) using ANFIS (Adaptive Neuro-Fuzzy Inference System) and PSO (Particle Swarm Optimization) in wire-EDM (electrical discharge machining). The findings recommended optimal operating settings to maximize dimensional accuracy and minimize both processing time and cost. Real-time monitoring was performed using a vibration device to assess the frequency of motion of the wire electrode during the cutting interaction with the workpiece. The main measurable aspects included vibration, surface finish, and overcut. A Taguchi L18 mixed-level design of experiments (DOE) has been used to conduct these tests. After testing, ANFIS suggested a common optimal setting for each feature: peak current (11.5 A), pulse-on-time (125 µs), pulse-off-time (58 µs), servo voltage (55 V), and wire feed (2 mm/min), resulting in feature values ​​of 0.113 K-Hz, 0.113 µm, and 0.0526 µm. In contrast, alternative optimal settings of PSO yielded feature values ​​of 0.68 K-Hz, 1.87 µm, and 0.648 µm. Comparative analysis demonstrated that ANFIS modeling provided better results, with excellent significant improvements in surface morphology and chemical composition, as investigated via FESEM and EDS testing at the optimal settings of machined and unmachined Zn coated brass wire. In addition, ANFIS effectively reduced the micro cracks and potholes compared to PSO.