Multi-response optimization of FSW parameters for Al–Mg–Zn alloys using Box–Behnken design and gray relational analysis and comparative study with ANFIS technique

In this study, cast Al–Mg–Zn and aging Al–Mg–Zn alloys were joined using the friction stir welding technique, with the welding parameters being the rotational speed of the tool, welding speed, tilt angle, and D/d ratio. The effects of these input parameters on output responses such as ultimate tensile strength, elongation percentage, hardness, and wear rate were investigated using Box–Behnken experimental design. Multi-response optimization was performed using the Box–Behnken Design (BBD) combined with Gray Relational Analysis (GRA). A rotational speed of 1200 rpm, welding speed of 100 mm/min, tilt angle of 1.5°, and D/d ratio of 2.25 were found to significantly enhance the mechanical properties and wear resistance. Furthermore, the Adaptive Neuro-Fuzzy Inference System (ANFIS) model was developed to predict outcomes based on the gray relational coefficient grades, leveraging its architecture to improve prediction accuracy. Finally, a comparative analysis was conducted between the BBD-GRA methods and the ANFIS technique. Results indicated that the ANFIS model was highly effective for predicting output parameters, achieving superior precision and accuracy with fewer iterative calculations. By selecting an appropriate ANFIS structure, the model’s performance is better than that of the BBD-GRA method.