Adaptive Neuro Fuzzy Inference System to Predict the Mechanical Properties of Friction Stir Welded AA7075-T651 Joints

The soft computing techniques are nowadays widely used in manufacturing industry for the modeling and optimization of processes parameters. The soft computing techniques give excellent predicted values which agree with the experimental results. In the present study, predictive model for the mechanical properties viz. ultimate tensile strength, micro hardness at weld nugget, and surface roughness in weld bead of friction stir weldedAA7075-T651 are developed. The adaptive fuzzy inference system technique is used for the development of the models. The models are developed using triangular, trapezoidal, Gaussian and generalized bell membership functions, and predicted values are compared. The triangular membership function shows minimum testing error of 19.1091, 12.3152, and 1.0018 for ultimate tensile strength, micro hardness at weld nugget, and surface roughness respectively. The validation experiment is performed at tool rotation speed of 1400 rpm and welding speed of 20 mm/min in order to check the predicted adaptive fuzzy inference system output. The observed values obtained after the validation experiment for ultimate tensile strength, micro hardness at weld nugget, and surface roughness are closer to the predicted adaptive fuzzy inference system output. The scanning electron microscopy images with energy dispersive X-ray spectrometer analysis confirmed the homogeneous mixing of material, laminar material flow with the equiaxed grain (size similar to 260 nm to 3 mu m)distribution at the weld nugget. The scanning electron microscopy images of fractured tensile specimen shows the large dimple with the failure of specimen in heat affected zone. (C) 2022 Jordan Journal of Mechanical and Industrial Engineering. All rights reserved