General Regression Neural Network-Based Frame Work for the Evaluation of Ultimate Tensile Strength of Vibratory-Assisted Welded Joints

Abstract

AbstractWelding is a widely used process to join the components. In recent years, several analytical methods are used to build the relationship between parameters of welding process and the quality of weldments. Soft computing tools are the one which can model the relationship between parameters of welding process and the quality of weldments at a shorter interval of time. In the present work, assessment of weld quality has been carried using general regression neural network (GRNN). A model is developed between vibratory tungsten inert gas welding process parameters and ultimate tensile strength of aluminum 5052 alloy weldments using experimental data. The developed GRNN model is validated with the experimental data. The predicted GRNN values closely matched with experimental values. This trained GRNN model can also predict the ultimate tensile strength of welded joints with an accuracy of 98.94%.KeywordsGeneral regression neural networksOptimizationVibratory-assisted weldingTIG welding