Investigation of Vibratory-Assisted TIG Welding on Al6063 Alloy: Microstructural Behavior, Mechanical Properties, and Machine Learning-Based Hardness Prediction

This study investigates the microstructural behavior and mechanical properties of Al 6063 alloy under vibratory-assisted TIG welding, aiming to develop a machine learning-enabled property assessment for the welded alloy. The research establishes correlations between vibratory welding parameters and weld joint hardness by employing machine learning techniques, specifically general regression neural network (GRNN) and artificial neural network (ANN) models. The methodology involved varying parameters such as the input voltage to the vibromotor and the vibration duration during vibratory-assisted TIG welding of Al 6063 alloy. The subsequent microstructural and hardness changes were analyzed. Experimental data were used to train and optimize GRNN and ANN models, which were validated using a separate dataset to ensure reliability and accuracy. A predictive tool was developed to forecast weld joint hardness based on specific vibratory welding parameters. Findings revealed that microstructural analysis showed the formation of long dendrites at 0 V and 230 V. In comparison, smaller dendrites were observed at 140 V. Hardness of the Al 6063 alloy increased up to 140 V but decreased with further voltage increase to 230 V. Vibration time had a less significant impact on hardness than the voltage input. Both ANN and GRNN models accurately predicted weld joint hardness, with an overall regression coefficient of 0.99.