Predictive analysis on mechanical properties, material flow and thermal analysis of friction stir welded dissimilar AA2014/AA7075 Al-alloy joints using machine learning.

Aluminum alloys are the most ideal materials for aerospace applications, and joining of a light weighted materials with an eco-friendly method is encouraging ongoing research into the green welding technique. In the current work solid-state joints of 10 mm thick AA2014/AA7075 Al-alloys dissimilar material joints with hybrid pin profiles were fabricated by using Friction Stir Welding (FSW) technique. The microstructure and mechanical properties of weld joints, temperature analysis of the welds processed under various conditions is investigated. An Artificial Neural Network (ANN) model is used to predict the mechanical properties of the welded joints fabricated at different conditions. The weld joints processed with hybrid pentagonal pin profile exhibits the highest UTS and YS measuring as 294 and 267 MPa respectively. The ANN showed strong prediction using a dataset of 27 experimentally obtained results providing a reliable data-driven method for forecasting UTS and optimizing FSW parameters. The weld joints processed at 900 rpm 31. 5 mm/min 2° tilt angle 0 mm tool offset and pentagonal pin profile revealed 62% of joint efficiency, which is suitable for the suggested AWS requirements for aerospace applications must be met by 60%. ANN model provided a good agreement with experimental results (R² = 0.93). However, due to the relatively small dataset (L27 design).