Abstract
<title>Abstract</title> <p>Joining dissimilar aluminium alloys is still a challenge due to the differences in their thermal and mechanical properties which cause non-uniform deformation and poor joint performance. The present study aims to study the mechanical behavior, microstructural evolution and intelligent optimization of AA1100–AA2014 dissimilar aluminium joints produced by rotary friction welding. The Taguchi L27 experimental design was used to change the rotational speed, friction pressure, forging pressure, friction time, and forging time. The tensile test, microhardness test, scanning electron microscopy (SEM) and X-ray diffraction (XRD) were used to assess the weld quality. A deep neural network (DNN) combined with the NSGA-II algorithm was designed to predict and optimize welding performance. The optimized joints had a maximum tensile strength of 142 MPa, hardness of 122.3 HV, and a predictive accuracy of R² = 0.987, indicating good agreement between the experimental and predicted results. Microstructural analyses showed fine grains, defect-free interfaces and no harmful intermetallic phases. The proposed hybrid experimental–AI framework is an efficient decision-support tool for intelligent process optimization and reliable dissimilar aluminium joining in the aerospace, automotive and lightweight structural manufacturing sector.</p>