A random forest-NSGA-II framework for optimizing heat treatment conditions in friction stir welding of Al7075 alloy

Friction stir welding (FSW) is a promising solid-state joining technique for high-strength aluminum alloys such as Al7075, as it minimizes defects common in fusion welding. However, the mechanical performance of FSW joints is highly sensitive to thermal cycles and microstructural evolution in the weld zones. This study evaluates the effects of pre- and post-weld heat treatments on the tensile properties and hardness distribution of Al7075 FSW joints using six designed processing conditions involving annealing and T6 treatments. In the experimental phase, tensile and hardness tests were conducted, supported by microstructural analysis. The results reveal that post-weld T6 heat treatment significantly enhances both tensile strength and hardness. In the optimization phase, a multi-output surrogate model based on the Random Forest algorithm was developed to capture nonlinear relationships between processing parameters and key mechanical properties, including yield strength, tensile strength, elongation, and hardness across the stir zone, thermo-mechanically affected zone, and heat-affected zone. The model achieved strong predictive performance, with an R² of 0.907 and a MAPE of 5.15%. Subsequently, the NSGA-II algorithm was employed for multi-objective optimization, generating a Pareto front that elucidates trade-offs between strength, hardness, and ductility. The optimal condition achieved a tensile strength of 466.9 MPa with a uniform hardness distribution of 180–195 HV.