Predictive modeling and optimization of ultimate tensile strength in friction stir welding of Al6061 using DOE and genetic algorithms

Friction Stir Welding (FSW) has emerged as an effective technique for joining aluminum alloys, particularly Al6061, due to its ability to produce high-quality welds without melting the material. However, optimizing welding parameters remains a critical challenge to enhance the mechanical strength of welded joints. In particular, the ultimate tensile strength (UTS) of Al6061 welds is highly dependent on operating conditions, such as temperature, feed rate, and fracture strain, whose influence on mechanical performance remains complex to model and optimize. In this study, an approach combining the Design of Experiments (DOE) methodology and a Genetic Algorithm (GA) is proposed to predict and optimize FSW parameters to maximize the UTS of Al6061 welds. A multiple linear regression model was developed to establish the relationship between welding parameters and UTS, while the genetic algorithm was employed to identify the optimal welding parameter configuration. The results demonstrate a significant improvement in UTS after optimization, increasing from 315 MPa to 354.74 MPa. Furthermore, the coefficient of determination (R2) of the predictive model improved from 0.828 to 0.905, while the root mean square error (RMSE) decreased from 10.6 MPa to 8.21 MPa, indicating enhanced prediction accuracy. These results confirm the effectiveness of the Genetic Algorithm in optimizing the mechanical performance of Al6061 welds and highlight its potential applicability to other welding processes and materials.