Machine learning-assisted prediction of mechanical and wear properties in friction stir welded AA7075/graphene nanoplatelet composites

The present study combines machine learning (ML) with experimental research to determine the mechanical and wear behavior of friction stir-welded AA7075/graphene nanoplatelets (GNPs) composite joints. Tool rotational speed (TRS), tool traverse speed (TTS), and tool plunge depth (TPD) were chosen as important input parameters, whereas ultimate tensile strength (UTS) and microhardness were used as important output responses. A total of 16 experiments were conducted, with 80% of the data used for training the ML models and 20% reserved for testing. There are five regression-based methods, including linear regression, support vector regression, random forest, extreme gradient boosting (XGBoost), and k-nearest neighbors, that were developed to evaluate predictive performance. Random Forest and XGBoost had better accuracy than the others. In the UTS prediction case, XGBoost achieves an MAE of 7.81 MPa and an RMSE of 12.02 MPa, while random forests yield an MAE of 11.20 MPa and an RMSE of 15.09 MPa. Random forest yielded the best results, with an MAE of 0.79 HV and an RMSE of 1.25 HV in microhardness prediction. The optimal process conditions (TRS of 800 rpm, TTS of 20 mm/min, TPD of 5.80 mm) resulted improvement in joint quality, yielding a UTS of 821 MPa and a maximum nugget-zone microhardness of 143 HV. Microstructural examination proved the refinement of grains (average grain size of approximately 610 µm) because of dynamic recrystallization and homogenous dispersion of GNPs. The ductile fracture morphology was characterized by fine dimples, and the wear depth was significantly reduced (from 98 µm to 35 µm) under optimal conditions. The experimental findings demonstrate that ML, especially random forests for microhardness prediction and XGBoost for tensile strength prediction, can be highly effective in predicting weld properties and guiding process optimization to produce high-performance AA7075/GNPs FSW joints, which can be used in aerospace fuselage tanks. • Integrated ML framework to predict strength and hardness of FSW AA7075/GNP composites. • GNP reinforcement significantly enhanced strength (821 MPa), hardness (143 HV), and wear resistance. • Data-driven optimization identified optimal FSW parameters for enhanced mechanical and tribological performance. • Random Forest and XGBoost showed highest prediction accuracy and reliability.