During Friction Stir Processing (FSP), the material is plasticized due to the heat generated by friction at the interface of the tool shoulder and the material. The mechanical properties are influenced by the changes in microstructure, or phase transformation, that occur during FSP. The peak temperature (T \(_p\) ) prevailing during FSP influences the flow stress and metallurgical bonding, which in turn, influences the microstructure and properties of the processed sample. Key parameters influencing T \(_p\) include tool rotation speed (TRS), tool traverse speed (TS), shoulder diameter (SD), tool geometry (TG), and axial load (AL). However, the exact correlation between these FSP parameters and peak temperature evolution is not fully understood. The current study employs four different machine learning (ML) techniques – Artificial Neural Network (ANN), Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting (GB) – along with data augmentation to develop predictive models for the average peak FSP temperature and the FSP process and tool geometry parameters for Mg \(_4\) Y \(_3\) NdZr/Al/Ti/Sn hybrid composite. The predictive models utilize three input features: two processing parameters (TRS and TS) and a single tool geometric parameter (SD). Experimental data obtained using the Taguchi L27 design of experiments, along with synthetically created data, are used to train and test the predictive models, enabling them to capture the underlying correlations between the input features TRS, TS, and SD, and the output feature T \(_p\) . The eventual goal is to identify the most accurate predictive model for T \(_p\) and to determine the relative influence of the input features on T \(_p\) . The results reveal thatSVR is the most optimalmodel for predicting Tp,as evidenced by its lowestnormalized MSE (0.0069)and highest \(R^{2}\) (99.14%) on the testing data. Furthermore, it is found that both TRS and SD have a positive impact on T \(_p\) , with an increase in both corresponding to a higher T \(_p\) , with SD being more dominant. In contrast, TS has a negative impact, where an increase in TS results in a lower T \(_p\) . To supplement the research findings, a multi-functional graphical user interface (GUI) has been developed, implementing the optimal SVR model and enabling the determination of T \(_p\) for any given combination of input features.
