Prediction of properties of friction stir spot welded joints of AA7075-T651/Ti-6Al-4V alloy using machine learning algorithms

In the present study, experimental works on friction stir spot welding (FSSW) of dissimilar AA 7075-T651/ Ti-6Al-4V alloys under various process conditions to weld joints have been reviews and multiple machine learning algorithms have been applied to forecast tensile shear strength. The influences of welding parameters such as dwell period and revolving speed on the mechanical and microstructural characteristics of weld joints were examined. Microstructural analyses were conducted using optical and scanning electron microscopy (SEM–EDS). The maximum tensile shear strength of 3457.2 N was achieved at the revolving speed of 1000 rpm and dwell period of 10 s. Dwell period has significant impact on the tensile shear strength of weld joints. A sharp decline (74.70%) in tensile shear strength was observed at longer dwell periods and high revolving speeds. In addition, a considerable improvement of 53.38% was observed in tensile shear strength at low dwell periods and high revolving speeds. Most significant machine learning data-driven methods used in welding such as, artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), support vector machine (SVM) and regression model were used to forecast the tensile shear strength of welded joints at selected welding parameters. The performance of each model was examined in training and validation stages and compared with experimental data. To evaluate the performance of the developed models, the two quantitative standard statistical measures of prediction error % and root mean squared error (RMSE) were applied. The performance of regression, ANN, ANFIS and SVM were compared and SVM regression model was found to perform better than ANN and ANFIS in forecasting the tensile shear strength of FSSW joints.