Artificial Intelligence Algorithms for Prediction of Tensile Strength for Process Parameter Variations in Friction Stir Welded Al7075 and Al 6063 Joints

Abstract

The most important breakthroughs in contemporary computer science can be traced to the interaction between machine learning and the optimisation process. Mechanical automobile industries rely heavily on optimisation because of the savings it brings in terms of manpower, time, and production. In recent studies, researchers have sought to maximise the Ultimate Tensile Strength (UTS) of the aluminium dissimilar welded joints, which are mostly used in vehicles, by optimising the friction stir welding process. The XGBoost method and the Decision Trees regression model are two of the machine learning techniques used for this task. The UTS is the output variable, whereas the input variables are the Rotational Speed (rpm), Weld Speed (mm/min), and Tool Profile (Cylindrical, Hexagonal and Taper). Mean Square Error (MSE), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) are found to be 6.77, 2.281 and 2.934 respectively for testing sets when using the XGBoost method. The MSE, RMSE, and MAE for Testing Sets with Decision Trees regression model, are 10.767, 3.281, and 3.255 respectively. We can therefore conclude that XGBoost is more effective and yields more reliable results than the Decision Tree approach