Friction Stir Welding of IS:65032 Aluminum Alloy and Predicting Tensile Strength Using Ensemble Learning

Abstract

Joining two metals play a significant role in the automobile and aerospace industries and it is a challenging job. To overcome the challenges, Friction Stir Welding (FSW) is considered as the solid-state joining technique in the industries as it is a new and unique way. As the field of machine learning is extending its applicability to different fields, it can even be applied in the field of welding. The objective of this paper is to give a review of FSW of aluminum alloys. In addition, machine learning regression models are developed to predict the tensile strength of IS:65032 aluminum alloy by taking rotational speed, welding speed, tool tilt angle, and tool pin shoulder diameter as the input parameters. In this work, the ensemble learning approach is adopted to develop models as it uses the wisdom of many learning algorithms to achieve better performance by filling in the gaps of learning ability. A comparative study was done by considering coefficient of determination and Root Mean Square Error (RMSE) of the Bagging and Boosting models in order to determine a robust regression model for predicting the tensile strength. From the conducted experiment it was concluded that Gradient Boosting Regressor performed better than the other ensemble models and feature importance of independent variables is also evaluated.KeywordsFrictions stir welding (FSW)IS:65032 aluminum alloyEnsemble learningTensile properties