Machine learning classification approach for predicting tensile strength in aluminium alloy during friction stir welding

Abstract

The machine learning (ML) methodology is receiving significant attention as a promising approach for addressing and modeling manufacturing challenges. This research designed an ML framework to forecast the tensile strength of aluminium alloys produced using friction stir welding (FSW). The dataset comprising 213 samples of aluminium alloy sourced from peer-reviewed literature was used. A range of ML algorithms such as decision tree, random forest, adaptive boosting classifier (ABC), k-nearest neighbors, gaussian naive Bayes classifier, and support vector machines were applied to the dataset. The findings revealed that the ABC algorithm attained the highest accuracy, reaching 81.6% among the models tested. This study highlights the effectiveness of ML methodologies in predicting the tensile properties of FSW-manufactured aluminium alloys, thus driving progress in the field of welding and joining.