Machine learning based tensile strength prediction and analysis on friction stir welded dissimilar joints (AA6082-AA5083) using conventional and hybrid tool pin profiles

Abstract

Friction stir welding technique is considered to produce weldments with promising strengths. Owing to the minimum set of parameters only to be controlled during welding, this green welding technology possesses noticeable advantages over other welding techniques. As the world of Artificial Intelligence is taking great leaps, the application of machine learning is serving as a great advantage in several research domains. This paper focuses on comparing different ensemble learning-based boosting approaches and implementing an optimum regression model for predicting tensile strength. This was trained upon the experimented data along with the analysis comprehending the influence of parameters involved during the process of welding with the concept of Explainable AI. The results obtained were competent with a 0.985 R2 error and 5.292 RMSE.