Hardness Prediction of Friction-Stir Processed Copper Composites Using Machine Learning Techniques

Friction-stir processing, a surface modification method is specifically used for the improvement of mechanical properties at exterior surfaces of different nonferrous alloys. The surface hardness of processed copper composites is largely dependent upon the input processing parameters. Considering the hardness which is a widely investigated material property relying upon the input processing conditions, the development of effective machine learning (ML) models would be a valuable tool to predict hardness in friction stir processing of copper composites. At present, an initial attempt in applying different ML models to predict the hardness of different friction-stir processed copper composites is made. Different ML algorithms such as decision tree, random forest, extra tree, and adaboost regressor are employed. The prediction accuracy of these models is investigated using data that includes various process parameters and hardness values obtained from previously published literature. The finding reveals that tree-based ML models have good prediction capabilities and may be used as suitable tools for predicting the hardness of friction-stir processed copper composites, to avoid intensive experimentations. Overall, the extra tree regressor model was found to have the best predictive performance with accuracy R2 score 84.28 among all the ML models.