Supervised Machine Learning Classifier for Mechanical Property Estimation of Friction Stir Spot Welded Acrylonitrile Butadiene Styrene Sheet in Butt Configuration

The desire for Acrylonitrile Butadiene Styrene (ABS) has risen in recent years owing to its applicability in many industrial sectors. A significant problem in using this material is identifying the ideal Friction Stir Spot Welding (FSSW) parameters to attain maximum weld strength while combining ABS sheets. This work used a complete factorial experimental design to assess the ultimate tensile strength and % elongation during friction stir spot welding of ABS sheets in a butt configuration. Logistic regression, kNN, Support Vector Machine, Decision Tree classifiers were used to predict the UTS and percentage elongation. For the prediction of elongation, the model for logistic regression was 77% accuracy and for UTS it was 70.27%. The obtained accuracy rates were as follows – kNN classifier for the UTS = 100 % and for elongation = 100 %. The SVM model attained 68.91% accuracy for UTS and 63.51% for elongation, while the Decision Tree classifier had 100% accuracy for both mechanical properties.