Condition monitoring of friction stir welding tool with vibration signals using support vector machine classifiers

Abstract Prediction of Friction Stir Welding (FSW) tool conditions has been a major highlight in recent years. Continuous monitoring and predicting tool life is the most difficult task on the industrial side. This is because the occurrence of high temperatures and vibrations during the welding process can affect tool life. Aluminium (Al5083) and magnesium (AZ31B) alloys are selected as a workpiece and H13 tool steel is selected as a tool for the FSW process. The FSW was carried out with different tool conditions namely, good condition, air gap between tool and workpiece, tapered tool, one side lift in base metals, and notch in workpieces. A piezoelectric transducer helps extract the vibration signal from the rotating tool. From the vibration signals, statistical features are extracted, and these features are used as input for the Support Vector Machine (SVM) Machine Learning (ML) classifiers. The various SVM algorithms are used to classify the tool condition. A Graphical User Interface (GUI) was developed to monitor the state of the tool, which is an added benefit. Results demonstrate that the SVM model can accurately classify different tool conditions, providing valuable insights into tool health and facilitating timely maintenance interventions. The fine gaussian SVM performs better with 95% and 88.6% accuracy in predicting tool conditions for Al5083 and AZ31B alloy respectively. The proposed SVM-based condition monitoring represents a significant step towards automated, real-time tool condition assessment in FSW, promising enhancements in process reliability and efficiency.