Condition monitoring of FSW tool using statistical approach

Friction stir welding (FSW) is used to combine high-strength aluminum/magnesium alloys and other metallic alloys that are difficult to weld by traditional fusion welding in aerospace, aircraft, and automotive applications. Since FSW tool is the main parameter for root cause for occurrences of rejection in Friction Stir Welding there is a need for monitoring these tools and this will help most of the automobile companies using this welding for achieving zero rejection ppm. Hence the tool is monitored for the different fault conditions like air gap, notch, good, misalignment and one side lift. Obtaining the vibration signals from the FSW tool using the piezoelectric accelerometer, major process of machine learning i.e., feature extraction using histogram approach, feature selection using attribute evaluator and feature classification using J48 decision tree classifier, Random Forest and Random tree classifier is performed to find the maximum prediction accuracy.