Friction stir weld classification by applying wavelet analysis and support vector machine on weld surface images

• Monitoring of surface welds are performed for FSW by analyzing the weld images.
• Wavelet features are extracted from the images by performing discrete wavelet analysis.
• The extracted features are classified into good weld defected weld by performing Support Vector Machine (SVM) classification.
• Average classification accuracy of 98.375% for Gaussian kernel and 97% for polynomial kernel is achieved.
• The proposed method is cost effective and reliable for online defect classification of FSW.

Online monitoring of friction stir welding (FSW) is inevitable due to the increasing demand of this process. Also the machine vision system has industrial importance for monitoring of manufacturing processes due to its non-invasiveness and flexibility. Therefore, in this research, an attempt has been made to monitor friction stir welding process by analyzing the weld surface images. Here, discrete wavelet transform has been applied on FSW images to extract useful features for describing the good and defective weld. These obtained features have been fed to support vector machine based classification model for classifying good and defective weld with 99% and 97% accuracy with Gaussian and polynomial kernel, respectively.