Arc spectral processing technique with its application to wire feed monitoring in Al-Mg alloy pulsed gas tungsten arc welding

► Principal component analysis (PCA) is applied for arc spectral characteristic signals extraction. ► Pulse interference resulted from the pulse current is eliminated by mean value of signal's pulse period (MSP) method. ► Wire feeding disturbance may give rise to welding seam defects and can be detected by the extracted spectral signals. ► A proper BP neural network trained with “Early Stopping” technique is used to detect these defects automatically.