Energy Detection and Identification of Welding Line Based on Arc Sound MFCC Characteristics

The weld line energy represents the arc energy received by the weld and is closely related to the weld penetration. Aiming at the detection and identification of MIG welding line energy, a recognition method based on arc sound MFCC characteristics is proposed. The arc acoustic signal is formed by the convolution of the excitation pulse of the sound source and the impulse response of the sound channel. Based on this, the auditory perception mechanism of the human ear is simulated, MFCC characteristics are extracted, and the channel envelope of arc acoustic signals in the Mel frequency domain is obtained. The MFCC feature of arc sound is used as the feature vector to establish a support vector machine model. The calculation results show that the overall recognition rate of different heat input in the welding process by this method is 99.25%, and the calculation speed is faster, which provides a new method for online monitoring of welding quality.