Abstract: Arc sound is well known as the potential source signals for monitoring the weld penetration status, which is very important to the welding process quality control and automatic production. The dynamic and non-stationary arc sound signal in MIG butt welding with spray transfer is selected as an object. The denoising principle and method based on wavelet transform are applied to arc sound signal, and the effect of noise reduction is achieved. Through a reasonable choice for sample data set on that basis, the internal relationship between arc sound signal and penetration status is analyzed from time domain, frequency domain and time-frequency domain. The energy and spectrum of arc sound signal are highly correlated with penetration status and signal energy variation. The signal from 1 500 Hz to 4 500 Hz can accurately detect the changes in penetration status in welding process. The results indicate the feasibility of the penetration state monitored by use of arc sound signal, and a cheap and effective way is provided for controlling the quality of MIG welding process.

Key words: Arc sound, MIG welding, Penetration, Wavelet transform