Arc Signal Characteristics Of Co <sub> 2 </ Sub> Welding Quality Control Methods

CO2 gas shield arc welding is of high effciency, low cost advance welding technology and widely used in the field of metal structure manufacturing. To deal with welding problem such as processing stability and quality monitoring online, the following works had been done in this thesis.Time-domain, frequency-domain, wavelet and statistics method were used to analysis current, voltage and arc sound in this thesis. Synchronization, self-correlation and mutually-correlation of three kind of signal were analyzed in time-domain. Fourier spectrum, little Fourier spectrum, power density spectrum and coherence function of signal were studied in frequency-domain. Wavelet analysis was used to extract signal from noise and to analyze frequency-sect energy of arc sound. Common parameters of signal in time domain and frequency domain were statistics.Short-circuit GMAW arc acoustic waveform, which contained plenty of information relevant to arc behavior, melting metal transition mode, arc stability, is analyzed in respect of time domain and frequency domain in this paper, it is found that arc acoustic wave is most related to differential of arc power which reveals itself as "ringing" form, occurs on arc re-ignite of short-circuit transfer, and mainly focused in the frequency band of 4-8kHz. Producing and formative mechanism of acoustic signal is approached according to its character. It is think that variation of arc power is the exciting source, shielded gas and arc column are tone channel, they are combined to generate the arc sound during the welding process. LPC forecast model was constructed according to the theory above.Several methods such as multi-resolution recognition based on power spectrum, multi-layer wavelet package decompose of arc energy, high-rank LPC linear forcast analysis and parameters statistics analysis are emplyed to extract characteristic parameters, which formed charactristic vectors mapped welding quality and stability of welding process. These vectors are employed as input parameters of Neural Network such as linear element and Back-propagation. The Neural network precented many advantages such as track and forecast welding parameters, classify welding qualities.