MIG Welding Arc Sound Signal De-noising Technique Research

As one form of the Acoustic Emission(AE) signal, generated in the welding process, with many information indicating the welding wuality, welding arc sound can be regarded as the valuable and potential source signal for online monitoring of the welding penetration depth and quality of welding.As the random non-stationary signal, welding arc sound signal is often heavily contaminated by sophisticated background noise such as the sound sensor noise, welding circumstance noise (welding sourcepowermechanical vibration noise, voice of operators and mechanical noise generated in the welding process), welding splashing arc sound. In order to better indicate the characterization of the arc weld pool, it is necessary to do the noise reduction with the noise. Thus this paper makes a thorough elaboration of MIG welding arc sound denoising technuiques including the software design for the welding arc sound denoising system, the construction of hardware circuits, and the techniques of arc sound signal blind source separation and noise reduction with the application of Independent Component Analysis(ICA) theory and Wavelet theory.First of all, by means of the powerful audio processing capabilities of LabVIEW and advanced signal processing toolkit, with the application of common PC sound card and voice microphone, the multi-channel audio signal acquisition system and threshold wavelet denoising system are constructed. In addition, by use of MATLAB simulation software, FastICA algorithm is edited to conduct the blind source separation of welding circumatance noise.Secondly, the possible welding noise signals are classified and a detailed analysis of the noise generating mechanism and signal characteristics is conducted, as well as the noise suppression and signal processing methods, and the influence to the arc sound.Finally, accoding to different characteristics of the noise signals, reasonable denoising shceme would be adopted to conduct the processing of actual MIG welding arc sound. By means of reasonable selecting and design the hardware circuit, necessary electromagnetic shielding, isolation and filtering, the maximum suppression of arc splashing, the real-time gathering of arc sound signal is conducted. Then the independent separation of arc sound signals gathered in the multichannel (Channel 2 and 3) is conducted to get the arc sound signal in conformity with the requirement and eliminate the welding circumstance noise by threshold wavelet denoising processing, eliminating the weld splashing arc sound signal, and purifying arc sound signal.Through a series of arc sound signal denoising, a good denoising effect has been gained both in the signal waveform in time domain, frequency domain power spectral analysis as well as Signal to Noise Ratio( SNR) evaluation.