Abstract: The working environment of inverter power sources is filled with strong interference, high voltage and current. To obtain useful signals of the inverter welding machine from harsh environment is the key to R&D and technological evaluation of all-digital welding power source. A new method is put forward after analysis and research on the principle of pulse width modulation (PWM) driving, i.e. the current signal of inverter welding machine at PWM driving level is denoised first, then noise filtering of the whole welding current signal is carried out. Hard threshold function is constructed to carry out wavelet decomposition of signal, and the detail coefficients are denoised. At the same time, modulus maxima discriminant function is constructed to search noise singular points from detail coefficients. The search results are used to denoise approximation coefficients linearly. Simulation experiments based on selection of appropriate correlation parameters show that the method can filter welding current noise effectively, ensure distortion-free signal and obtain useful signal characteristics, thereby laying the foundation for the research and performance evaluation of all-digital inverter power source.

Key words: All digital inverter, Modulus maxima, Performance evaluation, Threshold, Wavelet transform