Characteristic Analysis Of Dynamic Signals In CO₂ Arc Welding Processing

With a view to realize on-line welding quality diagnosis based on characteristics of dynamic signals in welding processing several studys on singnal feature and the correction between characteristic class and welding sprays are done in this paper as follows,which use arc sound as mainly studied object. In order to profoundly recognize the short-circuiting transferring of CO2 welding and correctly collect signals, a signal collecting system has been developed with KS2062. The experiment results have proved that this system has several advantages such as complete property, high reliability and so on. The signal colleted software programmed with TurboC3. 0 can be used to off-line display and previously treatment of dynamic signals. The signal processing software programmed with Matlab5.3 can be used successfully to fulfill characteristic analysis. The generated princz~les and frequency distribution of arc sound are investigated. Wavelet analysis is utilized on eliminating noises and interpreting zero crossing of arc sound and welding voltage signature. The results have been founded that arc sound is created by changed of the arc length and its zero crossing not only is of lower frequency distribution but also badly corresponds with welding processing. In addition, five layers decomposed of welding voltage signature by dbl wavelet, the zero crossings in dl, d2, and d5 can be used to divide the short-circuiting transferring. Wavelet packet analysis and Welch method has been advanced to delicately analysis the time-frequency characteristics of arc sound. The results have stated that there is a regulation in the power spectrum dig fribution of arc sound, and correction between max-value of different frequency range and spatter The arc sound energy in dfferent frequency ranges is extracted as characteristic class that can indicate changes in the welding processing Correction between every element and spatter is evaluated by statistic test to reduce dimensions of this class, and then get a simplest characteristic subclass. This subclass composes of energies in the 1, 3, 7, 9, 11, and 14 frequency ranges of arc sound signature 6 layers decomposed by db4 wavelet pocket. The base to fulfill on-line diagnose of CO2 welding processing in short circuiting transfer mode with characteristic class of arc sound has been advanced in this paper...