Research On Real-time Wavelet De-noising System Of Arc Sensor Signal Based On DSP

In recent years, arc sensor has been studied extensively because of its excellentperforming involving real time and stability in seam tracking. However, the intrinsicsignal of charactering the weld location information has always been submerged by theintensive interference noise in the actual welding process. Therefore, reducing noise andincreasing signal to noise ratio is the key to achieving the aim to successfully apply andpopularize the arc sensor in seam tracking. At present, although a lot of researches havebeen done in this respect, de-noising effect is not ideal. Through Analyzing the signalcharacteristics of the arc sensor signal, a real-time wavelet de-noising system of arc sensorsignal based on DSP was developed in this paper, which was based on the characteristic ofwavelet Time-frequency and the theory of moving data window in which the date length isfixed.Firstly, the basic theory of wavelet de-noising was introduced in the first chapter ofthis paper, briefly; and then according to the signal characteristics of the arc sensor signaland the real-time requirement of the seam tracking system, a real-time wavelet de-noisingmethod of arc sensor signal was proposed based on the moving data window, whichintercepts the arc sensor signals real-time and dynamically. In order to reduce timecomplexity of the sensing signal de-noising algorithm and ensure the real-time of thealgorithm, the orthogonal wavelet with compact support was chosen, the first and thesecond floor calculation of the wavelet decomposition was discarded, and the combinedmethod of the wavelet decomposing and reconstructing method and the non-linear wavelettransform threshold de-noising method was used to reduce the noise of arc sensor signal.Finally, the main factors which affect the de-noising effect and real-time in this methodwere analyzed in detail.Secondly, According to the real-time wavelet de-noising method of arc sensor signal,DSP hardware system was designed, which can quickly achieve real-time waveletde-noising algorithm for arc sensor signal with its high-speed floating-point capability.Thirdly, on the basis of the detailed analysis of the Mallat fast wavelet transformalgorithm characteristics, the fast implementation of the Mallat algorithm in software wasdesigned, and its program flow chart was given. In order to suppress the boundary effectcaused by the wavelet de-noising algorithm and reduce the signal phase shift, a newboundary extension method based on the intrinsic signal cycle of arc sensor signal was proposed according to the signal characteristics of the arc sensor signal. The validity andreliability of this extension method is verified through MATLAB simulation.Finally, In order to verify the de-noising effect and the real-time of the real-timewavelet de-noising method for arc sensor signal, a lot of on line real-time waveletde-noising experiments was done on the actual process of rotating CO2welding. Theexperimental results showed that this method not only can remove strong interferencenoise and improve signal to noise ratio, but also has very high real-time.