The Study On Information Processing And Controlling Of Arc Sensor For Automatic Tracking System In Welding

The arc sensor makes use of arc itself as a sensor. According to the characteristic of the arc, the variables of the current and the voltage are drawn asthe sensor signals during welding. In this thesis the basically theoretical andexperimental analyses on the arc sensor were done, aiming at the quantificationalanalysis on influence of the various processing parameters on the arc sensor biassignal. A welding robot was taken as welding manipulator. The adoptive power supply is an inverter type. The diffluent resistance was connected into the mainloop. The current signal from the resistance was transmitted to the AI port of thedata sample card after being magnified, and then was changed into digital value byA/D converter with high sensitivity of the data sample card. Under Windows operating system, Visual Basic 6.0 programming languages were utilized to carryon programming control of the data sample card. Sampling frequency was set up as 2 KHz according to the needs of analysis. The procedure was carried on in the course of welding and the wave can be displayed real-time. The gathered data canbe stored in the text file automatically after sampling in order to do analysis again. After collecting the signals of the electric current and the swing period within actual welding process, the various kinds of commonly used digital signal filterswere applied to process the signals, such as smoothing filter, mid-number filter and shape filter, in allusion to the short circuit transition nature of the CO₂ arc welding. The results showed that the multipoint smoothing filter can get rid of the burrcurrent and reduce the peak value of short circuit current but has limitation. If thewindow length is widened again, signals will be too smooth. Furthermore, thephase of the signals will be seriously brought forward compared to the originalsignals. In addition, if applied to control system, the multipoint filter would influence the real-time character of the tracking system. The mid-number filter hasthe good effect on getting rid of burr current because of the spray if the trackingsystem adopts the lower sampling frequency under low swing frequency and alsoon getting rid of short circuit peak current. The shape filter can get rid of noisevalidly from random high frequency signals. But when swinging in high frequency,the short circuit peak value currents are difficult to be got rid of which lie at thebeginning, half-period and end-period of the cycle. The short circuit transition with big droplets appears frequently in these three points when the length of arc is shortest, therefore this should be one phenomenon that one cycle has. The shape filter cannot get the good result within two cycles processing at high swinging frequency. Meanwhile the wavelet theory which was developed fast in recent years was used to resolve the problems of processing the signals of welding currents. On foundation of numerous experimental and theoretical analyses, db4 wavelet base was selected for use and Birge-Massart tactics were adopted to confirm threshold value by soft threshold value filtering on eight layers of wavelet to get the better results. Compared between the processed current waves and the original ones, the changes on the peak value and phase are negligible, but improvement on signal wave is very significant. The disturbing from various random signals to arc sensor can be got rid of better and this is very much meaningful to filtering of CO2 welding sensing current. After getting the better filtering results, the influence of the swing frequency, swing range and welding parameters on the sensitivity of arc sensor was effectively evaluated and analyzed. In the course of analyzing, on one hand, welding current value gathered actually was utilized; on the other hand, the theory modeling and emulation was used. Through analyzing, we found the improvement of arc sensor sensitivity by increasing swing frequency, that is to say, the response to the deviation signal was speeded up. Increasing swing range has no influence on the sensitivity of arc sensor. But the bigger swing range can improve the sensing current wave and raise Signal Noise Ratio (SNR). Meanwhile, the swing welding current was analyzed under large parameters. The results showed that in this case the signal waves were better than under small parameters. It is much easier to gain and process the deviation signal of arc sensor. Visual Basic, C and Matlab software were adopted to make united programming in the course of analyzing. There are two movement ways of arc sensor, swing and rotating. The principle of the rotation arc sensor is to get the higher sensitivity by quicker rotation frequency in order to acquire the higher tracking accuracy. The swing scan way has its advantage although the swing frequency is lower and the swing range can be adjusted wider. Such characteristics bring it to be applied in the groove and multi-layer welding seams. In allusion to this nature, the method to get frequency...