摘要
In order to recognize various metal transfer modes, by the creation of apattern recognition system for metal transfer mode, five kinds of spectrum signal in gas metal arcwelding (MIG, MAG and CO_2) are collected and taken as training samples. These samples arepretreated by computer. Several key characteristic parameters of the spectrum signal are creativelyextracted, and a corresponding recognition function and a minimum-distance-classifier areconstructed. The results show that using this method, the pattern recognition of several kinds ofmetal transfer mode for the metal gas arc welding can be done successfully. It has good accuracy andrecognition precision, basis for controlling the metal gas arc welding metal transferautomatically, and relative important parameters in welding process, such as the frequency ofdroplet transfer and the approximate diameter of each droplet, can also be obtained.
In order to recognize various metal transfer modes, by the creation of apattern recognition system for metal transfer mode, five kinds of spectrum signal in gas metal arcwelding (MIG, MAG and CO_2) are collected and taken as training samples. These samples arepretreated by computer. Several key characteristic parameters of the spectrum signal are creativelyextracted, and a corresponding recognition function and a minimum-distance-classifier areconstructed. The results show that using this method, the pattern recognition of several kinds ofmetal transfer mode for the metal gas arc welding can be done successfully. It has good accuracy andrecognition precision, basis for controlling the metal gas arc welding metal transferautomatically, and relative important parameters in welding process, such as the frequency ofdroplet transfer and the approximate diameter of each droplet, can also be obtained.
基金
This project is supported by National Natural Science Foundation of China (No.59990470).
