Of Co <sub> 2 </ Sub> Gas Shielded Arc Welding Interference Factors Research And Design Of The Computer Identification System,

With the rapid development in welding automatization, the problem how toautomatically indentify and mornitor the interfere factors in welding process is urgently tobe settled. Beginning with the dynamic welding current and arc voltage signals in weldingarc itself, the signals'characteristics that correspond with the welding interfere factorswhich directly result in the defect in welding seam was studied. A welding interfere factorsidentifying system that can be used to identify abnormal factors in welding process such asthe fluctuation of wire-feed speed, the oily parts on weld line and the lack of shielding gasin real time was successfully developed. The system's great value has been established inguaranteing the quality in welding, avoiding the waste of materials and power, andimproving welding productivity. The critical research method is to design a weldinginterfere condition identifier using ANN (Artifical Neural Network).The welding parameter under normal condition was determined by orthographic test,and the interfere factors was intentionally set to imitate the true welding circumstances.The welding electricity signal can be accurately and integrately acquired in real time by thesignal-sampling module that was developed by linking PCI-9112 multifunction dataacquisition card with computer and programming in Visual C++ 6.0. A number ofexperiments were done to acquire abundant sample data; the samples'original singalwaveform, their PDD (Probability Density Distribution) and PSD (Power Spectral Density)were employed to finely analyze their characteristics. The PDD method was determinedlyused to calculate the eigentvector of the original acquired sample data.With the help of ANN TOOLBOX provided in MATLAB, the two types of weldinginterfere factors identifier was seperately designed by using the SOM (Self OrganiziationMapping) and LVQ (Learning Vector Quntilization). The interfere factors'eigentvectorwas used to train and simulate the two ANN identifiers. The simulation results proved thatthe LVQ identifier's general recognition rate in identifying the interfere factors is 92.5percents, and it has higher speed in training and working, higher recognition rate and moreeasily to be put in practice than the SOM identifier.Finally, the different programs that were respectively developed in Visual C++ 6.0and MATLAB were seamlessly integrated into the final identifying system through VisualMATCOM. The advanced laboratory equipments and the scientific experimental techniqueensured the satisfied research achievements.The system can instantly identify the abnormal welding condition result from interferefactors in the welding process and has important practical value. This research inauguratesa new method for the use of LVQ ANN in welding interfere identifying field and hasimportant theory value.