Research On Detecting System Based On Multi-sensor Information Fusion For Narrow Gap Welding

NGW (Narrow gap welding) is high-efficiency and high quality process suitable for thick plates in industrial production, and thus has been widely used in shipbuilding and machinery etc. Fusion of the sidewalls is the main problem of the welding quality. In order to improve the NGW quality, a multi-sensor information fusion method has been used. Two CCD have been used to capture the images of the molten pool from the front and the back respectively, and the welding current and voltage have been collected at the same time. Then the relations between the image of the molten pool, the welding current, the arc voltage and fusion of the sidewalls have been analyzed.A double-vision and electrical parameters synchronous acquisition and save system has been designed. It can acquire the images of the molten pool, and the welding electrical parameters can be taken through the designed circuit. A software of the acquisition system has been designed, under the Windows XP environment, based on Visual C++6.0. It can synchronous achieve and save the multi-channel molten pool images and the welding electrical parameters.In this experiment, CO2gas shielded arc welding has been used, and the passive visual sensing method has been applied to capture the image of the molten pool. Because of the interference resulted from the welding light and the splash, a great deal of experiments have been taken to choose an adapted filtering system. Finally, the best combination of filters which is near the infra-red range has been found, and the wave range is1064nm. The image processing method of the molten pool includes a median filter processing, a double canny operator processing, false edge deleting, curve fitting. Then the whole outline of a molten pool can be acquired. Compare the image processing results with the original image, the outline corresponds with the really molten pool image.After analyze of the molten pool image, the following information can be acquired. Firstly, fusion of the sidewalls can be taken based on the width of the arc, so the more smaller width of the welding arc the better fusing result can been obtained. Secondly, the bilateral symmetry of the welding torch can be judged through the image250ms cycle. Thirdly, three-dimensional information can be extracted by the information integration of the molten pool images. The basic value welding current characteristic of the NGW has been learned based on analyze of the welding electrical parameters. Base on the one-to-one relationship between images and the welding parameters, the periodic variation reason of the welding current can be found. The timing of the welding torch reached the maximal swing could be acquired based on the characteristic of the welding current waveform; extracting the information of welding arc in the image which is corresponded with the timing, then fusion of the sidewalls and bilateral symmetry of the welding torch can be judged.