Study On Weld Pool Image Processing And Model Identification Of GTAW

As an important manufacturing technology, Gas Tungsten Arc Welding(GTAW) is widely used in industry. Real-time weld arc control is important to welding automation and to improve joint quality. Building an analysis model of weld pool dynamics is main step of welding process control, is very important to joint quality. Vision sensor can capture a lot of weld pool information, how to detect the pool information effectively, pick up character parameters of welding process and how to build a veracious identification model is precondition for welding automation control research.The thesis is focused on the weld pool identification model based on vision sense which the parameters are the weld pool width and penetration, and it has been researched by a set of weld arc control experiment apparatus which include the CCD (Charge coupled device) sensor, the working platform and the step drive control system and so on. Firstly, the CCD vision sensor and other assistant apparatus are used for gathering clear image of molten pool from the obverse welding line with many technical experiments. Secondly, the pool image is filtered by the traditional filtering methods, then, the edge of molten pool image is enhanced by top-hat and bottom-hat(top-bot) of Gray-scale Morphology, and the image is processed to binary image, in the next step ,the molten pool image is processed by dilation,erosion,open,close and combinatorial morphology between them. Thirdly, the edge is detected by edge detection operator, which make up of the basic operation, and then the shape information of molten pool in different welding process are extracted availably. Finally, based on this information, the weld pool image character parameters, for example, the pool width and so on, are calculated.As technical experiment research provides the foundation for capture of weld pool and modeling for weld pool dynamics. This paper makes many groups of technical experiment in the condition of changing parameters, and investigates the influence of technical parameters on weld pool width and penetration. The model of the welding process is designed in two different ways based on experiment parameters and the data of the weld pool width and penetration, which are multi-member regression analysis and factorial experimental design, and the models are testified with experimental data at the same time. At last, these two different ways for modeling are contrasted, and theses models reflect the impact of the parameters to width and penetration.