Research On Characteristic Of MAG Weld Pool Visual Image And Control Model In One-Side Welding With Back Formation Weld

In this paper, lots of one-side welding with back formation experiments of short-circuit transfer were performed, and visual features of MAG welding pool of different penetration was the core of this subject, based on the clear pool images obtained by passive visual sensing technology. A multi-parameter monitoring system of weld penetration formation was established, this can provide related data and models for online control of welding qualityA passive visual sensing methods coupled with near-infrared industrial CCD were used to acquire clear pool image of short-circuit transition MAG welding. The weld pool image processing system algorithm was optimized by using digital image processing technology, and the program execution efficiency was greatly enhanced, which meets the requirements of welding penetration real time control well.Weld pool obliquityθand tail contour angleβwere proposed based on the existing geometry parameters system of weld pool and related digital processing algorithm was designed. The geometry parameters of different type weld pool were extracted, which shows that when-5°<θ<5°, the torch is in the middle of the groove.The effects on welding penetration of heat input, root gap, groove edge size and electrode extension were analyzed on the basis of one-side welding with back formation experiments. Combined with the concept of back angle a, the geometry parameters of weld pool in different penetration state were summarized as follows:non-penetrated weld pool generally seemed like oblate withα>90°andβ> 120°; penetrated weld pool was near normal oval with a<90°andβ<110°; compared with penetrated pool, the collapsed type showed more slight withα<60°andβ<80°.The image of collapsed welding pool was identified from two aspects:the arc location and the weld pool width variation. The welding collapse phenomenon was indicated if the arc location declined ten pixels(△y>10) in the vertical axis and the pool width significantly shrinked 30 pixels (△W>30) between two adjacent frames. Mean of gray pixels (μ)was proposed to identify the image of burning through welding pool, ifμdroped rapidly by more the 20, the probability of burning through was great.Standardization curve models of weld pool tail contours for typical images of penetrated, non-penetrated and collapsedweld pool were established by using "Normalization-Discretization" method. Two characteristic parameters mean deviation E and mean square deviation s were proposed. The characteristic parameters threshold range were obtained from lots of experiments, which was used to estimate the pool penetration type.On this basis, a multi-parameter monitoring system for weld penetration formation was established, which can promote the developing steps of welding quality online control.