Abstract
There is a clear physical correspondence between the oscillatory behavior of weld pool surface and penetration. However, it is difficult to sense and resolve the oscillation characteristics of weld penetration from the multidimensional oscillation behavior of weld pool surface, which is influenced by the droplet transfer process during pulsed gas metal arc welding. In this work, a method for measuring the oscillation behavior of the weld pool of pulsed gas metal arc welding using laser vision is proposed. The image processing algorithm based on the reflected laser pattern is proposed to obtain the oscillation frequency, oscillation amplitude, curvature and width oscillation characteristics of the weld pool in different penetration states. The above oscillation characteristics are also analyzed in correlation with the surface fluctuation behavior of weld pool and the backside penetration state. And an artificial neural network model is built to predict the penetration state by oscillation features. The results show that the oscillation mode on the surface of weld pool in different penetration states are all traveling wave modes, and the oscillation frequency of weld pool is directly related to the oscillation mode, and there is no sudden change in frequency from partial to full penetration. Both curvature and width are correlated with the volume of the upper surface of the weld pool, respectively. The oscillation amplitude is characterized by abrupt changes at full penetration due to the variation of the bottom constraint of the weld pool. The established artificial neural network of weld pool oscillation characteristics and backside penetration state enables real-time sensing and prediction of weld penetration state.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have