Vision-based Sensing Of The Topside Weld Pool Geometry In PAW

As a high efficiency of the materials processing technology, PAW (Plasma Arc Welding) is widely used in Heavy machinery, pressure vessel, transportation, aerospace and other industries, with the advantages of focused power, less welding distortion and high efficiency. Detecting the PAW weld pool geometrical parameters from topside of work piece with visual sensor is of great theoretical and practical significance in verifying the numerical simulation results of welding thermal process and realizing the on-line quality monitoring.Aiming at the PAW process, a visual-based sensing system is designed. The proper sampling time and the focal length of the lens are selected, and clear images of top weld pool are obtained using the ordinary industrial CCD Camera.Based on the analysis of the weld pool images, a series of image processing algorithms are developed to extract the whole weld pool edge, including the Gaussian filtering, edge-enhanced algorithm and weld pool edge searching algorithm. According to the calibration result of the pinhole camera, the geometrical parameters of weld pool are acquired.The weld pool geometrical parameters including the width and the rear length are acquired under different welding current and welding speed. The regularity of the weld pool geometry is analyzed respectively. In the pulse key-holing plasma arc welding, the regularity of the welding parameters is analyzed in one period. The weld pool of the stationary PAW is acquired clearly by the CCD camera.Quantitative relationship between welding parameters and weld pool geometrical parameters are studied by the BP neural networks using the experiment data.