Weld pool image recognition of humping formation process in high speed GMAW

Aiming at the change rule of weld pool during the humping formation process in high speed GTAW, CCD vision system was used to track and collect. A segmentation method based on Fuzzy C-Means collaborative active contour model was proposed to segment the weld pool image. The results show that the step change of the weld length was the main image feature reflecting the humping formation. The weld length sequence was fitted into waveform and decomposed by Symlets2 wavelet. It was found that d₂ wavelet decomposition can identify the step change of weld length. Threshold was set on the d₂ wavelet to obtain the spike characteristic signal which reflected the step change of the well weld length. The formation of the hump defect can be recognized well and the monitoring and control of the humping bead can be realized preliminarily.