Abstract: There is a physical relationship between weld pool oscillation frequency and states of weld beam full penetration. But the detection of weld pool oscillation frequency is very hard. According to this problem, a new detecting and sensing method based laser vision for metal pool oscillation frequency is proposed. The method principle of laser vision is introduced and the experiment system of collecting weld pool oscillation signal for pulsed gas tungsten arc welding(GTAW) is designed. The video images which reflect weld pool oscillation variation are collected. An image processing algorithm used for dealing with the reflected laser lines image based Matlab software is designed and written, also weld pool oscillation waveform is captured by this algorithm. The weld pool oscillation characteristic frequency is extracted by Fast Fourier transform. The experimental results show that laser vision method not only can successfully accomplish detecting the oscillation frequency for pulsed GTAW, but also the defect of pre-detecting methods for pulsed GTAW weld pool oscillation which is not used for sequential welding is overcome. The oscillation characteristic frequency is detected by software system. The image processing algorithm elapses about time 20 ms and meet the real-time detecting requirement for metal pool oscillation. To further realize the continual pulsed GTAW penetration control laid a foundation.

Key words: Image processing, Laser vision, Weld pool oscillation frequency