Detection Of Weld Penetration Based Upon Visual Sensing During Arc Welding

In order to solve the problem of measuring and controlling the weld penetration state in industrial production, based on the characteristic of GTAW arc spectrum, pool vision sensing system is developed for weld pool images collection by using the method of passive light, and the pool images are clear. According to gray value features of weld pool images, a relevant algorithm for images processing is designed, the pool’s boundary is detected completely and the relevant geometric parameters are picked up. The influence law of weld process parameters to weld width on the back is found by doing lots of weld experiments. The model which describes the relationship between weld process parameters, geometric parameters of pool and weld width on the back is set up by adopting BP neural network technology. By means of fuzzy control theory, a fuzzy controller is designed with weld width error on the back and its variation as inputs, the regulating variation of welding current as output. The fuzzy controller lays a foundation for controlling weld penetration accurately.