Visual Characters Extraction Of Weld Pool And Shape Control During Aluminum Alloy GTAW Process

Visual sensing technology plays an important role in welding processcontrol, especially in shape control. Visual characters extraction is thefoundation of vision-based welding process control. Due to the specialproperties of aluminum alloy, how to establish a fast and stable imageprocessing algorithms to extract visual characters of molten pool has longbeen a difficult part. Systematic research is done in visual charactersextraction of weld pool in aluminum alloy pulse GTAW process. Besides,study on typical vision-based shape control system is also done.Firstly, the experimental system of pulse GTAW is described. Thesystem takes advantage of the reflection of arc. It uses the reflected light aslight source. Composite filter system has been used to reduce theinterference of high intensity arc light. In this way, clear image of weld poolzone can be collected. The software system is based on Visual C++.Regarding visual characters extraction in seam tracking, differentimage processing algorithms have been designed based on different traits ofstraight seam image and curve seam image. For straight seam image withsimple background, approach based on grayscale gradient of frontier is usedto split the image. For curve seam image with complex background,approach based on grayscale similarity of regions is used. The centerlineequation of weld seam and center point coordinate of weld pool can becalculated, so the deviation can be given.In GTAW process, shape parameter of weld pool has significantrelation with weld shaping quality, so four methods of extracting the shape parameters of weld pool have been designed. They are the algorithm basedon grayscale gradient, the algorithm based on edge extraction, the algorithmbased on image matching and the algorithm based on ROI detection. Thefour methods have been compared and the respective applicable scope hasbeen given. Among them, the ROI detection method has high processingspeed (about20.17ms per frame) and stable performance, so it is suitable forreal-time weld pool monitoring.Finally, the image processing algorithm has been added to GTAWsoftware system. Discrete incremental PID controller is introduced to thesystem. The system can control the weld shaping by monitoring the weldwidth and adjusting the welding current. Experiments show that the systemis effective under variable heat condition. Compared with constantspecification experiment, the system can significantly improve the weldshaping quality.