Study Of Automatic Seam Tracking System Based On Vision Sensor

Most welding robot in use work in"teach"mode , while some others run in track planning mode. In welding process, there must be offset between blowtorch center and seam center. As we all know, welding is a complicated, nonlinear process with multi-interference, so that the thermal deformation, undercut, unfitness of butt joint and the change of seam gap length, which may affect welding quality directly, are all unpredictable. Therefore, based on"teach"mode or track planning mode, real time seam tracking can improve welding precision, especially fit for the welding of products hard for automatic welding to control, like those which may deform easily or must be loaded complicatedly,.This essay is based on a pre-research project of the 11th 5-year program, with a background of groove-free aluminum alloy 2mm thin sheet welding, which are to be used in aircraft fuel container. We used impulsive TIG as our basic welding method, and did tracking experiments on plane straight seam and plane flange seam, improving tradition"teach"mode welding robot to robot system with the function of real time seam tracking. We first build a welding robot system, including:1. Designed an appropriate dimmer-filter system, which can clearly gather welding image under base current, and solved the problem that the reflection damages image quality greatly.2. Under Visual C++ environment, we use multi-thread technology, by the integrity of PC, A/D card, D/A card and image card, to complete image gathering, image processing, offset calculating and correct voltage sending at the same time.We studied the characters of seam right before welding pool, analyzed the change of gray value in the small windows, and brought forward a edge searching method based on gray value change velocity. Developed a series of image process including image recovery, middle value filtering, edge searching, false edge discarding and least-squares procedure fitting, we successfully got an algorithm which can adapt well to different environment. The precision is in±0.15mm.Because of MOTOMAN HP6 robot's character that its correct system is not open to user, we did random experiments to get the relations between correct voltage and offset value, and set an accurate model through which we can debug controller parameters off line.We bring forward a real time tracking technology, using CCD to gain offset between torch and seam center, so that torch position can be adjusted immediately. Under the condition that welding process must be kept stably, we design a new fuzzy PID controller, ensuring that robot can adjust torch quickly and stably.We used this technology on plane straight seam, plane flange seam, and the precision can reach±0.15mm and±0.2mm respectively. Taking the precision of image process into account, the offset is no more than±0.3mm and±0.35mm.