Research On The Weld Seam Tracking System Based On Visual Servo

As the welding manufacturing automation technology continues to improve, the welding manufacturing processes are experiencing from manual to automatic. The traditional manual welding operation is one of the hardest jobs. The welding process automation, robotization and intelligentization have become the development trend of welding industry. The usage of welding robot has acted a principle symbol of the automation technology modernization. With the development of the modern society and the higher demand on product quality, production efficiency, working conditions and environmental protection, there is a pressing need to achieve welding automation and intelligence. And moreover, the intelligent welding robot becomes particulately necessary in the welding operation of space station, the nuclear environment and deep-water condition.The seam tracking technology based on vision sensing not only makes a feature of the access to vast information, high reliability and wide use, but also greatly increases the external adaptability of welding robot combined with computer vision and image processing technology. In this dissertation, visual servo approach is introduced into the research on seam tracking, and the relevant technologies involved in the visual sensor system calibration, image processing recognition and weld seam tracking servo system design are in-deeply studied.During the establishment of seam tracking system in this paper, the visual servo sensing devices are set up and the active vision sensor technology is adopted. The vision sensor image acquisition device on the basis of laser technology is also designed to make the camera meet the requirements under two different capturing circumstances:indoor and arc. By analyzing the spectral characteristics of arc of MIG welding and CO2 welding, semiconductor laser which has a wavelength of 650nm is selected as the light source. In the meanwhile,650nm-wavelength and 15nm-half-bandwidth laser narrow-band filter and subtractive filte are chosen as supporting devices. These cannot only meet photosensitive range requirements of CCD vision sensor but also requirements of continuous output power.Seam tracking system on the one hand drives welding torch along the weld seam with a higher precision in the location and pose, on the other hand keeps the coordination of welding process parameters for precise trajectory control in the control procedure. Seam identification system uses TMS320DM642 as the master chip, mainly including video encode/decode module, memory module and communication module to achieve the weld image acquisition, processing and recognition. The weld image processing involves the density histogram, image smoothing, image enhancement, image restoration, image binarization and edge detection. According to actual welding conditions, this paper achieves the better results by improving some algorithms, and enhances the accuracy of the weld seam recognition. DSP/BIOS(variable size real-time multi-tasking operating system kernel) is transplanted to TMS320DM642 to meet real-time requirements better, and be easy to maintain and upgrade without impact on the response time of certain key thread. A splitting algorithm of multi-segment approximation is designed in the welding image recognition process which divides segmentation results of V-joint contour into four kinds of situations for classification. And SUSAN corner detection algorithm is used to obtain its feature extraction in accordance with the identification of the V-joints. Seam tracking servo system adopts MC9S12XDP512 as the master chip, mainly including optical isolation module, motor drive module, quadrature encoder module, signal shaping module, RS232/CAN communication module and human-machine interface module. Because of precise requirements on welding operation, this paper designs a double-loop negative feedback control model. The outer loop controller is the speed controller, and the inner loop controller is the current controller, both using PI control algorithm to automatically recognize the welding track and to accurately track and control.