Weld Seam Identification Based On Multi-source Data Fusion Positioning And Correction Method

Welding robots have high working efficiency and stable welding quality,which have been widely used in manufacturing industry.However,most of the current welding robots still use the "teach and repeat" work mode,which has poor real-time performance and high labor intensity.Therefore,this paper proposes a set of automatic welding seam recognition,localization,and real-time correction system based on multi-source data fusion.This system uses multiple sensors such as industrial cameras,laser sensors,and motor encoders to automatically recognize and plan welding seams and trajectories,and perform real-time trajectory correction.The main research contents are as follows:(1)Overall design of the welding seam recognition and correction system.The image processing module first recognizes and locates the welding seam from the workpiece image,and then the robot approaches the welding seam and uses the laser sensor to scan the weld groove to obtain the precise coordinate values of the middle points in the welding path.Then,the motion controller plans the welding path and welding trajectory in three-dimensional space,and finally,the robot performs welding trajectory tracking and real-time correction using data provided by the laser sensor.(2)Motion analysis of the welding robot.The improved D-H method is used to establish the kinematic model of the robot,analyze the forward and inverse kinematics,and the differential motion of the robot.The robot end tool coordinate system is established,and the pose calibration of the laser sensor is carried out to transform points in different coordinate systems into the robot base coordinate system.The kinematic parameters of the articulated robot are identified to improve the end positioning accuracy of the robot.(3)Research on welding seam recognition and localization method based on a single camera.The complete welding seam is automatically recognized using image filtering and edge extraction,and feature points are extracted from the welding seam image.Using the point-to-point method based on the imaging principle of a single camera,feature points are converted from the pixel coordinate system to the base coordinate system to obtain the positioning points,and provide the position information of the welding seam for the laser sensor installed at the end of the robot.(4)Automatic extraction of the welding seam groove center point based on laser sensor.Using the coordinate values of the positioning point and the readings of the laser sensor,the end effector’s posture is adjusted,and high-precision coordinate values for the middle points in the welding path are obtained by coordinate transformation.Three-dimensional curves are then fitted using cubic spline interpolation and welding trajectories are planned based on the arc length formula.(5)A real-time welding seam correction method is proposed,and a trajectory correction controller is designed.The theoretical derivation shows that it can reduce the tracking error.A V-shaped straight weld correction experiment is carried out,and after trajectory correction control,the maximum tracking error of the robot end effector is reduced to 0.0899 mm,and the root mean square error is 0.0392.The results show that the real-time trajectory correction method proposed in this paper is reasonable and feasible and can effectively improve the accuracy of the real-time tracking of welding seams.