Research On Robot Welding Seam Tracking System Based On Passive Vision

Compared with traditional manual welding,robot welding has absolute advantages in welding efficiency and welding quality.In today's society where labor costs are gradually increasing,the popularization of robot welding is an inevitable trend of industrial transformation in enterprises.At present,most of the robots work in the "teaching-reproduction" mode,and they have no ability to adapt to the assembly deviation of the workpiece or the deformation during the welding process.Therefore,the seam correction technology came into being,and the seam correction still needs manual teaching in advance.Therefore,the fully automatic seam tracking technology that can independently plan the welding trajectory is the future development direction.This paper designs a passive seam tracking system,which can automatically locate the welding starting point,dynamically plan the welding trajectory during the welding process,and automatically locate the welding end point.In the hardware design stage,based on the position-based visual servo control principle,this paper builds a welding seam tracking system based on passive visual welding seam tracking sensors,which includes industrial cameras,six-axis industrial robots,and host computers.In the system calibration stage,Matlab was used to calibrate the camera,and the open source calibration software easy-handeye was used for hand-eye calibration,and the accuracy of the system calibration was tested in the experimental stage.In the process of weld image processing,this paper proposes the method of inter-frame matching denoising to remove spatter,which is the key common problem of spatter interference in weld image.It uses ORB(Oriented FAST and Rotated BRIEF)and RANSAC(Random Sample Consensus)Combining the transformation matrix between the images,transforming the first frame image to be aligned with the second frame image and then performing pixel operation,can remove a lot of splash noise without losing details.Aiming at other line interferences that are easy to appear in fillet welding,this paperproposes to use a gradient operator-based image edge detection method,which can effectively remove the line interference of a specific gray gradient.It compares the three commonly used welding seam fitting methods and evaluates them in terms of main stability.In order to find the starting point and end point of the welding seam,this paper adopts the filtering method based on the direction convolution to collect the welding seam and the starting line and the end line of the welding line respectively.Among them,the dynamic ROI(Region of Interest)method is introduced in the identification of the welding end point In order to accurately identify the starting point and position of the weld seam in the case of more interference.In the trajectory planning stage of welding seam tracking,the kinematics model of the robot is established in this paper,which includes three aspects: robot DH parameters,robot forward kinematics,and inverse kinematics.For the attitude of the robot during the laser welding process,geometry-based Kinematics inverse solution algorithm,and the results of inverse solutions are screened,which effectively improves the reliability of inverse solutions.Aiming at the requirements of the welding trajectory in the laser welding process,trajectory planning is performed in the Cartesian space.In order to ensure that the laser welding head maintains a constant speed welding at all times during the welding process,the acceleration and deceleration control of the S curve is used to accelerate the welding start and end points In the planning of the area and deceleration area,it is proposed that the acceleration area needs to be reserved before reaching the welding starting point,and the segmented trajectory planning method to increase the acceleration area at the welding end point.It is proposed to use cubic polynomial spline interpolation for trajectory planning of the weld path,which can ensure that the trajectory is continuous at both the speed and acceleration of the line segment connection.In the design and experiment stage of the welding seam tracking system software,the currently popular open source robot operating system(ROS)is used as the development platform,and the robot model under the ROS system is established in conjunction with Solid Works and Rviz,and the visualization function verification of Rviz is adopted Thecorrectness of the robot model.Adopt MOVEIT under ROS operating system! Perform kinematics calculation and assist in trajectory planning,and introduce the software design methods of image acquisition module,image processing module and motion control module in detail.Finally,this paper tests the accuracy of the welding seam tracking system through the welding seam tracking experiment,and analyzes the causes of the errors.