Research On V-Shaped Welding System Based On Point Cloud Feature Extraction

At present,the welding industry generally uses manual welding or robotic teaching welding.Manual welding is harmful to the human body and the quality of welding is difficult to guarantee.There is a problem of the teaching track deviation in the robot teaching welding,and the welding position of different products is different,and the teaching operation is cumbersome,and it is difficult to complete the task of mass product welding.In this paper,the point cloud feature extraction technology is used to study the robot automatic welding system.The main work includes:1、The hardware selection and experimental platform construction of the V-weld welding system based on point cloud feature extraction were completed.The overall scheme of the system was designed,and the laser sensor was combined with the translational motion experimental platform to realize the collection of point cloud data on the weld surface.2、Robot modeling and system calibration were completed.According to the DH parameters,the kinematic model of the FYB6 six-axis robot used in this paper is established,and the kinematics positive and inverse solutions are calculated and the results are verified.The calibration of the tool coordinate system and the calibration of the hand eye were completed,and the calibration experiment was designed,and the calibration error was less than 0.5 mm.3、Aiming at the weld point cloud of this paper,the feature points based on the inter-point normal vector variation algorithm and the tangent plane linear algorithm are designed to extract the feature points.The former uses the covariance analysis method to calculate the point cloud unit normal vector,and then calculates the inner product of the unit point normal vector of the data point and the unit normal vector of the neighborhood point.If the absolute value of the inner product is less than the set threshold,the greater the degree of bending near the point,the more the point is determined to be the feature point.The latter uses the minimum bounding box algorithm to determine the approximate shape of the point cloud.A linear plane is inserted based on the resulting bounding box coordinates.The Hough transform is used to fit the points on the tangent plane to complete the extraction of the weld feature points.4、The path planning of the robot was completed and the welding experiment was designed.The extracted weld feature points are subjected to quadratic polynomial fitting by spatial curve fitting,and the curve equation is obtained and discretized to obtain the three-dimensional coordinates of the weld path.Through the inverse kinematics of the robot,according to the three-dimensional coordinates of the path and the set position of the welding torch,the position of each joint when the welding gun reaches the point on the trajectory is obtained.The motion trajectory of the robot is simulated by simulation.And by designing the welding experiment,the average error between the calculated feature points and the actual feature points is 0.8mm,which meets the requirements of welding technical indicators.