Research And Implementation Of Online Detection And Polishing Technology For High Speed Rail Body-in-White Putty

Nowadays,the application of robots polishing based on offline trajectory planning is becoming more and more widely used.However,due to the manual scraping of putty,the high-speed white body randomly generates local high-point defects,and there will still be residuals after offline polishing.Therefore,this paper proposes an online inspection and grinding program: combining the binocular structured light inspection system,the reconstructed surface、trajectory planning and the robot grinding operation to form an automated grinding system for high-point defects.In terms of high point defect detection,a method of normal plane projection screening based on least squares fitting is proposed.By fitting each row of projected point clouds,a reference line is obtained to filter high-point defect point clouds.Then,in order to cluster the scattered point clouds,a grid clustering method based on the Euclid kd tree(K-dimension tree)is proposed.Clustering each defect area according to the convex height and number of the contained defect point cloud to obtain the bounding box data.Finally,the average time for testing the defect extraction of a single point cloud is 1.76 s,which meets the requirement of online detection continuity.In terms of online trajectory planning,according to the cumulative chord length parameter method of the BSpline curve,the boundary surface of the defect bounding box is used to reconstruct the surface.Then the trajectory planning is performed on the reconstructed curved surface according to the equal chord deviation fitting and equal scallop height method to obtain the trajectory information of the surface defects of the polished body.For adaptive grinding,a relative position grinding method based on the vision system is proposed according to the system layout,so that the robot can accurately grind the defects that appear randomly on the surface of the car body.Finally,the robot is controlled remotely through the master control software,and the adaptive grinding process is realized.Based on the Opencascade 3D open source modeling library,the robot simulation software was independently developed to initially verify the operability of the robot grinding system.Finally,the system was tested in Qingdao Sifang: According to the real-time compression of the AOK(Active Orbital Kit),it was concluded that the fitting error of the normal to the body during the grinding process was within 4mm;then the measurement the average roughness of the grinded defect area is 2.722μm,which meets the process requirements.It is verified that the online inspection and polishing system can be used as a defect removal process before the offline full-surface grinding of the robot.