Point Cloud Based Free Form Surface Reconstruction And Grinding Path Planning For Removal Of Excessive Weldment

Robot grinding has gradually replaced manual grinding due to its strong adaptability,high degree of freedom,and flexible processing.At present,robot grinding path planning are based on manual teaching and offline programming.The path generated by these methods are mainly based on theoretical models.Due to the difference between the theoretical model and the actual model,the path cannot consider slight differences in the shape of the same batch of parts.This article aims at the path planning problem on complex curved surfaces in order to remove excessive weldment.By point cloud filtering,segmenting,and surface reconstruction,the weld zone is identified the path is planned automatically.First,an automatic welding seam grinding system is built.The system includes robot system,grinding system,data acquisition system and data analysis system.Communication interfaces between different systems are developed to achieve system integration and synchronous acquisition of point cloud data.In order to determine the coordinate relationship between the scanner and robot,a hand-eye calibration of the robot and the scanner was carried out to obtain the rotation matrix and translation vector.Point cloud data acquisition and efficient processing algorithms of the tank welds are studied.A coded octree is used for fast neighbor points retrieval.Statistical filtering and boundary propagation methods are used to remove outliers.The point cloud data is filtered by weighted least squares method.In order to achieve the segmentation of point cloud data,the principal component analysis method is used to calculate the normal vector of the surface.The segmentation line of the point cloud data is identified by Gaussian line detection.For the interior of the segmented area,continuous transformation is used to fill the inner points.Based on continuous transformation,the parametrized coordinates of the filled points are obtained and the point cloud data is simplified.In addition,reconstruction of the weld tank surface is studied.The Bspline model is used to establish the parametrized equations of the weld surface.The control points of the surface equation are solved by the least square method.The node vectors of the equation are optimized by the gradient descent method.In order to improve the accuracy of the model,a self-consistent fitting method is proposed.The self-consistent method alternately solves the control points and optimizes the node vector to reduce the fitting error.Rapid reconstruction of the model is achieved by first extracting the features of weld and then the superpose the centerline of the weld is onto the surface.Finally,based on the results of model reconstruction,a software system for welding seam grinding was developed.The software system can automatically perform point cloud data reading,preprocessing,segmentation,weld feature extraction and surface reconstruction.According to the generated surface model and weld seam feature,a curvature-based trajectory interpolation method is be used to convert the grinding path into a RAPID program that can be directly executed by ABB robots.Based on the experiment,the software system can generate the grinding path automatically and has high application value.