Reconstruction Of Multi-layer And Multi-pass Weld Contour And Analysis Of Weld Deformation

Welding technology plays an important role in the industrial manufacturing field,with a welding proportion of approximately 40%in the preparation of mining hydraulic supports.Currently,most structural components of mining hydraulic supports are of medium-thick plates which mainly adopt a multi-layer and multi-pass welding method.The angular deformation caused by each layer and pass of the weld seam is crucial to the safety and service performance of the medium-thick plate structural components.At present,manual measurement is the main method used for checking angular deformation,however,it is slow and subject to subjective factors.Therefore,the application of line laser profile cameras to measure welding deformation is of great significance.The mature development of machine vision technology enables non-contact measurement methods to reconstruct and analyze weld seam joints,thus more accurately describing the surface forming quality of the weld seam.The adoption of this method is of great significance for studying the quality inspection of welding forming.Based on thisr this paper takes the multilayer and multi-pass welding of Q460 high-strength steel in the mining hydraulic support as the research object and carries out the following research work:First,integrate a point cloud data acquisition device into the existing welding platform:the existing welding platform consists of a KUKA robot,welding power supply,arc welding gun,and wire feeder.In order to collect corresponding point cloud data during welding.a line laser contour camera needs to be added to the existing welding platform to obtain point cloud data.Using the KUKA robot as the carrier,the line laser contour camera is driven along the welding trajectory to scan and collect corresponding point cloud data.There are some noise points in the point cloud data collected by the camera.Therefore,the original point cloud needs to be filtered to remove noise points,adjust parameters according to different processing scenarios,delete duplicate and unnecessary points during sampling to reduce computational burden and speed up post-processing.Subsequently,the welded angle deformation is calculated based on the filtered point cloud data.The RANSAC plane segmentation method is used to segment the two plane equations of the base material,and a quadratic approximated RANSAC plane segmentation method is proposed to improve the accuracy of solving the plane equation.The angular deformation caused by each weld seam is then calculated using the formula for the angle berween planes in space.This is done to verify the welding passes that do not match the process parameters.The obtained angle values are used for nearby registration.The SAC-IA and ICP fusion registration algorithms are employed to register all point cloud images in the groove area of the weld seam.By comparing the registered cross-sections with the actual distribution of weld seams,the impact of the upper layer weld seams on the lower layer weld seams can be inferred.Secondly,all the scanned point cloud data of welds are reconstructed in 3D.Based on the characteristics of the point cloud data,the MLS and greedy projection triangulation fusion reconstruction methods are used for surface reconstruction of welds,which avoids excessive holes and detail loss that may occur with single reconstruction algorithms and improves the effectiveness of surface reconstruction.Finally,the algorithm is integrated into a graphical operating interface to form software for easy operation and use.The software is also designed to establish a connection with the camera,allowing it to directly read point cloud data and facilitate processing.