Realtime Routing Planning And Welding Seam Location For Multi-layer Welding Based On Machine Vision

Medium and thick plate welding workpieces are usually filled with multi-layer and multi-channel welding methods.Before welding,the initial point of the weld is generally positioned by manual teaching,and then multi-layer multi-channel welding is filled by manual preaching.In order to further reduce the workload of human intervention of the robot in the working process and improve the welding quality of the welding robot,based on the machine vision method,this paper conducts research on the initial positioning of the weld seam of the medium and thick plate welding and the dynamic drainage planning of multi-layer multi-channel welding.The main results of the full text are as follows:(1)A set of robot weld initial point positioning and multi-layer multichannel dynamic row welding experimental system were built.The mobile welding robot system is composed of Aobo robot,visual sensor,Autai welding machine and other related equipment,and the verification interface and communication program between each equipment are designed to facilitate the initial point positioning of the weld and multilayer multi-channel dynamic row welding experiments.(2)Based on the image convolutional neural network and the visual processing method of machine vision,the automatic and precise positioning of the initial point of the weld is realized.Using the convolutional neural network algorithm,a deep learning model for identifying the weld area is trained,and then by analyzing the structured light imaging of the V-shaped groove,an image processing algorithm for the imaging feature points of the structured light weld groove is given to realize For the purpose of positioning the initial point of the weld groove.(3)The visual processing algorithm of weld bead detection is written to realize multi-layer and multi-channel welding dynamic row.The relationship between the cross-sectional area of the weld bead and the welding current and other welding parameters is obtained through welding experiments,and the characteristic point identification algorithm of the weld contour of the filled layer is given in the multi-layer multi-channel welding process,and the remaining width and height information of the weld groove are calculated online during the welding process,so as to realize the multi-layer multi-channel dynamic row welding of the medium and thick plate workpiece.(4)In order to verify the effectiveness of the dynamic row and initial point positioning algorithm,according to the experimental platform built,the multi-layer multi-channel dynamic row and the initial point positioning test of the weld seam were carried out.The test results show that the filling of the medium and thick plate weld beads obtained by the dynamic row test is reasonable,and there are no welding defects such as porosity,unsolder penetration,slag clamping,and undercut,and the positioning accuracy of the initial point of the weld seam reaches 0.2mm,which meets the requirements of the initial point positioning of the robot welding.