Research On Multi-layer And Multi-channel Laser Vision Welding Seam Tracking Technology Of Arc Welding Robot

During the welding process,thick plate structural parts are affected by the deposition rate,and welding cannot be completed at one time.Multi-pass welding is one of the most common welding methods.At present,this welding process is frequently completed by manual or semi-automatic welding in the practical production,with great labor intensity and unstable welding quality.Therefore,this paper has carried out research work on multi-pass seam tracking technology of arc welding robot based on laser monocular vision,which has great scientific research value and practical significance for ensuring the quality of large-scale steel structural parts welding engineering,improving production efficiency and promoting the automation level of engineering units.The major results of the whole article are as follows:The multi-pass welding test conditions of plate were analyzed,and an overall multi-pass laser vision weld seam tracking test system was constructed.Focus on the design of laser vision sensor components,key parameters such as the laser oblique,CCD camera vertical receiving sensing method;the relative angle is 27°;the distance between the sensor and the torch is 80mm;the distance between the sensor and the workpiece is 160mm was determined.And the key components of the vision sensing system and the filter shading system in the laser vision sensor were selected.By analyzing the characteristics of multi-pass welding image noise,a laser stripe image processing method was developed.The median filtering method was used to perform preliminary denoising on the welded image,and the noise such as arc and smoke in the image was filtered out.The ROI of the binarized image was extracted.The noise caused by the welding spatter on the periphery of the image was effectively removed,the image processing time was shortened,and the real-time performance of weld seam tracking was improved.Morphological dilation was used to obtain a laser fringe curve with smooth edges and better connectivity.The thinning algorithm is used to extract the center line of the laser stripe,and an improved feature point recognition method based on distance filtering was proposed to process the multi-pass laser stripe feature points.Combined with the calibration of the laser vision sensor,the three dimensional localization of the weld feature information is realized.After the experimental verification,the maximum tracking error at the edge and bottom of the groove is no more than 0.5mm.The guidance of the weld initial point based on laser vision was achieved.Using the position information of the weld feature points obtained by image processing,a real-time adjustment algorithm for the pose of the touch is proposed to carry out the trajectory planning of the torch.On the basis of achieving multi-pass welding location,the welding process was improved.Combined with the Visual Studio and the interface module built in ABB robot,the data communication between the robot and IPC was realized.The multi-pass seam tracking test of plate was carried out to further verify the feasibility of the algorithm.And it could satisfy the needs of practical welding.