Research On Weld Seam Identification And Position Of Weld Robot System Based On Structured Light Vision Sensor

The traditional welding mode of welding robot is teaching programming or pre-programming according to the position and shape of welding workpiece,which needs a lot of manpower and time,and is suitable for standardized and small-scale welding environment.However,in the non-standard and large-scale complex welding environment,the disadvantages of this mode are low efficiency and low flexibility,which are often not applicable.Welding robot based on structured light vision sensing has been widely used in welding manufacturing field because of its advantages of high precision,non-contact,high flexibility and high efficiency.In the welding robot system based on structured light vision sensing,the accurate positioning and identification of welding seam and the reasonable trajectory planning are the important guarantee of welding quality.Therefore,this paper focuses on the welding seam type identification,V-type weld position extraction and V-type weld trajectory planning methods.The main contents are as follows:(1)The vision sensing of structural light for welding robot is studied.The imaging model based on structured light vision sensor is analyzed,including camera internal parameter model,triangulation principle and hand eye model;considering the actual welding site conditions,the structured light vision sensor is designed;the welding process is segmented,and the welding process and its vision sensor tasks are discussed.(2)The method of weld type recognition in complex welding environment is studied.Based on the analysis of the image features of each weld type,it is proposed that the deformation information of laser at the weld is used as the identification information of weld type,the slope distribution of laser curve and the interval of feature points of laser curve are extracted as feature vectors,and SVM(support vector machine)is used)At the same time,the optimal parameters are obtained by cross validation,and the method of weld type identification of welding workpiece in complex welding environment is established.The welding robot system can accurately identify the weld type before welding,and then select the corresponding welding parameters and weld feature points extraction algorithm according to the weld type.(3)The method of feature point extraction of V-shaped weld is studied.According to the characteristics of arc light,spatter and even laser submergence in the welding process of medium and thick plate,the welding process is divided into two stages: initial stage and welding stage.In the initial position stage,the traditional image denoising and ROI extraction from the connected region are used to extract the weld;in the welding process stage,the yolov3 tiny target detection method is used to extract the weld According to the position of the laser stripe in the seam,ROI is set in the source image to process the local image,and then the position of the weld feature point is determined by the geometric relationship.(4)The welding path planning method of V-shaped weld was studied.According to the characteristics of large weld spacing of medium and thick plates,the movement modes such as no gun and zigzag gun are designed;according to the characteristics of thick plates,the multi-layer welding planning is designed;and the corresponding position planning and interpolation calculation are carried out for different welding processes.(5)The experiment platform is designed and built,and the experiments of weld type identification,weld tracking and weld flaw detection are carried out.Five types of weld were identified by the experiments of multi type weld identification and noise verification.The time of weld identification was 148.24 MS,the recognition accuracy is more than 98.4%,which verifies the validity and robustness of the proposed method;the V-shaped seam tracking experiment under the actual welding environment is carried out,the results show that the maximum error of the welding robot system in the x-axis is 0.691 mm,and the maximum error in the z-axis direction is 1.296 mm,which verifies the validity and track of the proposed method of feature point extraction The rationality of the trace planning method and the accuracy of the weld positioning are discussed.The results of the radiographic test on the welding results of the workpiece show that there are not more than one porosity defect per 800 mm of the welding seam of the workpiece after welding,and there are no other defects,which meet the quality requirements of the international level II welding seam.It shows that the welding quality of the welding robot system with structured light vision sensing is good The system has the application prospect.