Weld Seam Profile Detection And Automatic Tracking System For Arc Welding Robot

As a fusion process,welding is widely used in industrial production,but at present,manual welding has low efficiency and high cost.Teaching-type welding robots(TWR)cannot have self-feedback correction deviation,and it is necessary to continuously teach the welding path in non-standard parts,adjust the welding accuracy.Besides,the TWR is only suitable for welding large-production standard parts,single welding type,etc.Based on the use of visual sensors instead of teaching to transmit weld path information,this paper studies an arc welding robot weld trajectory detection and automatic locating system,aiming at effectively reducing the welding deviation of robot,improving welding precision and enhancing ability of automation for welding robot to adapt to all kinds of non-standard parts independent welding.This article works as follows:(1)According to the detection requirements,the construction of the initial weld seam location,weld detection and tracking hardware system based on line structure light vision was completed.Firstly,the hardware selection of CCD camera,lens,line structure light and distance laser sensor is completed.Then the camera and manipulator end connection device is designed.The sliding table structure is designed on this device,and the sliding table motor signal is connected to the manipulator controller I/O to automatically adjust the working distance of the camera.(2)The forward and inverse kinematics of the Dobot four-axis manipulator are discussed.The forward kinematics equation for solving the attitude of the end coordinate system is established,and the joint coordinate value is inversely solved according to the attitude matrix of the path coordinate point to control the robot motion to the specified the target position,laid the foundation for the completion of hand-eye calibration and visual tracking.On the other hand,carbon steel and austenitic stainless steel based on the larger physical properties difference are used in engineering,the thermodynamic analysis of austenitic stainless steel welding is carried out,and the mathematical model of heat source is created,which provides a theoretical basis for the preparation of welding subprogram.(3)This paper utilize the Zhang Zhengyou algorithm in camera calibration toolbox of Matlab and the geometric camera calibration method based on Halcon circular control points to calibrate the camera parameters,and compares the calibration accuracy between the two calibration methods.Based on the hand-eye model,the relative position relationship between the robot and the camera is calibrated result in that the transition process from the camera coordinate system to the robot coordinate system is realized and the high calibration accuracy is obtained.(4)The traditional Canny method is improved.The improved method based on gradient amplitude histogram distribution statistics of image and intra-class feature minimization is used to determine the double threshold to realize the adaptive structure light stripe detection.Compared with the traditional Canny operator fixed threshold segmentation,the method has strong adaptive detection capability resulted from that the Gaussian filter of the traditional Canny operator is replaced by a nonlinear anisotropic filter,which ensures that the edge pixels are not lost while denoising.The window image method is used to determine the interesting area of the weld image,reduces the range of image processing,speed up the efficiency of image processing and reduce the impact of overall noise.What's more,the use of weld detection from coarse positioning to precise positioning greatly improves the detection accuracy.(5)A weld seam detection and tracking software system was developed under the weld tracking requirements.Combinating with the Halcon13 image library,the Winform interface and function blocks are developed on the VS2010 software platform.The function blocks include camera control,dobot robot control,hand-eye calibration,weld image processing,saving images,and weld data.Taking account to system calibration error and image pixel error,the real-time measurement portion of weld seam should process more than 10 frames per second with the positioning deviation is within 0.2mm.In addition,in the case of local environmental stability of the weld and the position of the welded joint can be arbitrarily placed,the weld can be detected normally,and the illumination change has a certain degree of fault tolerance.What's more,compared with the TWR,the proposed method has higher tracking accuracy and strong weld seam detection robustness.