| Now the welding robot is mainly used artificial "Teach-Reproduction" of the welding mode,with the factors thermal deformation of the workpiece and the clamping errors of the workpiece to be welded,resulting in larger error between the actual and theoretical track weld trajectory.Pre-teaching of the welding trajectory could not meet the actual welding requirements,so the actual weld trajectory on-line identification was very necessary.To this end,this paper presented a line laser detection technology,and installed a visual sensor at the end of the welding torch of the welding robot.Based on the Gaussian kernel-related line laser weld seam tracking method,the pixel coordinate values of the feature points of the weld were extracted in real time.The deviation amount of the coordinate value was sent to the robot through the adaptive fuzzy controller to guide the robot to complete the automatic welding process.This paper studied the hand-eye calibration method between laser vision sensor and welding robot.First,the calibration of the internal and external parameters of the industrial camera was completed,and then the plane equation of the line laser plane relative to the industrial camera coordinate system was established to complete the conversion from the pixel coordinates to the laser plane coordinates.Finally,aiming at the "Eye-in-Hand" system of vision sensor and welding robot,a hand-eye calibration method was proposed,and the parameters of homogeneous transformation matrix were obtained,thus completing the conversion from the two-dimensional pixel coordinate value to the robot three-dimensional coordinate value.In this paper,the target tracking method of weld feature points was researched.Before the start of welding,the images were collected by visual sensors for morphological processing to obtain an initial coordinate values of the target area and the feature point feature point where the weld.After the welding starts,the target tracking method was used to track and extract the feature points of the weld,and a classifier was used to classify the images.For the large noise image collected after the start of welding,the Gaussian kernel correlation algorithm was used to calculate the similarity between the candidate region and the target region.The region withthe highest similarity was selected as the target area,and a series of coordinate values of the feature points were obtained.This paper studied the intelligent control algorithm of robot displacement.By subtracting the obtained coordinate value of the weld feature point from the reference coordinate value,the displacement amount of the welding robot needed to be obtained.However,since the mathematical model of the welding robot was unknown,the displacement could not be sent directly to the robot.For this reason,this paper designed an adaptive fuzzy control algorithm,which used multi-input single-output fuzzy controller,the deviation value and the deviation rate were input to the fuzzy controller,and a series of smoothed robot bias voltages were output.The obtained bias voltage was sent to the controller of the welding robot to guide the robot to complete the automatic welding process.This paper designed and built a six-degree-of-freedom robot automatic tracking welding platform.The hardware system included Yasukawa 6-DOF welding robot and laser vision sensor.The control system included Yaskawa DX200 controller,A/D conversion module and the host computer software module run in Windows system.According to the difference of the offset distance,the platform has carried on the tracking welding experiment of six sets of weld characteristic points for three different types of curve welds.The experimental results showed that the system can realize the real-time automatic tracking welding of the workpiece to be welded under the interference of strong arc and splash noise.The tracking error is less than 0.32 mm,the measurement frequency of the sensor is 20 Hz,which can meet the requirements of actual welding situation. |
PDF Full Text Request