Research On Calibration Algorithms Of Welding Robot Based On 3D Visual Guiding

As the result of the thermal deformation and clamping error of the weldment in the robot welding procedure,the teaching welding trace is different with the actual welding trace,which makes the tradition teaching welding robot unable to match the need of high-precision welding.In recent years,the welding robot based on 3D visual guiding realizes the precise control of the welding-gun which is by using the localization method based on laser sensor,and it is an effective way to improve welding quality.However,the requirement of high-precision welding demand improving the calibration accuracy of the welding-gun,laser sensor and the transform relationship between them.To this end,the calibration in the welding process is explored in depth in this paper,and relevant calibration algorithms which can improve the calibration accuracy of the whole welding robot based on 3D visual guiding are proposed.In order to solve the transformation between the welding-gun coordinate frame and flange coordinate frame of the welding robot,the principle and the reasons for errors of the tool coordinate frame calibration are analyzed and explained in detail in this paper.To reduce the kinematic parameter errors which has the biggest influence on the welding robot errors,a new tool coordinate frame calibration method of parameter compensation which is based on the kinematic parameter calibration is proposed.The method combine the tool coordinate frame calibration with the kinematic parameter calibration,so the calibration will be more simplicity and precise.So as to deal with the problem of singular matrix in the calibration process,elastic network regression is used in parameter identification.Finally,the calibration method improves the absolute accuracy of the tool center point of the welding robot.On the basis of above-mentioned researches the transformation between line structured light sensor and welding robot is studied,and the hand-eye calibration algorithm is implemented.Firstly,the transformation between pixel coordinate frame and sensor coordinate frame is learned by analyzing the imaging principle of the line structured light sensor.Then,the hand-eye calibration model is built according to the traditional "eye-in-hand" algorithm.This model can make the two-dimensional pixel coordinates in the sensor into the three-dimensional in robot coordinate frame.In order to solve the problem of hand-eye matrix accuracy reduction caused by robot measurement error in the calibration process,a method which can improve the calibration model by adding the Euler Angle into the model is proposed to reduce the influence of the measurement error.Finally,the transformation matrix in the hand-eye calibration is solved by the Newton-Raphson method.For the sake of verifying the validity of the proposed calibration methods,the robot simulation platform is built,and the experiment accuracy is verified in the simulation experiment.And there two calibration methods are applied in the real welding robot experiment platform.Though the error of the robot used in experiment is far larger than the welding robot in the actual environment,the ball radius formed by the tool center point meets the industrial demand of 3mm-5mm.The result of the hand-eye calibration isn't ideal due to the large initial error,but it can basically meet the needs in practical use.The experimental results show that the above calibration methods can meet the accuracy requirements of the actual welding robot.