Self-planning And Online Correction Of Robot Multi-pass Welding Based On Laser Visual Information

Welding workpieces with medium thick plates are widely used in marine equipment,construction machinery and bridge structural parts.Compared with thin plate welding,the medium thick welding has longer groove,and the welding process is more complex.At present,the most welding robots are "teach-play" mode,they are unable to modify welding parameters of welding bead according to the actual situation of groove,and unable to determine the initial point of welding gun for each welding bead independently.In order to ensure the quality of medium thick plate welding and improve the production efficiency of plate welding,it is urgent to explore an intelligent method for medium thick plate welding.Aiming at the problem of plannig for multi-pass welding,using laser vision sensor to collect each weld groove profile after welding.Two feature points of weld groove are obtained by image processing and feature extraction algorithm,and convert they to 3d coordinate through the coordinate transformation.After that,we can use the feature points and the relation between the welding parameters of weld width and parameters to modify the welding parameters.At the same time,we can determine the starting torch point for next bead according to the feature point information.Firstly,we can get the relationship between the welding current of backing welding and the forming size of v-groove through the welding process test.At the same time,the influence of welding current and swing amplitude on the forming size of welding bead in the swing welding of medium and thick plate is analyzed by the classification of swing welding bead according to the constraint conditions.Then,the laser vision sensor is used to collect the profile of each groove after welding.The center line of laser stripe is extracted by binarization image,median filter and mean method,and the first feature point is extracted by using the feature at the lowest welding seam.Next,the second feature point is extracted by using the relationship between the second feature point and the first feature point.The corresponding coordinate transformation matrix is obtained through camera calibration,laser plane equation solving,and robot hand-eye calibration.And the two-dimensional image coordinate points are converted into three-dimensional base coordinate points of the robot.Finally,the width information of the weld bead to be filled is obtained by using the feature points information,and the corresponding welding parameters are calculated by using the relationship between the welding parameters and the width of the weld bead.And the pre-planned welding parameters are modified with the parameter value.At the same time,the starting point of each welding bead can be determined by using the feature points information.In order to verify the feasibility of the above method,taking medium thick plate of v-groove as the test object,pre-planning of the weld bead based on customized filling scheme,the welding parameters are modified by the above method,and the starting point position of the calculated welding gun is compared with the actual position of the welding gun to verify its accuracy.The results show that the welding parameters of the weld bead can be effectively corrected by this method,and the position of the welding torch of each bead can be accurately determined,and the error is within 1.2 mm.