Research On Key Technology Of Intelligent Trajectory Planning For Robot Welding Based On 3D Vision

There are still many industries(e.g.,mining equipment,shipbuilding)where automatic robot welding is difficult to implement due to factors such as laying-off cutting errors,assemblage and assembly errors and clamping errors.In addition,small and medium-sized enterprises(SMEs)or enterprises with small batch product welding needs are in urgent need of welding robots with more intelligence and automation.Therefore,the study of intelligent trajectory planning key technology of robot welding based on 3D vision can improve the level of robot welding automation in related fields in China and expand the application scenarios of welding robots.In this paper,the key technologies of intelligent trajectory planning for welding robot system equipped with 3D structured light vision sensor are studied with the workpiece which has complex welding trajectory as the object:the trajectory planning technology of complex welding trajectory workpiece and the actual welding trajectory deviation detection technology.According to whether the 3D model of the workpiece to be welded is known or not,the trajectory planning technology of complex welding trajectory workpiece is divided into the intelligent planning of complex welding trajectory based on structural features in the offline environment and the welding seam detection and welding trajectory planning based on 3D vision.The actual welding trajectory deviation detection technology contains 3D vision-based workpiece pose detection and weld seam capturing pose planning and 3D vision-based weld seam detection.First,an intelligent welding trajectory planning method incorporating non-rigid registration of the point cloud and offline processing of the 3D model is proposed to realize welding trajectory planning with known 3D model of the workpiece.For the workpiece with complex welding trajectory,a simplified 3D model with the same structure is established,and a structural feature point set containing structural feature information(edge feature point cloud)and welding trajectory planning information(welding trajectory point cloud and welding auxiliary trajectory point cloud)is constructed.The 3D model of the workpiece to be welded is converted into an edge feature point cloud,which is non-rigidly aligned with the edge feature point cloud and weld trajectory point cloud in the structural feature point set to obtain the weld trajectory point cloud of the workpiece to be welded.Then,according to the correspondence between the point cloud of the workpiece to be welded and the 3D model,the edge representing the welding trajectory in the 3D model is acquired.As a result,the 3D model of the workpiece to be welded is processed to construct a circular pipe surface with the axis of the weld trajectory and to find the intersection line with the neighboring sides of the weld trajectory to complete the planning of the weld trajectory.Secondly,a welding seam capturing posture planning method for visual sensors in offline environment is proposed to guide the visual sensors to optimally capture the weld seam area in response to the welding trajectory deviation brought about by the unloading and welding grouping.Based on the known 3D model of the workpiece and the theoretical welding trajectory planning,a mathematical model of the optimal capture posture of the weld seam is established based on the vision sensor’s field of view model.The welding trajectory is discretized,and the dynamic planning concept is applied to realize the optimal capturing posture planning of the weld seam point.The characteristics of the point cloud obtained by the 3D vision sensor are analyzed,and an improved nearest point iteration(lCP)algorithm is adopted to realize the detection of workpiece clamping position.Finally,a 3D vision-based weld seam detection and welding trajectory planning method is proposed.By analyzing the camera capturing principle and adopting a high grayscale expectation value method to enhance the weld seam features in the image,an image edge chain method is proposed to realize partial weld seam detection in the image.Combining the image and its corresponding depth map,a point cloud edge detection algorithm is proposed to realize complete weld seam detection,and the point cloud of the obtained welding trajectory is fitted with a spline curve.A geometric projection method is proposed to establish the current weld point coordinate system,and then the welding trajectory planning is realized.Combined with the above research content respectively established whether the known workpiece 3D model of the robot automatic welding process,especially in the unknown 3D model,build the structural feature point set in the online situation,to realize the detection of the welding trajectory of the workpiece to be welded.The robot welding experiment platform equipped with 3D structured light vision sensor carries out actual welding experiments on the impeller workpiece with known 3D model and C-type fillet weld workpiece and S-type fillet weld workpiece with unknown 3D model respectively.The final welding experiment results verify the effectiveness of the welding trajectory intelligent planning technology and the actual welding trajectory deviation detection technology.