Research And Application On Methods Of Welding Path Recognition And Autonomous Control In Welding Robot

Welding technology is one of the important processing technology in manufacturing industry.In the welding manufacture field,it is an inevitable trend to realize automation,flexibility and intellectualization of welding process.The intelligence research of the welding robot will become the key to welding intelligence technology.The dissertation mainly concerns methods of welding seam path recognition and autonomous control of welding robot based on visual system.Moreover,some experimental platforms are designed to validate the feasibility of the proposed methods.Detailed jobs are as follows:A novel welding sequential pattern recognition algorithm is proposed to measure the depth and shape of seam during V-shaped multilayer weld for rotating workpiece.Firstly,the weld image is preprocessed to reduce the computation burden and disturbance.Then the sequential method is proposed to extract and recognize weld seam feature angular points.Furthermore,the feature angular points' mean in several previous seam image frames can be employed to correct the location of the feature angular point in the current frame.Finally,the weld position,width and the slope information can be recognized by depending on the furthermore analysis of these extracted feature angular points.An accurate and flexible narrow seam visual tracking control method for welding robot is proposed to realize the real-time perception and online visual control for the narrow seam.Firstly,a weld image processing scheme is presented to reduce the strong disturbances such as arc light,spatters,etc..Then the reliable extraction of narrow seam features and communications transmission between image processing computer and controller is completed.Secondly,an image-based closed-loop control method is presented.Then the accurate narrow seam tracking of the welding robot is realized under guide of the vision system.Finally,some experimental results show that the visual control method and system is robust to some disturbances,such as arc,smoke,splash and lighting change.A visual predictive control method for manipulator is proposed.The object image is captured,and some feature is extracted.Then the relation between velocity of image feature and robotic motion velocity can be obtained.Furthermore,some visual control methods for manipulator are designed by employing predictive control frame to deal with multiple constraints.Finally the desired object tracking is realized,and some simulation results are satisfactory.The robotic trajectory tracking control method based on reinforcement learning is proposed.Firstly,the basic principle of reinforcement learning is presented.Then the robot trajectory tracking control strategy based on SARSA is proposed.Finally,an intelligent controlling system is designed and desired results are gained.