Research On Positioning Navigation And Control Technology Of Spherical Tank Inspection Robot

With the increasing number of large-scale special equipment in China,the demand for automatic inspection and maintenance equipment of large-scale petrochemical storage spherical tanks is also very urgent.Due to the high-altitude operation,the internal and external environment of spherical tanks is complex,and the inspection robots with high automation and strong adaptability have become an important direction of research and breakthrough.The traditional manual operation method requires the construction and support of external or internal scaffolding,manual measurement and inspection.The problems of long cycle,low efficiency and high cost have not been solved.At present,wall-climbing robots have been developed at home and abroad,which are mainly used in relatively flat walls,and most of them are concentrated in robot design,lacking the ability of automatic operation in industrial sites.Aiming at the curved wall and high-altitude environment of spherical tanks,this paper focuses on the mechanism design and analysis,weld identification and extraction,path tracking and control,spatial positioning and path planning,experimental verification and field testing of the spherical tank inspection robot.We have developed an intelligent inspection robot system suitable for spherical tanks.The main research contents include:(1)Mechanical design and analysis of the developed spherical tank inspection robot.First,according to the application environment of automatic inspection of spherical tanks,the performance and functional requirements of the robot were analyzed.The magnetic adsorption drive mechanisms,modular magnetic wheels,and adaptive shock-absorbing suspensions were designed,which significantly improves the adaptability and adsorption stability of the robot on curved walls.Then,based on the operation requirements,the overall mechanical design of the robot was completed.Additional inspection and maintenance mechanisms that could be quickly assembled were developed,such as grinding mechanisms,flaw detection mechanisms,cleaning mechanisms,and painting mechanisms.Finally,the robot forces and obstacle-surmounting process were analyzed,and the driving force of the robot are calculated.(2)Environmental perception and weld seam identification of spherical tanks by the robot.Weld seam images were collected and labeled,and the weld seam dataset was also expanded through data enhancement methods.Subsequently,a weld seam identification model based on Mask R-CNN was proposed.By training the weld seam dataset,the robot can achieve fast and accurate instance segmentation of weld images and perform pixel-level weld identification.Some image processing methods and the least square method were used to fit and calculate the deviation of the weld paths.By combining the deep learning networks,image processing,and path fitting algorithm,high-precision weld identification and path fitting were realized.(3)Precise weld path tracking by the inspection robot.The kinematic model of the robot and weld path tracking error were analyzed,the weld seam tracking coordinate system was established,the initial and expected state vector of the robot were analyzed,and the state error vector of the robot was obtained.Based on the analysis of the weld path tracking error,two weld path tracking controllers based on PID and sliding mode control were proposed to realize the robot tracking weld paths.A robot state feedback and control subsystem was established for real-time simulation and testing.Meanwhile,the simulation and analysis of the two designed weld path tracking controllers were carried out.(4)Spatial positioning and optimal path planning of the robot.The robot space positioning and navigation system was developed.The positioning sensors carried by the robot included a UWB positioning tag,two encoders,an inertial measurement unit and a camera.In order to prevent inaccurate data caused by interference of a single sensor,a robot spatial positioning method based on multi-sensor data fusion was adopted.Based on the Kalman filter,the multi-sensor positioning data were fused and updated to realize the estimation and prediction of the position and velocities of the robot in the space coordinate system.In order to realize the optimal weld path planning on the surface of spherical tanks,some 2D weld map of spherical tanks were constructed.The shortest repeated edges were added to the weld maps by graphical method based on an odd-even-point approach,a Euler circuit was solved based on the improved Fleury algorithm,so as to realize the optimal path planning of the robot in all weld maps.Virtual3D tank maps and software platform were constructed and developed to realize the real-time display of the robot’s spatial position and running route,and to feedback the robot status.(5)Robot experiments and field tests.In order to verify the performance and stability of the robotic system,a 4-m diameter cylindrical tank experimental platform was built.Through the experiments on the tank experimental platform,the payload ability,obstacle-surmouting performance and stability of the robot were verified.Weld seam recognition and path tracking experiments were performed to verify weld seam recognition accuracy and path tracking error of the robot.The spatial positioning and path planning experiments were also performed to verify the spatial positioning accuracy and path planning capabilities of the robotic system.Finally,field tests were carried out in a 3000-m~3 industrial spherical tank to verify the practicability and efficiency of the robot in weld seam tracking,spatial positioning and path planning in engineering applications.The experiments and field test of the robot provide the application basis for the automatic operation of the inspection robot on the large-scale special spherical tanks.