Intelligent maintenance robot for electric tower anti-falling guide rail

Abstract

The invention relates to an intelligent maintenance robot for an electric tower anti-falling guide rail, and solves the problems that in a guide type anti-falling device for overhead transmission line inspection, when the guide rail has a deformation defect, the connection state between two adjacent sections of rails is abnormal, and a bolt group is abnormal, safety accidents are easily caused by manual maintenance; the robot comprises a moving platform, and the moving platform comprises a supporting frame, a driving synchronous belt wheel, a driven synchronous belt wheel, a synchronous belt, a driving motor, a driving roller, a driven roller, a first rear pressing mechanism, a second rear pressing mechanism, a first front pressing mechanism and a second front pressing mechanism. A mechanical arm is connected to the movable platform. The method can be widely applied to detection and maintenance of the rigid guide rail for preventing the electric tower from falling.

Description

Technical Field

[0001] The present invention relates to the technical field of survey and maintenance of overhead high-voltage transmission lines, and in particular to an intelligent maintenance robot for an electric tower anti-falling guide rail. Background Art

[0002] Overhead transmission lines are the main way to transmit large-capacity electricity over long distances in my country. However, overhead transmission lines are generally erected outdoors and are susceptible to extreme weather conditions such as heavy rain, typhoons, lightning, and hail. Under the combined effects of huge alternating tension and bending stress caused by vibration and long-term fatigue caused by rapid changes in temperature, line components such as cables and insulators are prone to mechanical damage, which seriously threatens the safe operation of the line itself and its power transmission and transformation equipment. Regular inspections of overhead transmission lines can prevent the occurrence or deterioration of the above situations.

[0003] At present, the inspection of overhead power transmission lines mainly relies on electric workers to perform high-altitude operations. The biggest danger in high-altitude operations is that the workers accidentally fall to the ground, causing casualties. Therefore, anti-fall devices are of great significance for the safety of workers who climb high. The most basic anti-fall device is a safety rope. One end of the safety rope is wrapped around the waist and back of the tower climber through a lanyard, and the other end is equipped with a lifting ring, which can be buckled on the foot nail or the tower cross arm. Whenever the tower climber climbs up a distance on the foot nail, he needs to bend over to remove the lifting ring and buckle it on the upper foot nail or cross arm until the tower climber reaches the designated position. Although the use of safety ropes can prevent the occurrence of falling hazards to a certain extent, there are still major safety hazards during the period when the lifting ring is removed and reinstalled. In addition, with the rapid development of power grid construction, the height of iron towers is getting higher and higher. The average height of UHV lines has reached more than 70 meters, which means that tower climbers need to spend more time loading and unloading lifting rings, and their work efficiency is significantly reduced.

[0004] In order to further improve safety, a guided anti-fall device is adopted. Referring to the utility model patent with authorization announcement number CN218248181U and the patent application with application publication number CN110331883A, the guided anti-fall device consists of a rigid guide rail, a self-locking device, a connector, a lanyard and a safety rope. The rigid guide rail is fixedly installed on the pole tower. The rigid guide rail is made of several sections of track (the material of the track is aluminum alloy or Q235 steel) spliced ​​and fastened with bolts. The rigid guide rail is an I-shaped guide rail. The safety rope is fixed to the tower climber through a lanyard. The lifting ring at the end of the safety rope is buckled with the self-locking device. The self-locking device is processed with a groove that matches the cross-section of the guide rail. At the same time, a pulley is provided at the position in contact with the end face of the guide rail, so that the self-locking device can be put on the guide rail and move smoothly. When people climb up or down the tower normally, the self-locking device will rise and fall synchronously with the drag of the safety rope; and when there is a risk of falling, the huge instantaneous acceleration will trigger the self-locking structure inside the self-locking device, causing it to quickly lock on the guide rail. It can be seen that the rigid guide rail fall prevention system can not only effectively prevent the risk of falling, but also overcome the problem of reduced operating efficiency caused by the frequent loading and unloading of the lifting ring when relying only on the safety rope.

[0005] However, after long-term use of the guided anti-fall device, as the number of times the self-locking device slides back and forth along the guide rail increases, the guide rail, as the main force-bearing component, is prone to defects such as end face deformation and side deformation, which will cause the self-locking device to be dislocated and unable to slide; at the same time, the bolt group used to connect the two adjacent sections of the track will also loosen under the action of alternating stress, causing dislocation, relative inclination, gaps, etc. between the two sections of the track. Abnormalities such as dislocation, relative inclination, and gaps will seriously affect the normal operation of the self-locking device and threaten the personal safety of the operators. Therefore, it is necessary to regularly check the deformation defects of the guide rail, the state of the bolt group, and the connection state between the two adjacent sections of the track in response to the above problems. If personnel are still arranged to perform manual maintenance, if the anti-fall system already has the above-mentioned hidden dangers at this time, it is likely to cause safety accidents during the maintenance process. Therefore, how to use robots instead of manual maintenance is a technical problem that needs to be solved urgently by technical personnel in this field. Using robots for maintenance can ensure personal safety and improve the efficiency of maintenance operations. Summary of the invention

[0006] The present invention is to solve the technical problem of how to use robots for maintenance in the guide rail anti-fall device used for overhead power transmission line inspection, when the guide rail has deformation defects, the connection status between two adjacent sections of the track is abnormal, or the bolt group is abnormal, which may easily cause safety accidents through manual maintenance. The present invention provides an intelligent maintenance robot for the tower anti-fall rail that can automatically detect and repair the tower instead of manual labor.

[0007] The present invention provides an intelligent maintenance robot for an anti-falling guide rail of an electric tower, comprising a mobile platform, wherein the mobile platform comprises a supporting frame, an active synchronous pulley, a driven synchronous pulley, a synchronous belt, a driving motor, an active roller, a driven roller, a first rear clamping mechanism, a second rear clamping mechanism, a first front clamping mechanism, and a second front clamping mechanism;

[0008] The supporting frame includes a right side plate, a left side plate, a rear vertical plate, a front vertical plate, a top plate, a rear cross plate, a front cross plate, a first rear adjusting screw, a second rear adjusting screw, a first front adjusting screw, a second front adjusting screw, a first rear bearing seat, a second rear bearing seat, a first front bearing seat, and a second front bearing seat. The lower part of the right side plate and the lower part of the left side plate are fixedly connected by three connecting shafts, the rear vertical plate is fixedly connected to the rear end of the right side plate and the rear end of the left side plate, the front vertical plate is fixedly connected to the front end of the right side plate and the front end of the left side plate, the top plate is fixedly connected to the top of the right side plate and the top of the left side plate, and the rear cross plate is fixed to the rear end part of the top of the right side plate and the rear end part of the top of the left side plate The front cross plate is connected, and the front end part of the top of the right side plate and the front end part of the top of the left side plate are fixedly connected; the rear cross plate is provided with two threaded holes, and the first rear adjusting screw and the second rear adjusting screw are respectively connected with the two threaded holes on the rear cross plate, and the first rear adjusting screw and the second rear adjusting screw pass through the rear cross plate; the front cross plate is provided with two threaded holes, and the first front adjusting screw and the second front adjusting screw are respectively connected with the two threaded holes on the front cross plate, and the first front adjusting screw and the second front adjusting screw pass through the front cross plate; the first rear bearing seat is connected with the left side plate, and the second rear bearing seat is connected with the right side plate; the first front bearing seat is connected with the left side plate, and the second front bearing seat is connected with the right side plate;

[0009] An accommodating space is formed between the right side plate, the left side plate, the rear vertical plate, the front vertical plate and the top plate, and the driving motor, the active roller and the driven roller are located in the accommodating space. The right side plate is provided with a motor connecting hole, the driving motor is connected to the motor connecting hole, and the output shaft of the driving motor extends out from the motor connecting hole; the active synchronous pulley is connected to the output shaft of the driving motor; the active roller is provided with a rotating shaft, and the left end and the right end of the rotating shaft of the active roller are respectively connected to the first rear bearing seat and the second rear bearing seat, and the right end of the rotating shaft of the active roller extends outward from the right side plate; the driven synchronous pulley is fixedly connected to the right end of the rotating shaft of the active roller; the synchronous belt is connected between the driven synchronous pulley and the active synchronous pulley; the driven roller is provided with a rotating shaft, and the left end and the right end of the rotating shaft of the driven roller are respectively connected to the first front bearing seat and the second front bearing seat;

[0010] The second rear clamping mechanism includes a roller positioning block, a slider, an inner bearing, an outer bearing, a tension spring, a lower eye screw, an upper eye screw, a guide rail and a roller, the roller positioning block is provided with an inner bearing chamber and an outer bearing chamber, the inner bearing is located in the inner bearing chamber, the outer bearing is located in the outer bearing chamber, the roller is connected to the inner bearing, the roller is connected to the outer bearing, the slider is fixedly connected to the inner side surface of the roller positioning block, the lower eye screw is fixedly connected to the top surface of the roller positioning block, the lower end of the tension spring is hung on the lower eye screw, the upper end of the tension spring is hung on the upper eye screw, and the slider is connected to the guide rail;

[0011] The guide rail of the second rear clamping mechanism is fixedly connected to the rear end of the right side plate, the roller of the second rear clamping mechanism is located below the right side plate, the upper eye screw of the second rear clamping mechanism is connected to the right end of the rear horizontal plate; the lower end of the second rear adjusting screw contacts the top of the roller positioning block;

[0012] The first rear clamping mechanism and the second rear clamping mechanism are arranged symmetrically, and the first rear clamping mechanism includes a roller positioning block, a slider, an inner bearing, an outer bearing, a tension spring, a lower eye screw, an upper eye screw, a guide rail and a roller; the guide rail of the first rear clamping mechanism is fixedly connected to the rear end of the left side plate, the roller of the first rear clamping mechanism is located below the left side plate, the upper eye screw of the first rear clamping mechanism is connected to the left end of the rear cross plate, and the lower end of the first rear adjusting screw contacts the top of the roller positioning block of the first rear clamping mechanism;

[0013] The second front clamping mechanism includes a roller positioning block, a slider, an inner bearing, an outer bearing, a tension spring, a lower eye screw, an upper eye screw, a guide rail and a roller, the guide rail of the second front clamping mechanism is fixedly connected to the front end of the right side plate, the roller of the second front clamping mechanism is located below the right side plate, the upper eye screw of the second front clamping mechanism is connected to the right end of the front cross plate, and the lower end of the second front adjusting screw contacts the top of the roller positioning block of the second front clamping mechanism;

[0014] The first front clamping mechanism and the second front clamping mechanism are arranged symmetrically, and the first front clamping mechanism includes a roller positioning block, a slider, an inner bearing, an outer bearing, a tension spring, a lower lifting eye screw, an upper lifting eye screw, a guide rail and a roller. The guide rail of the first front clamping mechanism is fixedly connected to the front end of the left side plate, and the roller of the first front clamping mechanism is located below the left side plate. The upper lifting eye screw of the first front clamping mechanism is connected to the left end of the front cross plate, and the lower end of the first front adjusting screw contacts the top of the roller positioning block of the first front clamping mechanism.

[0015] Preferably, the mobile platform is connected with a camera for collecting images of the object to be detected.

[0016] Preferably, a sensor for detecting whether the self-locking device is abnormal is connected to the mobile platform.

[0017] Preferably, a sensor for detecting position information of the robot is connected to the mobile platform.

[0018] Preferably, a mechanical arm is connected to the mobile platform.

[0019] Preferably, the power tower anti-fall guide rail intelligent maintenance robot also includes a mechanical arm connecting device, a right mechanical arm and a left mechanical arm; the mechanical arm connecting device includes a mechanical arm supporting crossbeam, a linear slide assembly 1, a linear slide assembly 2, a right screw motor, a right nut, a right nut seat, a right mechanical arm connecting base, a right screw support seat, a left screw motor, a left nut, a left nut seat, a left mechanical arm connecting base and a left screw support seat, the linear slide assembly 1 and the linear slide assembly 2 are connected side by side to the bottom surface of the mechanical arm supporting crossbeam, the linear slide assembly 1 is provided with a first slider and a third slider, the linear slide assembly 2 is provided with a second slider and a fourth slider, the right screw motor is fixedly connected to the right end of the bottom surface of the mechanical arm supporting crossbeam, and the right screw support seat is fixedly connected to the mechanical arm supporting crossbeam In the middle of the bottom surface of the beam, the right nut is connected to the screw of the right screw motor, the end of the screw of the right screw motor is connected to the right screw support seat, the right nut seat is connected to the right nut, the right mechanical arm connecting base is fixedly connected to the right nut seat, and the two sides of the right mechanical arm connecting base are respectively fixedly connected to the first slider and the second slider; the left screw motor is fixedly connected to the left end of the bottom surface of the mechanical arm supporting beam, the left screw support seat is fixedly connected to the middle of the bottom surface of the mechanical arm supporting beam, the left nut is connected to the screw of the left screw motor, the end of the screw of the left screw motor is connected to the left screw support seat, the left nut seat is connected to the left nut, the left mechanical arm connecting base is fixedly connected to the left nut seat, and the two sides of the left mechanical arm connecting base are respectively fixedly connected to the third slider and the fourth slider;

[0020] A boss is connected to the top plate of the support frame, the bottom surface of the robot arm support beam is fixedly connected to the boss, the right robot arm is fixedly connected to the right robot arm connecting base, and the left robot arm is fixedly connected to the left robot arm connecting base; the free end of the right robot arm is connected to end effector 1, and the free end of the left robot arm is connected to end effector 2.

[0021] Preferably, the power tower anti-fall guide rail intelligent maintenance robot also includes an IMU, an encoder and a camera for visual positioning, the IMU is connected to the mobile platform, the encoder is connected to the drive motor or the active roller, and the camera for visual positioning is connected to the front end of the mobile platform;

[0022] The camera used for visual positioning collects images of the guide rail. After preprocessing, the image is used by the edge detection algorithm to identify the gaps at the connection between adjacent tracks in the guide rail, and the number of gaps is obtained. The number of gaps is multiplied by the length of a single track to obtain the absolute position information of the robot.

[0023] The robot posture information obtained by the trajectory calculation of the data collected by the encoder is αb =[x b ,y b ,z b ,θ b ] T , the robot posture information obtained by integrating the data collected by IMU is α i =[x i ,y i ,z i ,θ i ] T , the robot posture information obtained by visual positioning through the camera used for visual positioning is α v =[x v ,y v ,z v ,θ v ] T , then we get x k =[x,y,z,θ] T , where x, y, z, θ are given by α b , α i With α v The robot posture information is obtained after weighted fusion by the Kalman filter algorithm; from this, the prediction model of the robot's state at time k can be obtained:

[0024]

[0025] In formula (1), x k is the state vector x at time k k The predicted value, x k-1 is the state vector x at time k-1 k-1 The predicted value, F k is the state transfer matrix, B k is the control matrix, u k It is the control quantity of external input;

[0026] Similarly, there is an observation model of the system at time k:

[0027] z k =H k x k (2)

[0028] In formula (2), z k is the actual observed value of the system state at time k, H k is the observation matrix, which is used to map the state vector to the observation space;

[0029] At each moment, the Kalman filter first predicts the current state based on the prediction model, and the predicted covariance matrix P k It will also be updated as the state evolves:

[0030]

[0031] Kalman gain K in Kalman filter k The calculation formula is:

[0032]

[0033] In formula (4), R k is the noise covariance matrix of the observations;

[0034] The Kalman gain K is calculated by the neural network k Optimize, the optimized Kalman gain K opt :

[0035]

[0036] The Kalman filter can use the optimized Kalman gain K opt Update the state estimate and output:

[0037]

[0038] Preferably, the free end of the right mechanical arm is connected to a first depth camera, and the free end of the left mechanical arm is connected to a second depth camera;

[0039] The first depth camera and / or the second depth camera are configured to capture an image of the guide rail to be inspected;

[0040] The image of the guide rail to be inspected is input into the YOLOv5 model with the attention mechanism module introduced to identify the surface defects of the guide rail to be inspected;

[0041] After obtaining the surface defects of the guide rail to be inspected, the position information of the mobile robot is obtained and sent to the host computer.

[0042] Preferably, the free end of the right mechanical arm is connected to a first depth camera, and the free end of the left mechanical arm is connected to a second depth camera;

[0043] The first depth camera and / or the second depth camera are configured to capture an image of a bolt group on the guide rail to be inspected;

[0044] The image of the bolt group on the guide rail to be detected is input into the YOLOv5 model with the attention mechanism module introduced to identify the bolt group, and then the image of a single bolt is cropped according to the anchor box information in the recognition result;

[0045] The image of a single bolt is processed by an edge detection algorithm based on the Canny operator to obtain the edge contour of the bolt;

[0046] The edge contour is extracted by the Douglas-Peucker algorithm, and the bolt edge line equation is fitted based on the edge points and corner points by the least squares method. The angle θ between the straight line and the x-axis is calculated based on the slope k of the edge line equation using the inverse tangent function, and is converted from radians to degrees using the following formula (7):

[0047]

[0048] Formula (7) is used to solve the angle θ corresponding to the six edges of a single bolt: i , (i = 1, 2, ..., 6), and then the angle values ​​of the six edges are quantized to the angle value of the bolt itself θ by the following formula (8): bolt ,

[0049]

[0050] The quantitative angle value of the bolt in the current state is calculated by formula (8): The quantitative angle value θ of the bolt in normal state is calculated by formula (8): bolti , and then the bolt loosening angle Δθ is calculated by the following formula (9): i :

[0051]

[0052] The loose angle Δθ i Compared with the set threshold, if the loose angle Δθ i If it is greater than the threshold, it is judged as loose;

[0053] The three-dimensional coordinates of the loose bolts in the robot coordinate system are calculated, and the three-dimensional coordinate information is transmitted to the robot arm path planning algorithm to obtain the desired motion path of the robot arm. Then the right robot arm or the right robot arm moves to make the end effector one or the end effector two move to tighten the loose bolts.

[0054] The beneficial effect of the present invention is that it replaces manual maintenance work on the rigid guide rail and enables robots to perform maintenance work, with a high degree of automation and intelligence, avoiding the occurrence of safety accidents and ensuring the safety of personnel.

[0055] It greatly improves the efficiency of maintenance work and has the advantages of high efficiency, good operability and high flexibility.

[0056] The detection accuracy is high. It can accurately detect defects such as deformation of the guide rail, abnormal conditions such as loose bolt group, misalignment, relative inclination, gap, etc. between two adjacent sections of track, and abnormal connection conditions such as misalignment, relative inclination, gap, etc.

[0057] It can adapt to working conditions under different lighting conditions, suitable for weather such as strong sun, haze or heavy fog.

[0058] It can accurately determine the position information of the robot and precisely locate the robot. It can also accurately determine the position information of defects and abnormalities on the guide rail.

[0059] The robot has various functions, and the mechanical arm installed on it can repair bolts in abnormal conditions.

[0060] Further features and aspects of the present invention will be clearly described in the following description of specific embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is an axonometric view of the maintenance robot;

[0062] Figure 2 yes Figure 1 A front view of the robot shown;

[0063] Figure 3 yes Figure 1 a left side view of the robot shown;

[0064] Figure 4 yes Figure 1 An axonometric view of the robot from another perspective;

[0065] Figure 5 It is an exploded view of the mobile platform;

[0066] Figure 6 It is an axonometric view of the mobile platform;

[0067] Figure 7 It is an axonometric view of the mobile platform;

[0068] Figure 8 It is an exploded view of the second rear clamping mechanism;

[0069] Fig. 9 is a cross-sectional view of the second rear clamping mechanism;

[0070] Fig.10 It is a schematic diagram of the structure in which two robotic arms are installed on the robotic arm support beam;

[0071] Fig.11 It is a schematic diagram of the structure of two robot arm connection bases installed on the linear module on the bottom surface of the robot arm support beam;

[0072] Fig.12 is a structural schematic diagram of a first depth camera and a second depth camera installed at the free ends of a right robotic arm and a left robotic arm respectively;

[0073] Fig.13 yes Fig.12a side view of the structure shown;

[0074] Fig.14 It is a flow chart of the robot localization algorithm;

[0075] Fig.15 This is a flowchart for improving the Canny algorithm;

[0076] Fig.16 It is a flow chart of the bolt loosening determination algorithm;

[0077] Fig.17 It is a flowchart of the robot's operation.

[0078] Explanation of symbols in the figure:

[0079] 1. Support frame, 1-1. Right side plate, 1-1-1. Motor connection hole, 1-2. Left side plate, 1-3. Rear vertical plate, 1-4. Front vertical plate, 1-5. Top plate, 1-6. Rear closing plate, 1-7. Front closing plate, 1-8. Connecting shaft, 1-9. Triangular bracket, 1-10. Triangular bracket, 1-11. Triangular bracket, 1-12. Triangular bracket, 1-13. Eye screw for self-locking device connection, 1-14. First rear adjusting screw, 1-15. Second rear adjusting screw, 1-16. First front adjusting screw, 1-17. Second front adjusting screw, 1-1 8. Boss, 1-19. First rear bearing seat, 1-20. Second rear bearing seat, 1-21. First front bearing seat, 1-22. Second front bearing seat, 1-23. Motor connection hole; 2. Active synchronous pulley, 3. Driven synchronous pulley, 4. Rear cross plate, 5. First rear clamping mechanism, 5-1. Roller positioning block, 5-2. Slider, 5-3. Tension spring, 5-4. Roller, 5-5. Upper eye screw; 6. Second rear clamping mechanism, 6-1. Roller positioning block, 6-1-1. Threaded hole, 6-2. Slider, 6-3. Inner bearing, 6-4. Outer bearing , 6-5. Shaft end retaining ring, 6-6. Bearing end cover, 6-7. Screw, 6-8. Tension spring, 6-9. Lower lifting eye screw, 6-10. Upper lifting eye screw, 6-11. Guide rail, 6-12. Roller; 7. Front cross plate, 8. Second front clamping mechanism, 8-1. Roller positioning block, 8-2. Tension spring, 8-3. Upper lifting eye screw, 8-4. Lower lifting eye screw, 8-5. Guide rail, 8-6. Roller, 9. Active roller, 10. Driven roller, 11. Robot arm support beam, 12. Linear guide rail assembly one, 12-1. First slider, 12-2. Three sliders, 13. Linear slide assembly 2, 13-1. Second slider, 14. Right screw motor, 15. Right nut, 16. Right nut seat, 17. Right robot arm connecting base, 18. Right screw support seat, 19. Left screw motor, 20. Left robot arm connecting base, 21. Left screw support seat; 22. Right robot arm, 23. Left robot arm; 24. End effector 1, 25. End effector 2; 26. First depth camera, 27. Second depth camera; 28. I-shaped rigid guide rail; 29. ​​Foot nails, 30. Inter-segment connecting bolts; 31. Tension sensor. DETAILED DESCRIPTION

[0080] The present invention will be further described in detail below with reference to the accompanying drawings by way of specific embodiments.

[0081] like Figure 1-4As shown, the intelligent maintenance robot for the anti-fall guide rail of the power tower includes a mobile platform, a robotic arm connecting device, a right robotic arm 22, a left robotic arm 23, an end effector 1 24, and an end effector 2 25. The robotic arm connecting device is installed on the mobile platform, and the right robotic arm and the left robotic arm are respectively installed on the robotic arm connecting device; the end effector 1 24 is connected to the free end of the right robotic arm 22, and the end effector 2 25 is connected to the free end of the left robotic arm 23.

[0082] The mobile platform includes a supporting frame 1, an active synchronous pulley 2, a driven synchronous pulley 3, a synchronous belt, a driving motor, an active roller 9, a driven roller 10, a first rear clamping mechanism 5, a second rear clamping mechanism 6, a first front clamping mechanism, and a second front clamping mechanism 8.

[0083] like Figure 5-7As shown, the support frame 1 includes a right side plate 1-1, a left side plate 1-2, a rear vertical plate 1-3, a front vertical plate 1-4, a top plate 1-5, a rear horizontal plate 4, a front horizontal plate 7, a rear closing plate 1-6, a front closing plate 1-7, a connecting shaft 1-8, a triangular bracket 1-9, a triangular bracket 1-10, a triangular bracket 1-11, a triangular bracket 1-12, a lifting screw 1-13 for connecting the self-locking device, a first rear adjusting screw 1-14, a second rear adjusting screw 1-15, a first front adjusting screw 1-16, a second front adjusting screw 1-17, a boss 1-18, a first rear bearing seat 1-19, a second rear bearing seat 1-20, a first front axle The bearing seat 1-21, the second front bearing seat 1-22, the lower part of the right side plate 1-1 and the lower part of the left side plate 1-2 are fixedly connected by three connecting shafts 1-8, the rear vertical plate 1-3 is fixedly connected to the rear end of the right side plate 1-1 and the rear end of the left side plate 1-2 by screws, the front vertical plate 1-4 is fixedly connected to the front end of the right side plate 1-1 and the front end of the left side plate 1-2 by screws, the top plate 1-5 is fixedly connected to the top of the right side plate 1-1 and the top of the left side plate 1-2 by screws, the rear horizontal plate 4 is fixedly connected to the rear end of the top of the right side plate 1-1 and the rear end of the top of the left side plate 1-2 by screws, and the front horizontal plate 1-4 is fixedly connected to the front end of the right side plate 1-1 and the front end of the left side plate 1-2 by screws. The plate 7 is fixedly connected to the front end of the top of the right side plate 1-1 and the front end of the top of the left side plate 1-2; since the middle of the top of the right side plate 1-1 is low and the two ends are high, and the middle of the top of the left side plate 1-2 is low and the two ends are high, the rear closing plate 1-6 is fixedly connected to the side of the rear end of the top of the right side plate 1-1 and the side of the rear end of the top of the left side plate 1-2 by screws, and the front closing plate 1-7 is fixedly connected to the side of the front end of the top of the right side plate 1-1 and the side of the front end of the top of the left side plate 1-2 by screws; the boss 1-18 is fixedly connected to the top plate 1-5 by screws; two triangular brackets 1-10 are connected to the rear horizontal plate 4 and Between the right side plate 1-1, two triangular brackets 1-9 are connected between the rear horizontal plate 4 and the left side plate 1-2, two triangular brackets 1-11 are connected between the front horizontal plate 7 and the left side plate 1-2, and two triangular brackets 1-12 are connected between the front horizontal plate 7 and the right side plate 1-1; the self-locking device connection eye screw 1-13 is connected to the rear vertical plate 1-3; the first rear adjustment screw 1-14 and the second rear adjustment screw 1-15 are respectively connected to the corresponding threaded holes on the rear horizontal plate 4 and pass through the rear horizontal plate 4; the first front adjustment screw 1-16 and the second front adjustment screw 1-17 are respectively connected to the corresponding threaded holes on the front horizontal plate 7 and pass through the front horizontal plate 7. The first rear bearing seat 1-19 is connected to the left side plate 1-2, and the second rear bearing seat 1-20 is connected to the right side plate 1-1. The first front bearing seat 1-21 is connected to the left side plate 1-2, and the second front bearing seat 1-22 is connected to the right side plate 1-1.

[0084] An accommodation space is formed between the right side plate 1-1, the left side plate 1-2, the rear vertical plate 1-3, the front vertical plate 1-4 and the top plate 1-5, and the driving motor, the driving roller 9 and the driven roller 10 are located in the accommodation space. The right side plate 1-1 is provided with a motor connection hole 1-1-1, and the driving motor is fixedly installed at the motor connection hole 1-1-1. The output shaft of the driving motor extends out from the motor connection hole, and the driving synchronous pulley 2 is connected to the output shaft of the driving motor. The driving roller 9 is provided with a rotating shaft, the left end of which is connected to the bearing in the first rear bearing seat 1-19, the right part of which is connected to the bearing in the second rear bearing seat 1-20, and the right end of which extends outward from the right side plate 1-1. The driving roller 9 can rotate under the support of the two bearing seats. The driven synchronous pulley 3 is fixedly connected to the right end of the rotating shaft of the driving roller 9. The synchronous belt is connected between the driven synchronous pulley 3 and the driving synchronous pulley 2. The driven roller 10 is provided with a rotating shaft, the left end of which is connected to the bearing in the first front bearing seat 1-21, and the right end of which is connected to the rotating shaft in the second front bearing seat 1-22. With the support of the first front bearing seat 1-21 and the second front bearing seat 1-22, the driven roller 10 can rotate. When the driving motor is working, the driving roller 9 is rotated through the driving synchronous pulley 2, the synchronous belt, and the driven synchronous pulley 3. An electromagnetic power-off brake can be installed at the hub rib of the driven roller 10 to reliably stop the driven roller 10 and achieve reliable braking of the entire robot.

[0085] like Fig.10 and 11 As shown, the robot arm connecting device includes a robot arm supporting beam 11, a linear slide assembly 12, a linear slide assembly 2 13, a right screw motor 14, a right nut 15, a right nut seat 16, a right robot arm connecting base 17, a right screw support seat 18, a left screw motor 19, a left nut, a left nut seat, a left robot arm connecting base 20, and a left screw support seat 21. The linear slide assembly 12 and the linear slide assembly 2 13 are connected side by side to the bottom surface of the robot arm supporting beam 11. The linear slide assembly 12 is provided with a first slider 12-1 and a third slider 12-2. The linear slide assembly 2 13 is provided with a second slider 13-1 and a fourth slider. The right screw motor 14 It is fixedly installed on the right end of the bottom surface of the robot arm support beam 11, the right screw support seat 18 is fixedly installed in the middle of the bottom surface of the robot arm support beam 11, the right screw motor 14 is provided with a screw, and the right nut 15 is connected to the screw, the end of the screw of the right screw motor 14 is connected to the right screw support seat 18, the right nut seat 16 is connected to the right nut 15, the right robot arm connecting base 17 is fixedly connected to the right nut seat 16 by screws, and the two sides of the right robot arm connecting base 17 are respectively fixedly connected to the first slider 12-1 and the second slider 13-1, and the right screw motor 14 can drive the right robot arm connecting base 17 to translate along the length direction of the robot arm support beam 11.

[0086] The left screw motor 19 is fixedly installed at the left end of the bottom surface of the robot arm support beam 11, and the left screw support seat 21 is fixedly installed in the middle part of the bottom surface of the robot arm support beam 11. The left screw motor 19 is provided with a screw, and the left nut is connected to the screw, the end of the screw of the left screw motor 19 is connected to the left screw support seat 21, the left nut seat is connected to the left nut, and the left robot arm connecting base 20 is fixedly connected to the left nut seat by screws, and the two sides of the left robot arm connecting base 20 are respectively fixedly connected to the third slider 12-2 and the fourth slider. The operation of the left screw motor can drive the left robot arm connecting base 20 to translate along the length direction of the robot arm support beam 11.

[0087] like Figure 5 , 8 As shown in Figure 9, the second rear clamping mechanism 6 includes a roller positioning block 6-1, a slider 6-2, an inner bearing 6-3, an outer bearing 6-4, a shaft end retaining ring 6-5, a bearing end cover 6-6, a screw 6-7, a tension spring 6-8, a lower eye screw 6-9, an upper eye screw 6-10, a guide rail 6-11, and a roller 6-12. The top surface of the roller positioning block 6-1 is provided with a threaded hole 6-1-1, and the roller positioning block 6-1 is provided with an inner bearing chamber and an outer bearing chamber. The inner The bearing 6-3 is located in the inner bearing chamber, the outer bearing 6-4 is located in the outer bearing chamber, the roller 6-12 is connected to the inner bearing 6-3, the roller 6-12 is connected to the outer bearing 6-4, the roller 6-12 can rotate under the support of the inner bearing 6-3 and the outer bearing 6-4, the shaft end retaining ring 6-5 is connected to the end of the roller 6-12, the bearing end cover 6-6 is connected to the outer side of the roller positioning block 6-1, and the bearing end cover 6-6 blocks the outer bearing 6-4. Four screws 6-7 fix the slider 6-2 to the inner side of the roller positioning block 6-1, the lower eye screw 6-9 is connected to the threaded hole 6-1-1, the lower end of the tension spring 6-8 is hung on the lower eye screw 6-9, and the upper end of the tension spring 6-8 is hung on the upper eye screw 6-10, the slider 6-2 is connected and matched with the guide rail 6-11, and the slider 6-2 can slide along the guide rail 6-11.

[0088] refer to Figure 5 , 6 and 7, the guide rail 6-11 is fixedly installed at the rear end of the right side plate 1-1, and the upper eye screw 6-10 is connected to the right end of the rear cross plate 4. The roller 6-12 is located below the right side plate 1-1. The lower end of the second rear adjusting screw 1-15 contacts the top of the roller positioning block 6-1. When the second rear adjusting screw 1-15 is turned, the lower end of the second rear adjusting screw 1-15 moves downward to press the roller positioning block 6-1, and the roller positioning block 6-1 moves downward. Then, when the second rear adjusting screw 1-15 is turned in the opposite direction, the lower end of the second rear adjusting screw 1-15 moves upward. Under the tension of the tension spring 6-8, the roller positioning block 6-1 moves upward, and the roller 6-12 moves upward under the drive of the roller positioning block 6-1.

[0089] The structure of the first rear clamping mechanism 5 is the same as that of the second rear clamping mechanism 6, the structure of the second front clamping mechanism 8 is the same as that of the second rear clamping mechanism 6, and the structure of the first front clamping mechanism is the same as that of the second rear clamping mechanism 6. The entire robot is equipped with four sets of clamping mechanisms, two sets at the rear and two sets at the front.

[0090] refer to Figure 1 , 3 Under the tension of the tension spring 6-8, the roller 6-12 of the second rear clamping mechanism 6 presses upward on the top inner side of the I-shaped rigid guide rail, and the roller 6-12 applies a certain pressure to the top inner side of the I-shaped rigid guide rail. The roller 6-12 is inclined to adapt to the top inner side of the I-shaped rigid guide rail.

[0091] refer to Figure 3 The first rear clamping mechanism 5 and the second rear clamping mechanism 6 are arranged symmetrically. As can be seen from the figure, the roller positioning block 5-1 is symmetrical to the roller positioning block 6-1. The guide rail of the first rear clamping mechanism 5 is fixedly installed at the rear end of the left side plate 1-2. The guide rail is connected and matched with the slider 5-2. The upper eye screw 5-5 of the first rear clamping mechanism 5 is connected to the left end of the rear cross plate 4 (as shown in FIG. Figure 5 As shown in the figure, the upper end of the tension spring 5-3 is hung on the upper eye screw, and the lower end of the tension spring 5-3 is hung on the lower eye screw of the first rear clamping mechanism 5. The roller 5-4 is located below the left side plate 1-2. As can be seen from the figure, the roller 5-4 presses upward on the top inner side of the I-shaped rigid guide rail, and the roller 5-4 applies a certain pressure to the top inner side of the I-shaped rigid guide rail. The roller 5-4 is inclined to adapt to the top inner side of the I-shaped rigid guide rail. The lower end of the first rear adjusting screw 1-14 contacts the top of the roller positioning block 5-1. When the first rear adjusting screw 1-14 is turned, the lower end of the first rear adjusting screw 1-14 moves downward, which will press the roller positioning block 5-1 downward, and the roller positioning block 5-1 moves downward; then, when the first rear adjusting screw 1-14 is turned in the opposite direction, the lower end of the first rear adjusting screw 1-14 moves upward, and under the tension of the tension spring 5-3, the roller positioning block 5-1 moves upward.

[0092] refer to Figure 2 In the second front clamping mechanism 8, the lower eye screw 8-4 is connected to the roller positioning block 8-1, the lower end of the tension spring 8-2 is hung on the lower eye screw 8-4, and the upper end of the tension spring 8-2 is hung on the upper eye screw 8-3. Figure 5, the guide rail 8-5 is fixedly installed at the front end of the right side plate 1-1, and the upper eye screw 8-3 is connected to the right end of the front cross plate 7. The slider in the second front clamping mechanism 8 is connected and cooperated with the guide rail 8-5. The roller of the second front clamping mechanism 8 is located below the right side plate 1-1. Similarly, the roller in the inclined state in the second front clamping mechanism 8 presses upward on the top inner side of the I-shaped rigid guide rail. Adjust the second front screw 1-17 to make its lower end move downward, and then press the roller positioning block 8-1 downward, and the roller positioning block 8-1 moves downward, and then reversely rotate the second front adjustment screw 1-17, and under the tension of the tension spring 8-2, the roller positioning block 8-1 moves upward.

[0093] The first front clamping mechanism is symmetrically arranged with the second front clamping mechanism 8. Similarly, the roller in the inclined state in the first front clamping mechanism presses upward on the top inner side of the I-shaped rigid guide rail. Screwing the first front adjustment screw 1-16 can make the roller positioning block in the first front clamping mechanism move downward or upward. The roller of the first front clamping mechanism is located below the left side plate 1-2.

[0094] The bottom surface of the robot arm support beam 11 is fixedly connected to the boss 1-18 by screws, and then the right robot arm 22 is fixedly installed on the right robot arm connecting base 17 by screws, and the left robot arm 23 is fixedly installed on the left robot arm connecting base 20 by screws.

[0095] The end effector 1 24 is connected to the free end of the right mechanical arm 22, and the end effector 25 is connected to the free end of the left mechanical arm 23. As shown in the figure, the end effector 1 24 and the end effector 2 25 are specifically electric impact wrenches. The motor in the electric impact wrench is a brushless DC motor. The electric impact wrench uses the torque-angle method to tighten the bolts on the guide rail. The work process is divided into three stages, wherein the first stage is the thread recognition stage, in which the wrench needs to overcome the friction torque of the thread. The stroke in this stage is longer, the resistance is the smallest, and the required speed is the largest; the second stage is the fitting stage, when the torque value measured by the sensor exceeds the threshold value specified in the second stage, the fitting stage is entered, in which a certain resistance needs to be overcome, and the stroke to the next stage is short, and the motor speed is reduced; the third stage is the tightening stage, when the torque measured by the torque sensor reaches the threshold value of the third stage, the tightening stage is entered, in which the angle method is used for tightening, and the motor needs to be controlled to rotate at a certain angle, with the lowest speed and the most precise control. When the torque measured by the sensor reaches the set value, the motor stops suddenly and the operation is completed. The first and second stages use the torque method for tightening, and the third stage uses the angle method for precise tightening. A three-loop control architecture consisting of a current loop, a speed loop, and a position loop is used to achieve precise control of the brushless DC motor.

[0096] The present invention uses multi-sensor fusion to locate the robot, and three types of sensors can be used, namely, IMU inertial measurement unit, encoder, and camera. IMU inertial measurement unit is usually composed of accelerometer, gyroscope and magnetometer, among which, accelerometer is used to measure the acceleration of the robot moving along the guide rail, and the speed and displacement are obtained after integral calculation; gyroscope is used to measure the angular velocity change of the robot, and the inclination of the robot can be obtained after integral calculation; magnetometer is used to measure the direction of the object relative to the earth's magnetic field. IMU inertial measurement unit has a higher operating frequency, so it has a better local positioning effect, but it will drift as time goes by, and the smaller constant error will become extremely large after multiple integrations, so it usually needs to be combined with other positioning methods to reduce the error. IMU inertial measurement unit can be installed on the front cross plate of the mobile platform.

[0097] The encoder is installed on the drive motor or the shaft of the active roller. When the robot is running, the encoder will generate a constant number of pulses for each rotation of the active roller. When the sampling period is fixed, the number of pulses generated by the rotation of the wheel during the measurement period can be calculated to calculate the number of rotations of the wheel during the time interval and then calculate the displacement of the robot relative to the guide rail. There are three main methods for measuring the speed of rotary encoders: M method (Master method), T method (Tracking method) and M / T method (Master-Tracking method). The M / T method combines the advantages of the M method and the T method. It can directly use the number of pulses output by the encoder for speed estimation, and can also correct the speed measurement through the time interval to improve the stability and accuracy of the measurement. Therefore, the speed measurement formula of the M / T method is used to obtain the rotation speed of the active roller, and then the posture parameters are obtained by the robot's motion model using the dead reckoning algorithm.

[0098] The controller can be installed on the front horizontal plate of the mobile platform, and the controller is provided with a wireless transmission module. The encoder is connected to the controller through a signal line. The IMU inertial measurement unit is connected to the controller through a signal line.

[0099] IMU positioning and encoder positioning are both relative positioning, with accumulated errors, and require regular calibration to ensure the accuracy and stability of their measurements. As a commonly used absolute positioning system, GPS can measure height, but its accuracy is usually lower than that of measuring horizontal position, ranging from a few meters to more than ten meters. To make up for the above defects, a camera-based visual positioning solution is introduced to correct the relative positioning error. According to the industry standard "Anti-fall Device for Pole Tower Operation", the length of a single-section vertical guide rail should be a fixed value of 4000mm, with an allowable deviation of ±1.5mm, a small difference, and the gap characteristics at the connection between adjacent tracks in the guide rail are obvious, which can be used for visual positioning. Therefore, the camera used for positioning is installed at the front end of the mobile platform (for example, on the front horizontal plate or the front vertical plate). When the robot rises and falls along the guide rail, the camera mounted on the front end of the robot captures the image of the guide rail to obtain continuous image frames, and the controller pre-processes and detects the edge of the image frame. The edge detection uses the Hough transform algorithm, which can calculate the possible straight line formed by the edge pixel points in the image to identify the gap at the connection between adjacent tracks. A gap is identified and counted once to obtain the number of gaps. The number of gaps is multiplied by the length of a single track to obtain the absolute position information of the robot, thereby realizing the visual positioning of the robot.

[0100] IMU positioning and encoder positioning are methods to infer the robot's position based on the robot's kinematic model. It does not require external environmental information and has high accuracy in a short period of time, but there is an error accumulation problem. Although the visual positioning method has high accuracy, its positioning is discontinuous, and there may be misidentification and missed counting of the guide rail gap. For this reason, a suitable fusion algorithm is selected to fuse the obtained positioning data to achieve the complementary advantages of the three positioning methods. Commonly used fusion algorithms include particle filtering and Kalman filtering algorithms. Among them, particle filtering can be used for nonlinear systems, but the calculation amount is large; Kalman filtering has a small amount of calculation, but it is only suitable for linear systems. In order to overcome the shortcomings, neural networks can be introduced or improved algorithms such as extended Kalman filtering or error Kalman filtering can be used. After comprehensive consideration, the Kalman filter algorithm is selected to complete the multi-sensor fusion positioning, and the neural network is used to improve it to improve the positioning accuracy and stability of the system. Reference Fig.14 The core idea of ​​the Kalman filter algorithm is to construct a vector x from the various states of the robot at time k. k Let the robot posture information obtained by trajectory calculation based on the data collected by the encoder be α b =[x b ,y b ,z b ,θ b ] T , the robot posture information obtained by integrating the data collected by IMU is α i =[xi ,y i ,z i ,θ i ] T , the robot posture information obtained by visual positioning through the camera is α v =[x v ,y v ,z v ,θ v ] T , then we get x k =[x,y,z,θ] T , where x, y, z, θ are given by α b , α i With α v The robot posture information is obtained after weighted fusion by the Kalman filter algorithm. From this, the prediction model of the robot's state at time k can be obtained:

[0101]

[0102] In formula (1), x k is the state vector x at time k k The predicted value, x k-1 is the state vector x at time k-1 k-1 The predicted value, F k is the state transfer matrix, B k is the control matrix, u k It is the control quantity of external input.

[0103] Similarly, there is an observation model of the system at time k:

[0104] z k =H k x k (2)

[0105] In formula (2), z k is the actual observed value of the system state at time k, H k is the observation matrix, which is used to map the state vector to the observation space.

[0106] At each moment, the Kalman filter first predicts the current state based on the prediction model, and the predicted covariance matrix P k It will also be updated as the state evolves:

[0107]

[0108] The trade-off between the reliability of each sensor and the fusion of the predicted value and the actual observation value is determined by the Kalman gain K k In the traditional Kalman filter, K k The calculation formula is

[0109]

[0110] In formula (4), R k is the noise covariance matrix of the observations.

[0111] For complex nonlinear systems, P k With R k The change of is difficult to estimate accurately, which means that the traditional Kalman filter may no longer be effective. The BP neural network has nonlinear mapping and adaptive capabilities. The present invention uses the BP neural network to fuse the extended Kalman filter to improve the accuracy of attitude solution. Other neural networks, such as Elman neural network, convolutional neural network, etc., can also be used.

[0112] The BP neural network generates an optimized Kalman gain by learning the deviation between the real state and the estimated state in the historical data. The input of the network includes the observed value, predicted value, and current covariance information of the system state, and the output is the optimized gain.

[0113] The BP neural network structure includes input layer, hidden layer and output layer. The input quantity in the input layer includes the predicted value x k-1 , observation value z k , the predicted covariance matrix P k and the noise covariance matrix R k ; The hidden layer is set to 3 layers according to engineering experience, with 15 neurons in each layer, and the activation function is the Sigmoid function; the output layer uses a single neuron, and the activation function is a linear activation function. After receiving the observation and prediction information at the current moment, the trained BP neural network will output the optimized Kalman gain K opt :

[0114]

[0115] The Kalman filter can use the optimized Kalman gain K opt Update the state estimate and output:

[0116]

[0117] Since the estimated value of the above state vector is generated based on the fusion of IMU unit and encoder data, it belongs to relative positioning, which will produce cumulative errors over time, so it needs to be calibrated by absolute positioning. As mentioned above, the vision-based absolute positioning idea uses the Hough transform algorithm to identify the gaps at the connection of adjacent tracks and count them. Every time the robot passes through a gap at a track connection, the counter increments by one, and the count is multiplied by the length of the single track section to obtain the absolute position of the robot at this moment. When the robot detects that it has passed through a gap, the α of the input BP neural network is v A mutation will occur. At this time, the BP neural network is solving the Kalman gain K opt When α v Assign a sufficiently large weight so that the final positioning estimate x output by the Kalman filter is k Mainly from the absolute positioning information α obtained this time v This process realizes the positioning calibration of the robot. After the calibration is completed, the calibrated positioning estimate x is recorded. k , and assign it to α v As the subsequent input of the BP network, until the robot passes the next section of the rail gap to generate a new α v .

[0118] like Fig.12 and 13 As shown, the first depth camera 26 is connected to the free end of the right mechanical arm 22 through a camera bracket, and the second depth camera 27 is connected to the free end of the left mechanical arm 23 through a camera bracket. Considering that the ambient light may be too strong or too weak due to weather factors, affecting the effect of robot visual detection, the present invention uses a visual algorithm based on deep learning to detect surface defects of the guide rail and identify abnormal bolt groups. The image of the guide rail is collected by the first depth camera 26 and / or the second depth camera 27.

[0119] The present invention selects the YOLOv5 algorithm to identify and detect surface defects such as guide rail deformation and bolt group anomalies. YOLOv5 is a target detection model based on anchor frames. Its core idea is to divide the target detection task into three subtasks: bounding box regression, target classification and confidence prediction. The YOLOv5 model predicts the position of the target through the anchor frame, and filters the redundant prediction frames through the non-maximum suppression (NMS) method to obtain the final detection result. The network structure of the YOLOv5 model includes the following three parts: Backbone, Neck and Head. The Backbone of YOLOv5 is used to extract multi-scale features of the input image. It uses CSPDarknet53 as the backbone network, and effectively reduces the amount of calculation through the cross-stage partial network CSP module, while enhancing the expression ability of the features; the Neck part of YOLOv5 is mainly used for multi-scale feature fusion to improve the detection ability of targets of different sizes. The Neck part also introduces the SPPF module, which increases the receptive field and aggregates key information by performing multi-scale pooling operations on the feature map; the Head part of YOLOv5 is responsible for generating the final detection result. It adopts an anchor box-based prediction method, divides the feature map into multiple grids, predicts multiple anchor boxes for each grid, and outputs the target category, bounding box coordinates and confidence information. NMS is used to remove low-confidence or overlapping prediction boxes to obtain the final detection result.

[0120] When the robot is actually working, it may encounter extreme weather such as sun exposure and fog, which will reduce the image quality and thus affect the recognition effect. In order to further improve the adaptability of the selected YOLOv5 model to the working environment of this project, the attention mechanism can be introduced to improve it. Commonly used modules include SE, CBAM, BAM, CA, etc. For example, the CA module is used to improve the YOLOv5 model. In the CA module, the input feature map is averaged and pooled in both width and height directions, and the position information is embedded in the channel attention, so as to achieve the effect of obtaining channel attention and spatial attention at the same time. The Backbone part of YOLOv5 contains the key component C3 module, which consists of three Conv modules and a Bottleneck module. The feature extraction ability of the network is further enhanced through residual connection and feature fusion. By adding the CA attention module to the BottleNeck residual connection of the C3 module, and then replacing all the C3 modules in the original YOLOv5 backbone network with the improved C3 module, the network model of YOLOv5 with the fusion attention mechanism is constructed. Using the improved YOLOv5 model, the robot can pay more attention to the area of ​​interest during the detection process and overcome the impact of strong or weak light environments to a certain extent.

[0121] The training data set required to train the improved YOLOv5 model can be expanded by taking photos on the spot and performing data enhancement operations such as positive and negative sample balance and geometric transformation.

[0122] For bolts that connect adjacent rails, if the bolts are loose, the image recognition method is used to judge whether they are abnormally loose. First, the bolt group is identified by the improved YOLOv5 model, and then the sub-image containing only a single bolt is cropped according to the anchor frame information in the recognition result. Then, the edge detection algorithm based on the Canny operator is used to perform edge detection on the single bolt image. The traditional Canny operator can be improved by using methods such as bilateral filtering, non-maximum suppression, and dual threshold detection to eliminate the pseudo edges caused by background noise as much as possible and improve the accuracy of the algorithm. Reference Fig.15 , because the traditional Canny operator uses the first-order derivative of the two-dimensional Gaussian function as an image filter, and the Gaussian filter will cause the image to lose detail information after smoothing the image, resulting in blurred edges. The present invention uses bilateral filtering as an improved Canny operator image filter to obtain a more stable and accurate smoothed image; after obtaining the grayscale image information, the Canny operator obtains a series of potential edge pixels by performing gradient calculation on the image, and introduces a non-maximum suppression method to compare and analyze each potential pixel with its surrounding neighborhood along its specific gradient direction. If the gradient intensity of the pixel significantly exceeds that of the adjacent pixels on both sides of its gradient direction, it is considered to be a valid edge pixel; otherwise, it is considered to be a false detection result and is removed. Finally, a double threshold detection is used to perform secondary processing on the processed pixel set, and the pixels are screened according to the high and low threshold coefficients and histogram information of the image to obtain the final edge point information with high confidence.

[0123] refer to Fig.16 After extracting the edge point information, the bolt profile is obtained. The Douglas-Peucker algorithm is used to extract the corner points of the edge profile, and the bolt edge equation is fitted by the least squares method based on the edge points and corner points. After obtaining the edge equation expression, the bolt looseness assessment can be performed. Using the inverse tangent function, the angle θ between the straight line and the x-axis is calculated based on the slope k of the edge equation, and it is converted from radians to degrees using the following formula (7).

[0124]

[0125] The angle θ corresponding to the six edges of a single bolt is solved in turn from the above formula: i , (i=1,2,...,6). Then the angle values ​​of the six edges are quantized to the angle value θ of the bolt itself by the following formula (8): bolt ,

[0126]

[0127] The range of the angle value of the bolt itself is θ bolt ∈[0,60].

[0128] Before the robot is put into use for inspection, the edge detection and angle quantification of the bolts in normal state on the guide rail are performed to obtain the quantified angle value θ in normal state. bolti During the robot operation, the actual detected quantitative angle value of the bolt in the current state is Then, the quantized angle value in the current state is obtained by the following formula (9): Compared with the quantized angle value θ in the normal state bolti After making the difference, take the absolute value to get the loosening angle Δθ of the i-th bolt i ,

[0129]

[0130] The loose angle Δθ i Compared with the set threshold, if the loose angle Δθ i If it is greater than the threshold, it is judged as loose.

[0131] The specific value of the threshold can be 3σ, where σ is the quantized angle value θ under multiple normal conditions. bolti The standard deviation of the error.

[0132] refer to Fig.17 , the process of using the above robot is as follows:

[0133] In the first step, the robot is mounted on an I-shaped rigid guide rail 28 located in a vertical direction.

[0134] The staff uses an Allen wrench to tighten the first rear adjusting screw 1-14, the second rear adjusting screw 1-15, the first front adjusting screw 1-16, and the second front adjusting screw 1-17 on the front and rear horizontal plates of the mobile platform to make them move downward, thereby pushing the roller positioning block in contact with them to move downward until the distance between the roller of the clamping mechanism and the bottom of the mobile platform is greater than the thickness of the flange of the I-shaped rigid guide rail 28. At this time, the robot can be installed on the I-shaped rigid guide rail 28, so that the active roller 9 and the driven roller 10 are in contact with the upper surface of the flange of the I-shaped rigid guide rail 28, and the four rollers are in contact with the lower surface of the flange of the I-shaped rigid guide rail 28. After that, the staff reversely rotates the first rear adjusting screw 1-14, the second rear adjusting screw 1-15, the first front adjusting screw 1-16, and the second front adjusting screw 1-17 to make them rise, and each roller positioning block will move upward under the action of the corresponding tension spring restoring force, and finally a sufficient positive pressure (such as 1.50°) is generated between the four rollers and the lower surface of the flange of the I-shaped rigid guide rail 28. Figure 1 , 2, 3), the friction force provided by the positive pressure is sufficient to overcome the robot's own gravity. At this point, the robot is installed on the guide rail.

[0135] In the second step, the staff sends a start command to the robot through the host computer, and the controller controls the drive motor on the mobile platform to work, and the robot will climb along the I-shaped rigid guide rail 28. Figure 2 The arrow in the figure indicates the direction.

[0136] The control of the drive motor adopts PID closed-loop control of the speed outer loop and the current inner loop to ensure the smoothness of the lifting and the accuracy of the control. The speed loop uses the manually input target speed value as input, and the actual speed feedback from the motor encoder as feedback. The two are fuzzy PID adjusted and the output is used as the input of the current loop; the current loop takes the output of the speed loop PID adjustment as input, and takes the actual motor current measured by the Hall element inside the motor as the feedback of the current loop. The input and feedback of the current loop are used to adjust the fuzzy PID input to the drive wheel motor to achieve closed-loop control.

[0137] In the third step, the robot traverses along the I-shaped rigid guide rail 28 while climbing. During the traversal, the IMU, encoder, and camera used for positioning collect information, and the real-time position of the robot is calculated and recorded through the multi-sensor fusion positioning algorithm. The guide rail defect detection algorithm is used to detect whether there are deformation, dents, corrosion and other defects on the upper surface of the guide rail. The detection of loose bolt groups is completed by the bolt recognition detection algorithm.

[0138] When it is confirmed that the guide rail has defects such as deformation, dents, corrosion, etc., the controller sends the robot's position information to the host computer, and the staff will then climb the tower to perform maintenance.

[0139] The first depth camera 26 and the second depth camera 27 at the end of the two sets of robotic arms traverse the bolts on the guide rails, and the objects traversed include the inter-segment bolt group and the U-shaped bolts. When the bolts are detected to be loose, the robot's drive motor stops to brake automatically, and the three-dimensional coordinates of the loose bolts in the robot coordinate system are calculated, and then the three-dimensional coordinate information is transmitted to the robotic arm path planning algorithm, and then the desired motion path of the robotic arm is calculated through forward and inverse kinematics, and then the robotic arm is controlled to move the end effector at the end of the robotic arm to the loose bolt, and the end effector 1 24 (electric impact wrench) and the end effector 2 25 (electric impact wrench) act to tighten the loose bolts.

[0140] In the fourth step, the robot climbs to the top of the guide rail to complete the maintenance work of the entire guide rail. After that, the robot begins to descend and return to the starting position, waiting for the staff to recover it.

[0141] It can be seen that the above robot inspection and maintenance is highly efficient, and can automatically detect defects in the guide rail with high detection accuracy; it can automatically detect abnormalities such as loosening of inter-section bolts on the guide rail with high detection accuracy. It can accurately determine the position information of the robot and the position information of defects and abnormalities on the guide rail.

[0142] It should be noted that the staff can hang the anti-fall self-locking device on the self-locking device connection eye screw 1-13 behind the mobile platform, and position the self-locking device on the guide rail (the self-locking device passes through the guide rail flange). The robot climbs and moves to drive the self-locking device to move on the guide rail. If the guide rail has defects such as warping, misalignment, and excessive gaps, the self-locking device will have adverse reactions, and the self-locking device will move abnormally, be difficult to move, or be unable to move. Based on these adverse reactions, it can be determined that the guide rail has defects and needs maintenance. Furthermore, the eye screw 1-13 for the self-locking device connection can be replaced with a tension sensor 31, such as Fig.12 As shown, the signal line of the tension sensor 31 is connected to the controller. When the self-locking device moves at a constant speed and smoothly, the output signal of the tension sensor 31 will be stable near a specific value; when the self-locking device jumps, stalls or cannot pass, the output signal of the tension sensor 31 will show a significant mutation. The controller receives this mutation and determines that there is a defect, or the controller controls the robot to retreat and re-pass the suspicious section after receiving this mutation. If the mutation signal appears three times, it is considered that there is a connectivity defect in this section of the guide rail, and finally the position information of the robot at this time is fed back to the host computer. The staff will then perform tower maintenance.

[0143] The above description is for the inspection and maintenance of vertical guide rails. It should be noted that some power towers, such as angle steel towers, have horizontally arranged guide rails. The robot can be installed on the horizontally arranged guide rails for inspection and maintenance operations.

[0144] It should be noted that a camera for collecting images of the I-shaped rigid guide rail 28 may be installed on the mobile platform, and the controller determines whether there is an abnormality in the I-shaped rigid guide rail based on the collected images.

Claims

1. An intelligent maintenance robot for anti-falling rails of power towers, characterized in that: The mobile platform includes a supporting frame, a driving synchronous pulley, a driven synchronous pulley, a synchronous belt, a driving motor, a driving roller, a driven roller, a first rear clamping mechanism, a second rear clamping mechanism, a first front clamping mechanism, and a second front clamping mechanism; The support frame comprises a right side plate, a left side plate, a rear vertical plate, a front vertical plate, a top plate, a rear cross plate, a front cross plate, a first rear adjusting screw, a second rear adjusting screw, a first front adjusting screw, a second front adjusting screw, a first rear bearing seat, a second rear bearing seat, a first front bearing seat, and a second front bearing seat. The lower part of the right side plate and the lower part of the left side plate are fixedly connected by three connecting shafts. The rear vertical plate is fixedly connected to the rear end of the right side plate and the rear end of the left side plate. The front vertical plate is fixedly connected to the front end of the right side plate and the front end of the left side plate. The top plate is fixedly connected to the top of the right side plate and the top of the left side plate. The rear cross plate is fixedly connected to the rear end part of the top of the right side plate and the rear end part of the top of the left side plate. The front cross plate is fixedly connected to the front end part of the top of the right side plate and the front end part of the top of the left side plate; the rear cross plate is provided with two threaded holes, the first rear adjusting screw and the second rear adjusting screw are respectively connected to the two threaded holes on the rear cross plate, and the first rear adjusting screw and the second rear adjusting screw pass through the rear cross plate; the front cross plate is provided with two threaded holes, the first front adjusting screw and the second front adjusting screw are respectively connected to the two threaded holes on the front cross plate, and the first front adjusting screw and the second front adjusting screw pass through the front cross plate; the first rear bearing seat is connected to the left side plate, and the second rear bearing seat is connected to the right side plate; the first front bearing seat is connected to the left side plate, and the second front bearing seat is connected to the right side plate; An accommodating space is formed between the right side plate, the left side plate, the rear vertical plate, the front vertical plate and the top plate, and the driving motor, the active roller and the driven roller are located in the accommodating space, the right side plate is provided with a motor connecting hole, the driving motor is connected to the motor connecting hole, and the output shaft of the driving motor extends out from the motor connecting hole; the active synchronous pulley is connected to the output shaft of the driving motor; the active roller is provided with a rotating shaft, and the left and right ends of the rotating shaft of the active roller are respectively connected to the first rear bearing seat and the second rear bearing seat, and the right end of the rotating shaft of the active roller extends outward from the right side plate; the driven synchronous pulley is fixedly connected to the right end of the rotating shaft of the active roller; the synchronous belt is connected between the driven synchronous pulley and the active synchronous pulley; the driven roller is provided with a rotating shaft, and the left and right ends of the rotating shaft of the driven roller are respectively connected to the first front bearing seat and the second front bearing seat; The second rear clamping mechanism comprises a roller positioning block, a slider, an inner bearing, an outer bearing, a tension spring, a lower eye screw, an upper eye screw, a guide rail and a roller, the roller positioning block is provided with an inner bearing chamber and an outer bearing chamber, the inner bearing is located in the inner bearing chamber, the outer bearing is located in the outer bearing chamber, the roller is connected to the inner bearing, the roller is connected to the outer bearing, the slider is fixedly connected to the inner side surface of the roller positioning block, the lower eye screw is fixedly connected to the top surface of the roller positioning block, the lower end of the tension spring is hung on the lower eye screw, the upper end of the tension spring is hung on the upper eye screw, and the slider is connected to the guide rail; The guide rail of the second rear clamping mechanism is fixedly connected to the rear end of the right side plate, the roller of the second rear clamping mechanism is located below the right side plate, the upper eye screw of the second rear clamping mechanism is connected to the right end of the rear horizontal plate; the lower end of the second rear adjusting screw contacts the top of the roller positioning block; The first rear clamping mechanism and the second rear clamping mechanism are arranged symmetrically, and the first rear clamping mechanism comprises a roller positioning block, a slider, an inner bearing, an outer bearing, a tension spring, a lower eye screw, an upper eye screw, a guide rail and a roller; the guide rail of the first rear clamping mechanism is fixedly connected to the rear end of the left side plate, the roller of the first rear clamping mechanism is located below the left side plate, the upper eye screw of the first rear clamping mechanism is connected to the left end of the rear cross plate, and the lower end of the first rear adjusting screw contacts the top of the roller positioning block of the first rear clamping mechanism; The second front clamping mechanism comprises a roller positioning block, a slider, an inner bearing, an outer bearing, a tension spring, a lower eye screw, an upper eye screw, a guide rail and a roller, the guide rail of the second front clamping mechanism is fixedly connected to the front end of the right side plate, the roller of the second front clamping mechanism is located below the right side plate, the upper eye screw of the second front clamping mechanism is connected to the right end of the front cross plate, and the lower end of the second front adjusting screw contacts the top of the roller positioning block of the second front clamping mechanism; The first front clamping mechanism is symmetrically arranged with the second front clamping mechanism, and the first front clamping mechanism includes a roller positioning block, a sliding block, an inner bearing, an outer bearing, a tension spring, a lower lifting eye screw, an upper lifting eye screw, a guide rail and a roller. The guide rail of the first front clamping mechanism is fixedly connected to the front end of the left side plate, and the roller of the first front clamping mechanism is located below the left side plate. The upper lifting eye screw of the first front clamping mechanism is connected to the left end of the front cross plate, and the lower end of the first front adjusting screw contacts the top of the roller positioning block of the first front clamping mechanism.

2. The intelligent maintenance robot for preventing the tower from falling according to claim 1 is characterized in that: The mobile platform is connected with a camera for collecting images of objects to be detected.

3. The intelligent maintenance robot for preventing the tower from falling according to claim 1 is characterized in that: The mobile platform is connected with a sensor for detecting whether the self-locking device is abnormal.

4. The intelligent maintenance robot for preventing the tower from falling according to claim 3 is characterized in that: The sensor used to detect whether the self-locking device is abnormal is a tension sensor.

5. The intelligent maintenance robot for preventing the tower from falling according to claim 1 is characterized in that: The mobile platform is connected with a sensor for detecting the position information of the robot.

6. The intelligent maintenance robot for anti-falling guide rails of power towers according to claim 1 is characterized in that: The mobile platform is connected with a mechanical arm.

7. The intelligent maintenance robot for preventing the tower from falling according to claim 1 is characterized in that: The electric tower anti-fall guide rail intelligent maintenance robot also includes a mechanical arm connecting device, a right mechanical arm and a left mechanical arm; the mechanical arm connecting device includes a mechanical arm supporting beam, a linear slide assembly 1, a linear slide assembly 2, a right screw motor, a right nut, a right nut seat, a right mechanical arm connecting base, a right screw support seat, a left screw motor, a left nut, a left nut seat, a left mechanical arm connecting base and a left screw support seat, the linear slide assembly 1 and the linear slide assembly 2 are connected side by side to the bottom surface of the mechanical arm supporting beam, the linear slide assembly 1 is provided with a first slider and a third slider, the linear slide assembly 2 is provided with a second slider and a fourth slider, the right screw motor is fixedly connected to the right end of the bottom surface of the mechanical arm supporting beam, and the right screw support seat is fixedly connected to the bottom surface of the mechanical arm supporting beam In the middle part, the right nut is connected to the screw of the right screw motor, the end of the screw of the right screw motor is connected to the right screw support seat, the right nut seat is connected to the right nut, the right mechanical arm connecting base is fixedly connected to the right nut seat, and the two sides of the right mechanical arm connecting base are respectively fixedly connected to the first slider and the second slider; the left screw motor is fixedly connected to the left end of the bottom surface of the mechanical arm supporting beam, the left screw support seat is fixedly connected to the middle part of the bottom surface of the mechanical arm supporting beam, the left nut is connected to the screw of the left screw motor, the end of the screw of the left screw motor is connected to the left screw support seat, the left nut seat is connected to the left nut, the left mechanical arm connecting base is fixedly connected to the left nut seat, and the two sides of the left mechanical arm connecting base are respectively fixedly connected to the third slider and the fourth slider; A boss is connected to the top plate of the support frame, the bottom surface of the robotic arm supporting beam is fixedly connected to the boss, the right robotic arm is fixedly connected to the right robotic arm connecting base, and the left robotic arm is fixedly connected to the left robotic arm connecting base; the free end of the right robotic arm is connected to end effector 1, and the free end of the left robotic arm is connected to end effector 2.

8. The intelligent maintenance robot for anti-falling guide rails of power towers according to any one of claims 1 to 7, characterized in that: The electric tower anti-fall guide rail intelligent maintenance robot also includes an IMU, an encoder and a camera for visual positioning, wherein the IMU is connected to a mobile platform, the encoder is connected to a drive motor or an active roller, and the camera for visual positioning is connected to the front end of the mobile platform; The camera used for visual positioning collects and detects an image of the guide rail. The image is pre-processed and then used with an edge detection algorithm to identify the gaps at the connection between adjacent tracks in the guide rail, and the number of gaps is obtained. The number of gaps is multiplied by the length of a single track to obtain the absolute position information of the robot; The robot posture information obtained by performing dead reckoning on the data collected by the encoder is α b =[x b ,y b ,z b ,θ b ] T , the robot posture information obtained by integrating the data collected by IMU is α i =[x i ,y i ,z i ,θ i ] T , the robot posture information obtained by visual positioning through the camera used for visual positioning is α v =[x v ,y v ,z v ,θ v ] T , then we get x k =[x,y,z,θ] T , where x, y, z, θ are given by α b , α i With α v The robot posture information is obtained after weighted fusion by the Kalman filter algorithm; from this, the prediction model of the robot's state at time k can be obtained: In formula (1), x k is the state vector x at time k k The predicted value, x k-1 is the state vector x at time k-1 k-1 The predicted value, F k is the state transfer matrix, B k is the control matrix, u k It is the control quantity of external input; Similarly, there is an observation model of the system at time k: z k =H k x k (2) In formula (2), z k is the actual observed value of the system state at time k, H k is the observation matrix, which is used to map the state vector to the observation space; At each moment, the Kalman filter first predicts the current state based on the prediction model, and the predicted covariance matrix P k It will also be updated as the state evolves: Kalman gain K in Kalman filter k The calculation formula is: In formula (4), R k is the noise covariance matrix of the observations; The Kalman gain K is calculated by the neural network k Optimize, the optimized Kalman gain K opt : The Kalman filter can use the optimized Kalman gain K opt Update the state estimate and output:

9. The intelligent maintenance robot for preventing the tower from falling according to claim 7 is characterized in that: The free end of the right mechanical arm is connected to a first depth camera, and the free end of the left mechanical arm is connected to a second depth camera; The first depth camera and / or the second depth camera are configured to capture an image of the guide rail to be inspected; The image of the guide rail to be inspected is input into a YOLOv5 model that introduces an attention mechanism module, thereby identifying surface defects of the guide rail to be inspected; After obtaining the surface defects of the guide rail to be inspected, the position information of the mobile robot is obtained and sent to the host computer.

10. The intelligent maintenance robot for anti-falling guide rails of power towers according to claim 7 is characterized in that: The free end of the right mechanical arm is connected to a first depth camera, and the free end of the left mechanical arm is connected to a second depth camera; The first depth camera and / or the second depth camera are configured to capture an image of a bolt group on the guide rail to be inspected; The image of the bolt group on the guide rail to be detected is input into the YOLOv5 model with the attention mechanism module introduced, so as to identify the bolt group, and then crop the image of a single bolt according to the anchor frame information in the recognition result; The image of the single bolt is processed by an edge detection algorithm based on a Canny operator to obtain an edge contour of the bolt; The edge contour is subjected to corner point extraction by the Douglas-Peucker algorithm, and the bolt edge line equation is fitted based on the edge points and corner points by the least squares method. The angle θ between the straight line and the x-axis is calculated based on the slope k of the edge line equation using the inverse tangent function, and is converted from radians to degrees using the following formula (7): Formula (7) is used to solve the angle θ corresponding to the six edges of a single bolt: i , (i = 1, 2, ..., 6), and then the angle values ​​of the six edges are quantized to the angle value of the bolt itself θ by the following formula (8): bolt , The quantitative angle value of the bolt in the current state is calculated by formula (8): The quantitative angle value θ of the bolt in normal state is calculated by formula (8): bolti , and then the bolt loosening angle Δθ is calculated by the following formula (9): i : The loose angle Δθ i Compared with the set threshold, if the loose angle Δθ i If it is greater than the threshold, it is judged as loose; The three-dimensional coordinates of the loose bolts in the robot coordinate system are calculated, and the three-dimensional coordinate information is transmitted to the robot arm path planning algorithm to obtain the desired motion path of the robot arm. Then the right robot arm or the right robot arm moves to make the end effector one or the end effector two move to tighten the loose bolts.

Images