Water-cooled wall detection wall-climbing robot and obstacle crossing method thereof

Abstract

The invention relates to the technical field of visual defect detection and robot obstacle crossing, in particular to a wall-climbing robot for water-cooled wall detection and an obstacle crossing method of the wall-climbing robot. The robot comprises a vehicle body, a walking and obstacle crossing mechanism arranged on the vehicle body and used for walking and obstacle crossing of the vehicle body, a detection and positioning system used for defect detection and defect position positioning, and a mechanical arm mechanism carrying the detection and positioning system for surface detection. A control and monitoring system for fully autonomous and semi-autonomous control of a wall-climbing robot. According to the method, the ResNet101 and the YOLOV5 are combined to carry out defect classification detection and defect target identification respectively, and defect classification and defect positioning can be carried out on the defects of the water cooling wall of the boiler rapidly and precisely.

Description

Technical Field

[0001] The invention relates to the field of visual defect detection technology and robot obstacle surmounting technology, in particular to a water-cooled wall detection climbing robot and an obstacle surmounting method thereof. Background Art

[0002] Boilers are the most important production equipment in power plants. The water-cooled wall tubes in the boilers are eroded by flue gas, coal ash, and flames during long-term use. They are prone to wear and corrosion, resulting in local thinning of the tube wall and eventually boiler tube bursts. Once a boiler leaks, it cannot be handled during operation and often has to be shut down for repairs, resulting in huge economic losses. Therefore, all power plants attach great importance to effectively reducing and avoiding boiler tube bursts. Every time the boiler is shut down, a furnace scaffolding needs to be set up to inspect the water-cooled wall, which is a large workload and takes a long time. It is of great significance to be able to use robots to inspect the water-cooled wall without setting up a full furnace rack during a short-term shutdown.

[0003] The traditional manual inspection method has the following five problems. To address the following problems, a wall-climbing robot water-cooled wall defect detection solution is proposed to replace manual inspection. Defect recognition and detection based on machine vision is a research hotspot for industrial inspection. The automated, efficient and intelligent defect recognition and detection of water-cooled wall tubes of power plant boiler units is of great significance to reducing the non-stop time of boiler units and avoiding boiler safety accidents:

[0004] 1. The boiler is operated in a limited space. The dust in the furnace is large and the working environment is bad, which is harmful to the human body.

[0005] 2. The furnace needs to be cooled after the furnace is finished. Manual inspection requires a large workload, slow progress, and is time-consuming and labor-intensive.

[0006] 3. Problems such as high cost, high missed detection rate, and high risk factor:

[0007] 4. Affected by human factors, it is easy to miss inspections;

[0008] 5. Working at heights is very risky. Summary of the invention

[0009] In order to solve the above problems, the present invention proposes a water-cooled wall detection climbing robot and an obstacle surmounting method thereof.

[0010] A water-cooled wall inspection climbing robot comprises a body, a walking and obstacle-crossing mechanism arranged on the body for the body to walk and cross obstacles, a detection and positioning system for detecting defects and locating defect positions, a mechanical arm mechanism carrying the detection and positioning system for surface inspection, and a control and monitoring system for fully autonomous and semi-autonomous control of the wall-climbing robot.

[0011] The walking and obstacle-crossing mechanism comprises two groups and is distributed at the front and rear of the vehicle body. Any group of the walking and obstacle-crossing mechanism comprises a reduction motor arranged in the vehicle body and a magnetic wheel assembly matched with the reduction motor.

[0012] The magnetic wheel assembly includes a Y-shaped bracket and a magnetic wheel rotatably mounted on the three outer ends of the Y-shaped bracket. The combination of the magnetic wheel and the Y-shaped bracket forms a three-wheel structure. The magnetic wheel is connected to the output end of the reduction motor through a synchronous belt structure to achieve power transmission.

[0013] The mechanical arm mechanism comprises a six-axis mechanical arm and a mounting fixture arranged at the end of the six-axis mechanical arm.

[0014] The detection and positioning system includes a high-definition camera and an ultrasonic wave thickness meter installed at the end of a six-axis robotic arm, and a 3D laser radar installed on the vehicle body.

[0015] The control and monitoring system comprises a controller, an obstacle avoidance sensor connected to the controller, and a panoramic camera.

[0016] A method for overcoming obstacles of a water-cooled wall detection climbing robot, the specific steps of which are as follows:

[0017] S1. Place the wall-climbing robot into the starting position specified by the boiler;

[0018] S2, remotely start the robot's control and monitoring system;

[0019] S3, remotely control the robot to perform inspection along the designed route;

[0020] S4, check surface defects: the wall-climbing robot extends a high-definition camera to detect whether there are defects on the surface of the water-cooled wall. If the detection result shows that there are defects at the current position, the 3D laser radar is used to locate the current position and identify and record the type of damage. If there are no defects, the robot proceeds to step S5 to proceed to the next step.

[0021] S5. Check inner wall defects: The wall-climbing robot extends an ultrasonic wave thickness meter to detect whether there are defects in the water-cooled wall. If the detection result is yes, the 3D laser radar is used to locate the current position and identify and record the damage type. If not, the robot enters step S6 to proceed to the next step.

[0022] S6. The car continues to move along the planned path: if an obstacle is detected during the journey, the six-axis robotic arm will be automatically retracted to detect whether there is an obstacle in front of the car body. If there is an obstacle, the obstacle crossing process will be started. The unique three-wheel design of the walking and obstacle crossing mechanism will automatically switch between walking and obstacle crossing modes when encountering obstacles to keep the center of gravity stable;

[0023] S7, Obstacle Crossing: During the obstacle crossing process, there will be a situation where only two magnetic wheels touch the water-cooled wall. At this time, the magnetic wheels must provide sufficient adsorption force to ensure that the robot does not fall from a high altitude during the obstacle crossing process;

[0024] S8, the car continues to move and inspects the next water-cooled wall;

[0025] S9, judging whether all water-cooled walls have been inspected, if the inspection is completed, proceeding to step S10, if the inspection is not completed, returning to step S4 to continue the inspection;

[0026] S10, ending all defect detection processes, and the wall-climbing robot returns to the designated position.

[0027] The magnetic wheel in step S6 is selected to adopt a DC solenoid electromagnet, and the adsorption force of the magnetic wheel when the wall-climbing robot is working in a steady state is calculated according to the formula:

[0028] Where φ is the working air gap flux, Wb; B is the working air gap magnetic induction intensity, T; u0 is the vacuum permeability, its value is 4π×10 -7 Wb / A·m; S is the magnetic circuit cross-sectional area, m 2 .

[0029] The specific method for the high-definition camera 6 in step S4 to detect whether there are defects on the surface of the water-cooled wall is as follows: using the boiler water-cooled wall tube data collected by the wall-climbing robot, a boiler water-cooled wall tube defect classification and recognition data set and a boiler water-cooled wall tube target detection data set are constructed as training data, the PyTorch deep learning framework is used to train the ResNet101 defect image classification model and the YOLOV5 defect target detection model, LibTorch is used to deploy the algorithm model, the Tornado server framework is used to interact with the algorithm and the Web end, and the algorithm model classification, recognition and detection results are displayed on the Web end.

[0030] The specific method for the high-definition camera in step S4 to detect whether there are defects on the surface of the water-cooled wall also includes: in the structure adopted by the Web server, a request for classification recognition or target detection is issued on the Web side, and the algorithm server side processes data by calling the corresponding algorithm, returns the classification recognition and detection results, and performs data interaction; the Web-side boiler water-cooled wall tube defect recognition and detection system mainly includes two modules, namely a defect classification recognition module and a defect target detection module. The transfer learning method is used to train the boiler water-cooled wall tube image defect classification recognition model and the defect target detection model, and a boiler water-cooled wall defect detection and recognition model is constructed. The defect classification and recognition module adopts the ResNet101 algorithm model with the best classification performance, the highest accuracy, precision, and recall rate, and the defect target detection module adopts the YOLOV5 algorithm model with high detection accuracy and fast detection speed.

[0031] The beneficial effect of the present invention is that the present invention adopts a combination of ResNet101 and YOLOV5 to respectively perform defect classification detection and defect target recognition, which can quickly and accurately classify and locate boiler water-cooled wall defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0033] Figure 1 It is a schematic diagram of the three-dimensional structure of the present invention;

[0034] Figure 2 This is a diagram of the algorithm system architecture of step S4 of the present invention;

[0035] Figure 3 It is a schematic diagram of the process structure of the present invention;

[0036] Figure 4 It is a bottom view structural schematic diagram of the present invention;

[0037] Figure 5 It is a schematic structural diagram of the magnetic wheel assembly of the present invention;

[0038] Figure 6 The magnetic wheel scale meter of the present invention;

[0039] Figure numerals: 1. vehicle body; 2. reduction motor; 3. magnetic wheel assembly; 4. six-axis robotic arm; 5. mounting fixture; 6. high-definition camera; 7. ultrasonic wave thickness meter; 8. 3D laser radar; 9. controller; 10. obstacle avoidance sensor; 11. panoramic camera; 12. magnetic wheel; 13. Y-shaped bracket. DETAILED DESCRIPTION

[0040] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further described below.

[0041] like Figures 1 to 6 As shown, a water-cooled wall inspection climbing robot includes a body 1, a walking and obstacle-crossing mechanism arranged on the body 1 for the walking and obstacle-crossing functions of the body 1, a mechanical arm mechanism carrying visual inspection for surface inspection, a detection and positioning system for defect detection and positioning of defect positions, and a control and monitoring system for fully autonomous and semi-autonomous control of the equipment.

[0042] The robot can not only automatically switch between walking and obstacle crossing modes when encountering obstacles, maintain a stable center of gravity, and ensure that the wall-climbing robot can safely cross various types of obstacles that may be encountered during water-cooled wall defect detection. It can also realize automatic detection of boiler water-cooled wall tube defect targets through deep learning target detection algorithms, and has a good defect detection effect for water-cooled wall cracks, providing an automated, efficient and intelligent method for defect detection and identification of boiler water-cooled wall tubes.

[0043] The walking and obstacle-crossing mechanism includes two groups and is distributed at the front and rear of the vehicle body 1 . Any group of the walking and obstacle-crossing mechanism includes a reduction motor 2 arranged in the vehicle body 1 and a magnetic wheel assembly 3 matched with the reduction motor 2 .

[0044] The magnetic wheel assembly 3 includes a Y-shaped bracket 13 and a magnetic wheel 12 rotatably mounted on the three outer ends of the Y-shaped bracket 13. The combination of the magnetic wheel 12 and the Y-shaped bracket 13 forms a three-wheel structure. The magnetic wheel 12 is connected to the output end of the reduction motor 2 through a synchronous belt structure to achieve power transmission.

[0045] The robotic arm mechanism comprises a six-axis robotic arm 4 and a mounting fixture 5 arranged at the end of the six-axis robotic arm 4 .

[0046] The detection and positioning system includes a high-definition camera 6 and an ultrasonic wave thickness meter 7 installed at the end of a six-axis robotic arm 4 , and a 3D laser radar 8 installed on the vehicle body 1 .

[0047] The control and monitoring system includes a controller 9 , an obstacle avoidance sensor 10 connected to the controller 9 , and a panoramic camera 11 .

[0048] A method for overcoming obstacles of a water-cooled wall detection climbing robot, the specific steps of which are as follows:

[0049] S1. Place the wall-climbing robot into the starting position specified by the boiler;

[0050] S2, remotely start the robot teleoperation system;

[0051] S3, remotely control the robot to perform inspection along the designed route;

[0052] S4, check surface defects: the wall-climbing robot extends the high-definition camera 6 to detect whether there are defects on the surface of the water-cooled wall. If the detection result shows that there are defects at the current position, the 3D laser radar 8 is used to locate the current position and identify and record the type of damage. If there are no defects, the process proceeds to step S5 to continue the next step. The algorithm system architecture diagram is shown in the figure below. Figure 2 As shown;

[0053] S5, check inner wall defects: the wall-climbing robot extends the ultrasonic wave thickness meter 7 to detect whether there are defects in the water-cooled wall. If the detection result is yes, the 3D laser radar 8 is used to locate the current position and identify and record the damage type. If not, enter step S6 to continue the next step;

[0054] S6. The car continues to move along the planned path: if an obstacle is detected during the movement, the six-axis robot arm 4 is automatically retracted to detect whether there is an obstacle in front of the car body. If there is an obstacle, the obstacle crossing process is started. The unique three-wheel design of the walking and obstacle crossing mechanism automatically switches to the walking and obstacle crossing mode when encountering an obstacle, keeping the center of gravity stable, which can ensure that the wall-climbing robot can safely cross various types of obstacles that may be encountered during the water-cooled wall defect detection process;

[0055] S7. Obstacle crossing: The schematic diagram of the wall-climbing robot crossing obstacles is as follows Figure 3 As shown, it should be noted that during the obstacle crossing process, there will be a situation where only two magnetic wheels 12 contact the water-cooled wall. At this time, the magnetic wheels 12 must provide sufficient adsorption force to ensure that the robot will not fall from a high altitude during the obstacle crossing process. The obstacle crossing mechanism design of the wall-climbing robot is as follows Figure 4 If not, go to step S7;

[0056] S8, the car continues to move and inspects the next water-cooled wall;

[0057] S9, judging whether all water-cooled walls have been inspected, if the inspection is completed, proceeding to step S8, if the inspection is not completed, continuing the inspection in step S4;

[0058] S10, ending all defect detection processes, and the wall-climbing robot returns to the designated position.

[0059] In the step S4, Figure 2 As shown, the specific method for the high-definition camera 6 in step S4 to detect whether there are defects on the surface of the water-cooled wall is: using the boiler water-cooled wall tube data collected by the wall-climbing robot, a boiler water-cooled wall tube defect classification and recognition data set and a boiler water-cooled wall tube target detection data set are constructed as training data, the PyTorch deep learning framework is used to train the ResNet101 defect image classification model and the YOLOV5 defect target detection model, LibTorch is used to deploy the algorithm model, the Tornado server framework is used to interact with the algorithm and the Web end, and the algorithm model classification, recognition and detection results are displayed on the Web end.

[0060] The specific method for the high-definition camera 6 in step S4 to detect whether there are defects on the surface of the water-cooled wall also includes: in the structure adopted by the Web server, a request for classification recognition or target detection is issued on the Web side, and the algorithm server side processes data by calling the corresponding algorithm, returns the classification recognition and detection results, and performs data interaction; the Web-side boiler water-cooled wall tube defect recognition and detection system mainly includes two modules, namely a defect classification recognition module and a defect target detection module. The transfer learning method is used to train the boiler water-cooled wall tube image defect classification recognition model and the defect target detection model, and a boiler water-cooled wall defect detection and recognition model is constructed. The defect classification and recognition module adopts the ResNet101 algorithm model with the best classification performance, the highest accuracy, precision, and recall rate, and the defect target detection module adopts the YOLOV5 algorithm model with high detection accuracy and fast detection speed.

[0061] In the B / S structure adopted by the Web server, a request for classification recognition or target detection is issued on the Web side. The algorithm server side processes the data by calling the corresponding algorithm, returns the classification recognition and detection results, and conducts data interaction.

[0062] The web-based boiler water-cooled wall tube defect recognition and detection system mainly includes two modules, namely, a defect classification and recognition module and a defect target detection module. The present invention adopts the method of transfer learning to train the boiler water-cooled wall tube image defect classification and recognition model and the defect target detection model, and constructs a boiler water-cooled wall defect detection and recognition model. The defect classification and recognition module adopts the ResNet101 algorithm model with the best classification performance, the highest accuracy, precision and recall rate. The defect target detection module adopts the YOLOV5 algorithm model with high detection accuracy and fast detection speed.

[0063] The magnetic wheel 12 in step S6 is selected to adopt a DC solenoid electromagnet, and the adsorption force of the magnetic wheel 12 when the wall-climbing robot is working in a steady state is calculated according to the formula:

[0064] Where φ is the working air gap flux, Wb; B is the working air gap magnetic induction intensity, T; u0 is the vacuum permeability, its value is 4π×10 -7 Wb / A·m; S is the magnetic circuit cross-sectional area, m 2 ;

[0065] According to the formula S = (Dd)π·W / 2 and Figure 6 The magnetic wheel scale table can be used to calculate W, i.e., when the magnetic wheel radius is 50 mm, S, i.e., the magnetic circuit cross-sectional area, is 0.0175 m 2 ;

[0066] It is measured that when W is 50mm, B, i.e. the magnetic induction intensity of the working air gap, is 0.2T. According to the adsorption force calculation formula and repeated verification of the actual object, the adsorption force that a single magnetic wheel 12 can provide is 900N. Even in the process of surmounting obstacles, it can provide an adsorption force far greater than the weight of the wall-climbing robot itself to ensure that the robot will not fall from a high altitude.

[0067] The present invention adopts a combination of ResNet101 and YOLOV5 to perform defect classification detection and defect target recognition respectively, which can quickly and accurately classify and locate boiler water-cooled wall defects.

[0068] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A water-cooled wall inspection climbing robot, comprising a body (1), characterized in that: It also includes a walking and obstacle-crossing mechanism for the vehicle body (1) to walk and cross obstacles, a detection and positioning system for detecting defects and locating the defect positions, a mechanical arm mechanism carrying the detection and positioning system for surface detection, and a control and monitoring system for fully autonomous and semi-autonomous control of the wall-climbing robot.

2. A water-cooled wall inspection climbing robot according to claim 1, characterized in that: The walking and obstacle-crossing mechanism comprises two groups and is distributed at the front and rear of the vehicle body (1). Any group of the walking and obstacle-crossing mechanism comprises a reduction motor (2) arranged in the vehicle body (1) and a magnetic wheel assembly (3) matched with the reduction motor (2).

3. A water-cooled wall inspection climbing robot according to claim 2, characterized in that: The magnetic wheel assembly (3) comprises a Y-shaped bracket (13), a magnetic wheel (12) rotatably mounted on the three outer ends of the Y-shaped bracket (13), the combination of the magnetic wheel (12) and the Y-shaped bracket (13) forms a three-wheel structure, and the magnetic wheel (12) is connected to the output end of the reduction motor (2) through a synchronous belt structure.

4. The water-cooled wall inspection climbing robot according to claim 1, characterized in that: The mechanical arm mechanism comprises a six-axis mechanical arm (4) and a mounting fixture (5) arranged at the end of the six-axis mechanical arm (4).

5. The water-cooled wall inspection climbing robot according to claim 1, characterized in that: The detection and positioning system comprises a high-definition camera (6), an ultrasonic wave thickness meter (7) and a 3D laser radar (8).

6. The water-cooled wall inspection climbing robot according to claim 1, characterized in that: The control and monitoring system comprises a controller (9), an obstacle avoidance sensor (10) connected to the controller (9), and a panoramic camera (11).

7. A method for detecting an obstacle crossing of a wall-climbing robot using a water-cooled wall according to any one of claims 1 to 6, characterized in that: The specific steps are as follows: S1. Place the wall-climbing robot into the starting position specified by the boiler; S2, remotely start the robot's control and monitoring system; S3, remotely control the robot to perform inspection along the designed route; S4, check surface defects: the wall-climbing robot extends a high-definition camera (6) to detect whether there are defects on the surface of the water-cooled wall. If the detection result shows that there are defects at the current position, the 3D laser radar (8) is used to locate the current position and identify and record the type of damage. If there are no defects, the robot proceeds to step S5 to proceed to the next step. S5, check inner wall defects: the wall-climbing robot extends an ultrasonic wave thickness meter (7) to detect whether there are defects in the water-cooled wall. If the detection result is yes, the 3D laser radar (8) is used to locate the current position and identify and record the type of damage. If no, the robot enters step S6 to proceed to the next step; S6. The car continues to move along the planned path: if an obstacle is detected during the movement, the six-axis robot arm (4) is automatically retracted to detect whether there is an obstacle in front of the car body. If there is an obstacle, the obstacle crossing process is started. The unique three-wheel design of the walking and obstacle crossing mechanism automatically switches to the walking and obstacle crossing mode when encountering an obstacle to keep the center of gravity stable; S7, the car continues to move and inspects the next water-cooled wall; S8, judging whether all water-cooled walls have been inspected, if the inspection is completed, proceeding to step S10, if the inspection is not completed, returning to step S4 to continue the inspection; S9, all defect detection processes are completed and the wall-climbing robot returns to the specified position.

8. The obstacle surmounting method of a water-cooled wall detection climbing robot according to claim 7, characterized in that: The magnetic wheel (12) of step S6 is a DC solenoid electromagnet, and the adsorption force of the magnetic wheel (12) when the wall-climbing robot is working in a steady state is calculated according to the formula: Where φ is the working air gap flux, Wb; B is the working air gap magnetic induction intensity, T; u0 is the vacuum permeability, its value is 4π×10 -7 Wb / A·m; S is the magnetic circuit cross-sectional area, m 2 .

9. The obstacle surmounting method of a water-cooled wall inspection climbing robot according to claim 7, characterized in that: The specific method of using the high-definition camera (6) in step S4 to detect whether there are defects on the surface of the water-cooled wall is as follows: using the boiler water-cooled wall tube data collected by the wall-climbing robot, constructing a boiler water-cooled wall tube defect classification and recognition data set and a boiler water-cooled wall tube target detection data set as training data, using the PyTorch deep learning framework to train the ResNet101 defect image classification model and the YOLOV5 defect target detection model, using LibTorch to deploy the algorithm model, using the Tornado server framework to interact with the algorithm and the Web end, and displaying the algorithm model classification, recognition and detection results on the Web end.

10. The obstacle surmounting method of a water-cooled wall inspection climbing robot according to claim 9, characterized in that: The specific method for the high-definition camera (6) in step S4 to detect whether there are defects on the surface of the water-cooled wall also includes: in the structure adopted by the Web server, a request for classification recognition or target detection is issued on the Web end, and the algorithm server end processes data by calling the corresponding algorithm, returns the classification recognition and detection results, and performs data interaction; the Web-end boiler water-cooled wall tube defect recognition and detection system includes two modules, namely a defect classification recognition module and a defect target detection module, and a transfer learning method is used to train the boiler water-cooled wall tube image defect classification recognition model and the defect target detection model to construct a boiler water-cooled wall defect detection and recognition model, the defect classification recognition module adopts the ResNet101 algorithm model, and the defect target detection module adopts the YOLOV5 algorithm model.

Images