Multi-foot wall surface crawling welding robot, groove wall leakage point identification method of robot and storage medium

Abstract

The invention discloses a multi-foot wall surface crawling welding robot, a groove wall leakage point identification method of the robot and a storage medium, and relates to the technical field of industrial mechanical devices. Comprising a shell, a plurality of sets of mechanical leg mechanisms evenly arranged at the bottom of the shell, a mechanical arm with the tail end provided with a welding device or a grinding device in a replaceable mode, a visual perception module used for collecting wall face images and a master control host used for executing a preset groove wall leakage point recognition method. The mechanical arm is controlled to execute grinding or welding operation based on the recognition result of the groove wall leakage point, the specific situation of the groove wall leakage point is accurately recognized through the driving part by means of stable crawling of multiple sets of mechanical legs on the wall face, grinding and repair welding maintenance operation is conducted on the leakage part, and the robot integrates the whole-process operation capacity of inspection, recognition, grinding and welding. The operation mode of high-risk environment maintenance is fundamentally changed, 24-hour continuous operation can be achieved, the single-time maintenance operation time is greatly shortened, and the manual inspection time is effectively shortened.

Description

Technical Field

[0001] This invention relates to the field of industrial machinery and equipment technology, and in particular to a multi-legged wall-crawling welding robot, a method for identifying leaks in the robot's groove walls, and a storage medium. Background Technology

[0002] In the alumina production process in the chemical industry, tanks are core equipment that are exposed to a corrosive environment of high temperature, strong alkali (NaOH solution), and fine slurry for a long time. The tank walls (usually made of carbon steel Q345) are prone to cracks or perforations due to corrosion, wear, or stress concentration, which can lead to material leakage and affect production safety and efficiency.

[0003] Currently, high-altitude welding is generally carried out by welders erecting scaffolds or suspended platforms, which has problems such as high risk, low efficiency, and poor welding results. Therefore, developing an automated welding maintenance device that can adapt to the dangerous environment of high altitudes is of great significance to the current market. To this end, a multi-legged wall-climbing welding robot is proposed. Summary of the Invention

[0004] The main objective of this invention is to provide a multi-legged wall-crawling welding robot. The robot uses multiple sets of mechanical legs to crawl stably on the wall through a drive unit. It adopts a camera structure and a preset method for identifying leaks in the wall, which can accurately identify the specific situation of leaks in the wall. The robot arm welding device can grind and repair the leaking parts, effectively solving the problems in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A multi-legged wall-climbing welding robot includes: a robotic arm, a camera gimbal, a main control unit, welding rods, a welding rod storage cylinder, robotic leg structures, and a shell; A robotic arm is provided on one side of the housing. A main control host and a welding rod storage tube are respectively provided on the left side of the robotic arm. The camera gimbal is fixed above the main control host. The welding rods are stored in the welding rod storage tube. Openings are provided at the front and rear ends and the center of the lower side of the housing. Multiple sets of robotic leg structures are provided at the bottom of the openings. The main control unit is used to execute a preset method for identifying leaks in the tank wall, and controls the robotic arm to perform grinding or welding operations based on the identification results.

[0006] Furthermore, the robotic arm includes a welding device, a robotic arm base, a first-level robotic arm, a second-level robotic arm, a third-level robotic arm, a fourth-level robotic arm, a base drive motor, a first-level servo cylinder, a second-level servo cylinder, a third-level drive motor, a rotary sleeve, a right motor, a left motor, a front camera, a grinding wheel, and a rear camera. The robotic arm base is fixedly mounted on the housing. The bottom of the first-stage robotic arm and the bottom of the first-stage servo cylinder are respectively fixed to the front and rear ends of the upper circular plate of the robotic arm base. The upper part of the first-stage servo cylinder is fixedly connected to the middle support of the first-stage robotic arm, allowing the first-stage robotic arm to perform circular motion under the action of the first-stage servo cylinder. The middle part of the second-stage robotic arm is connected to the front end of the first-stage robotic arm, and the end of the second-stage robotic arm is connected to the upper end of the second-stage servo cylinder. The end of the second-stage servo cylinder is connected to the bottom of the first-stage robotic arm, allowing the second-stage robotic arm to perform circular motion under the action of the second-stage servo cylinder. A front camera is fixed on the top of the second-stage robotic arm. The third-stage robotic arm... A rotating sleeve is connected to the front end of the secondary robotic arm, and a tertiary drive motor is fixed inside the secondary robotic arm and connected to the rotating sleeve, so that the tertiary robotic arm rotates under the action of the tertiary drive motor. The end of the tertiary robotic arm is connected to the front end of the tertiary robotic arm, and a right motor and a left motor are fixed inside the tertiary robotic arm, so that the tertiary robotic arm performs circular motion under the action of the right motor and the left motor. A rear camera is fixed to the outside of the tertiary robotic arm, and a welding device is installed at the front end, so that the welding device can move under the action of the right motor and the left motor. The base drive motor is fixed on the base of the robotic arm, so that the robotic arm rotates under the drive of the base drive motor.

[0007] Furthermore, the mechanical leg mechanism includes a mechanical leg fixed base plate, a universal joint, an X-axis tie rod drive motor, a Y-axis tie rod drive motor, an X-axis tie rod, a Y-axis tie rod, an X-axis drive push rod, a Y-axis drive push rod, a servo electric cylinder, a tie rod fixing ring, a mechanical leg foot plate, an electromagnet, a laser rangefinder, a drive motor fixing bracket, and a nine-axis attitude sensor; The mechanical leg fixing base plate is fixed to the bottom of the housing. Drive motor fixing brackets are fixed on both sides of the bottom of the mechanical leg fixing base plate. A servo electric cylinder is installed at the bottom of the mechanical leg fixing base plate between the drive motor fixing brackets. A nine-axis attitude sensor is installed on the upper part of the servo electric cylinder. The top of the servo electric cylinder is connected to the bottom of the mechanical leg fixing base plate through a universal joint. The movable end of the bottom of the servo electric cylinder is fixedly connected to an electromagnet through a mechanical foot plate. A laser rangefinder is fixed on the mechanical leg foot plate.

[0008] The X-axis tie rod drive motor and the Y-axis tie rod drive motor are respectively fixed on the drive motor mounting bracket. The drive shafts of the X-axis tie rod drive motor and the Y-axis tie rod drive motor are respectively connected to one end of the X-axis tie rod and the Y-axis tie rod through the X-axis drive push rod and the Y-axis drive push rod, respectively. The other end of the X-axis tie rod and the Y-axis tie rod are respectively connected to two adjacent side walls of the tie rod fixing ring. The tie rod fixing ring is sleeved and fixed to the cylinder body of the servo electric cylinder.

[0009] Furthermore, the rear camera is fitted onto the upper part of the fourth-level robotic arm via a fixed bracket, and the front camera is fitted onto the upper part of the second-level robotic arm via a fixed bracket.

[0010] Furthermore, the camera gimbal, welding device, base drive motor, first-stage servo cylinder, second-stage servo cylinder, third-stage drive motor, right motor, left motor, front camera, rear camera, X-axis tie rod drive motor, Y-axis tie rod drive motor, servo cylinder, electromagnet, laser rangefinder, and nine-axis attitude sensor are all connected to the main control host. The camera gimbal, front camera, and rear camera are connected to the welding device host via wires, enabling them to transmit real-time captured images and photos.

[0011] Furthermore, the robot performs a self-check process upon startup, which includes a vertical angle correction process, an extension movement of the mechanical leg mechanism, a forward swinging process of the mechanical leg mechanism, a backward swinging process of the mechanical leg mechanism, a leftward swinging process of the mechanical leg mechanism, a rightward swinging process of the mechanical leg mechanism, a forward and backward movement process of the robot, and a left and right movement process of the robot; specifically: The mechanical leg mechanisms 6-1 and 6-2 form one group, the mechanical leg mechanisms 6-4 and 6-5 form another group, and the mechanical leg structure 6-3 forms an independent group. The main control host identifies the angle between the direction of the cylinder centerline of the servo electric cylinder and the fixed base plate of the mechanical leg based on the nine-axis attitude sensor on any mechanical leg mechanism. Based on the identification results, the vertical angle of the mechanical leg mechanism is corrected in sequence to ensure that the angle between the direction of the cylinder centerline of the servo electric cylinder on any group of mechanical leg mechanisms and the fixed base plate of the mechanical leg is 90°, thereby adjusting the robot's posture to the optimal state. The mechanical leg mechanism performs an extension motion as follows: During extension, the internal lead screw of the servo electric cylinder rotates, causing the piston cylinder at the tail of the electric cylinder to move downward. The piston cylinder pushes out the part of the mechanical leg below the foot plate, thus achieving extension.

[0012] The forward swinging process of the mechanical leg mechanism is as follows: The X-axis and Y-axis tie rod drive motors are controlled to drive the X-axis and Y-axis tie rods respectively, which in turn drive the servo cylinder to move through the tie rod fixing ring. This causes the central axis of the servo cylinder of the mechanical leg mechanism to form a certain angle with the mechanical leg fixing base plate, thereby achieving movement. Specifically: When moving forward, the X-axis drive push rod rotates clockwise around the transmission shaft of the X-axis pull rod drive motor, which increases the angle between the X-axis drive push rod and the X-axis pull rod. Meanwhile, the Y-axis drive push rod rotates clockwise around the transmission shaft of the Y-axis pull rod drive motor, which decreases the angle between the Y-axis drive push rod and the Y-axis pull rod. Both the X-axis pull rod and the Y-axis pull rod are connected to the pull rod fixing ring, which is fixed to the servo electric cylinder, thus causing the corresponding mechanical leg mechanism to move forward. The backward swinging process of the mechanical leg mechanism is as follows: When moving backward, the X-axis drive push rod rotates counterclockwise around the transmission shaft of the X-axis tie rod drive motor, making the angle between the X-axis drive push rod and the X-axis tie rod smaller. Meanwhile, the Y-axis drive push rod rotates counterclockwise around the transmission shaft of the Y-axis tie rod drive motor, making the angle between the Y-axis drive push rod and the Y-axis tie rod larger. Both the X-axis tie rod and the Y-axis tie rod are connected to the tie rod fixing ring, which is fixed to the servo electric cylinder, thus causing the corresponding mechanical leg mechanism to move backward. The mechanical leg mechanism swings to the left as follows: The X-axis and Y-axis tie rod drive motors are controlled to move the X-axis and Y-axis tie rods respectively. This, in turn, drives the servo cylinder to move to the left via the tie rod fixing ring. This causes the central axis of the servo cylinder of the mechanical leg mechanism to form a certain angle with the mechanical leg's fixed base plate, thus achieving the movement. Specifically: When moving to the left, the Y-axis drive push rod is driven by the Y-axis pull rod drive motor 66 and rotates counterclockwise around its transmission shaft, which increases the angle between the Y-axis drive push rod and the Y-axis pull rod. Meanwhile, the X-axis drive push rod is driven by the X-axis pull rod drive motor and rotates counterclockwise, which decreases the angle between the X-axis drive push rod and the X-axis pull rod. Both the X-axis pull rod and the Y-axis pull rod are connected to the pull rod fixing ring, which is fixed to the servo electric cylinder, thus causing the corresponding mechanical leg mechanism to move to the left. The mechanical leg mechanism swings to the right as follows: When moving to the right, the Y-axis drive push rod is driven by the Y-axis pull rod drive motor to rotate clockwise around its transmission shaft, which makes the angle between the Y-axis drive push rod and the Y-axis pull rod smaller. Meanwhile, the X-axis drive push rod is driven by the X-axis pull rod drive motor to rotate counterclockwise, which makes the angle between the X-axis drive push rod and the X-axis pull rod larger. Both the X-axis pull rod and the Y-axis pull rod are connected to the pull rod fixing ring, which is fixed to the servo electric cylinder, thus causing the corresponding mechanical leg mechanism to move to the right.

[0013] The robot's forward and backward movements are as follows: All the mechanical leg mechanisms swing backward, and the robot moves forward a short distance. Mechanical leg mechanism 6-3 extends vertically and swings to the left, mechanical leg mechanisms 6-1 and 6-2 extend vertically and swing forward, mechanical leg mechanism 6-3 extends vertically and swings to the right, and mechanical leg mechanisms 6-4 and 6-5 extend vertically and swing forward. During the movement of the mechanical leg mechanisms, the electromagnets are started and stopped accordingly. The robot performs forward and backward movements in this manner.

[0014] The robot's left and right movements are as follows: When all the mechanical leg mechanisms swing to the right, the robot body moves to the left a short distance. Mechanical leg mechanism 6-3 extends vertically and swings to the left, mechanical leg mechanisms 6-1 and 6-2 extend vertically and swing to the left, mechanical leg mechanism 6-3 extends vertically and swings to the right, and mechanical leg mechanisms 6-4 and 6-5 extend vertically and swing to the left. During the movement of the mechanical leg mechanisms, the electromagnets are started and stopped accordingly. The robot performs left and right movements in this manner.

[0015] A method for identifying leaks in the wall of a multi-legged wall-crawling welding robot includes the following steps: S1: Obtain a multi-view image sequence of the wall surface through the visual perception module; S2: Perform multi-scale Gaussian pyramid decomposition on the acquired image to obtain image subbands of different scales; S3: A set of biomimetic compound eye filters with different directions are used to convolve the image sub-bands at each scale to generate a multi-scale, multi-directional visual response map. S4: Fuse the multi-scale, multi-directional visual response maps to generate a visual saliency map;

[0016] Fusion visual saliency map The calculation method is as follows: ,in Indicated as from camera The scale is , direction is Visual response diagram; This is for the operation of retrieving the maximum value; For scale Weighting coefficients; This represents the total number of scales.

[0017] S5: Based on the visual saliency map, identify and locate the leak area on the tank wall.

[0018] The leak detection method also includes the following steps:

[0019] The local curvature of the wall surface measured by the laser rangefinder 613 is obtained. and roughness ;

[0020] Based on the obtained local curvature and roughness Generate tactile aid map The calculation method is as follows: ,in, For the first The projected coordinates of the sensor in the image; The average roughness of the wall surface; , All are weighting coefficients; For the number of sensors; It is an exponential function;

[0021] The visual saliency map With the tactile aid diagram Weighted fusion is performed to obtain the final multimodal fusion feature map.

[0022] The leak detection method also includes the following steps:

[0023] For any candidate leak region extracted from the multimodal fusion feature map Define its time The dynamic state vector on is The dynamic state vector It should include at least the confidence component, the size component, and the stability component;

[0024] The dynamic state vector is obtained through evolutionary equations. In time The evolutionary process described above is expressed by the following equation: ,in, For time Image features extracted from frames; , Both are weight matrices; Represented as a region of the image With the region Interaction weights between them; , All are nonlinear functions;

[0025] Based on the calculation results of the evolution equation, candidate leak point regions are determined according to the following rules:

[0026] When the confidence component continuously exceeds a preset threshold and the stability component meets preset requirements, the candidate leak area is determined to be a confirmed leak.

[0027] The robot's operation steps are as follows: Step 1: As the robot crawls forward on the wall, the camera pan-tilt unit captures images of the wall as the robot moves.

[0028] Step 2: During the camera pan-tilt-zoom (PTZ) shooting and recognition process, the main control unit uses a preset slot wall leak point recognition method to analyze the captured image and further determine the candidate leak point areas in the image.

[0029] Step 3: When the candidate leak area is determined to be a confirmed leak, the main control host sends a command to the first-level servo cylinder, the second-level servo cylinder, and the base drive motor, so that they drive the grinding wheel to the leak area, control the grinding wheel to align with the leak, and drive the third-level drive motor to rotate the grinding wheel to grind the leak area.

[0030] Step 4: During the polishing process, the visual perception module acquires multi-view image sequences of the wall surface and obtains the local curvature and roughness of the polishing area in real time. When the local curvature and roughness reach the preset threshold, the polishing is completed.

[0031] Step 5: The main control unit controls the robotic arm to retract and the robotic leg mechanism to move from top to bottom. The third-level robotic arm and the fourth-level robotic arm are installed manually. Then, Step 1 is repeated to move the robotic arm up to the target position.

[0032] Step 6: After moving to the target position, the main control host controls the first-level servo cylinder, the second-level servo cylinder, the third-level drive motor, the base drive motor, the right motor, the left motor, the rear camera, and the camera gimbal to start. At the same time, the front camera is turned off. The camera gimbal rotates 360° to identify the welding rod, controls the base drive motor to rotate the robotic arm, and uses the welding device to clamp the welding rod, completing the clamping of the welding rod.

[0033] Step 7: After the welding rod is clamped, the main control host controls the base drive motor to rotate the robotic arm and apply the welding rod to the leak area, and then perform the repair welding operation.

[0034] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any embodiment of the method for identifying leaks in the groove wall of a multi-legged wall-crawling welding robot, and generates control instructions based on the identification results for controlling the robotic arm of the multi-legged wall-crawling welding robot to perform grinding or welding operations.

[0035] The present invention has the following beneficial effects: Compared with existing technologies, this solution enables the robot to climb stably on vertical and inclined walls by using multiple sets of mechanical legs with electromagnetic adsorption and multi-degree-of-freedom swing, completely replacing manual scaffolding or hanging baskets. This fundamentally changes the operation mode of high-risk environment maintenance. Furthermore, the multi-level joint design of the robotic arm and the quick-change function of the end tool enable a single robot to integrate the entire process of inspection, identification, grinding, and welding.

[0036] Compared with existing technologies, the proposed method for identifying leaks in the trench wall by integrating vision (multiple cameras) and tactile sensation (laser ranging) can effectively identify defects such as microcracks, perforations, and corrosion points. The identification accuracy is higher than that of human eye observation, reducing missed detections and achieving millimeter-level positioning of the three-dimensional coordinates of leaks, thus providing a foundation for subsequent precise operations.

[0037] Compared with existing technologies, this solution ensures that the robot maintains stable adhesion and movement on uneven walls with welds by setting up posture self-correction of the mechanical leg mechanism and feedback from nine-axis sensors. At the same time, the grouped collaborative control strategy of the mechanical leg mechanism (such as alternating movement of the front, middle and rear groups) can ensure that there are at least three points of adhesion during the movement, resulting in high safety against falls.

[0038] Compared with existing technologies, the robot proposed in this solution can operate continuously for 24 hours without the need to build / dismantle scaffolding. The time for a single maintenance operation is expected to be reduced by more than 60%. Through automatic identification and positioning, it effectively reduces manual inspection time. Automated grinding and welding ensures consistent operation, high first-pass yield, and reduced rework. The robot can also accurately detect and repair micro-defects in tanks at an early stage, preventing them from developing into large-scale leaks or downtime accidents. This results in a double reduction in maintenance costs and production losses. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the overall structure of the multi-legged wall-crawling welding robot of the present invention;

[0040] Figure 2 This is a schematic diagram of the structure of the robotic arm (welding) of the present invention;

[0041] Figure 3 This is a schematic diagram of the structure of the robotic arm (grinding) of the present invention;

[0042] Figure 4 This is a schematic diagram of the mechanical leg structure of the present invention;

[0043] Figure 5 This is a flowchart illustrating the method for identifying leaks in the tank wall proposed in this invention.

[0044] In the picture,

[0045] 1. Robotic arm;

[0046] 11. Welding device; 12. Robotic arm base; 13. First-stage robotic arm; 14. Second-stage robotic arm; 15. Third-stage robotic arm; 16. Fourth-stage robotic arm; 17. Base drive motor; 18. First-stage servo cylinder; 19. Second-stage servo cylinder; 110. Third-stage drive motor; 111. Rotary sleeve; 112. Right motor; 113. Left motor; 114. Front camera; 115. Grinding wheel; 116. Rear camera; 2. Camera pan / tilt head; 3. Main control unit; 4. Welding rod; 5. Welding rod 6. Storage cylinder; 6. Mechanical leg structure; 61. Mechanical leg fixing base plate; 62. Universal joint; 63. X-axis tie rod drive motor; 64. Y-axis tie rod drive motor; 65. X-axis tie rod; 66. Y-axis tie rod; 67. X-axis drive push rod; 68. Y-axis drive push rod; 69. Servo electric cylinder; 610. Tie rod fixing ring; 611. Mechanical leg foot plate; 612. Electromagnet; 613. Laser rangefinder sensor; 614. Drive motor fixing bracket; 615. Nine-axis attitude sensor; 7. Housing. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] See Figures 1-5 The specific implementation steps of the multi-legged wall-crawling welding robot proposed in this invention are as follows:

[0049] Phase 1: Robot Deployment and Initialization Phase

[0050] Step S101: Environmental Survey and System Placement

[0051] Operators remotely or on-site survey the environment of the tank wall to be maintained and confirm the work area;

[0052] Transport the multi-legged wall-crawling welding robot to the bottom of the tank or near the predetermined starting point;

[0053] Check the integrity of each component of the robot, including the mechanical leg mechanism 6, the mechanical arm 1, the vision perception module and the main control host 3;

[0054] Connect the robot to an external power source or confirm that the built-in power source is sufficient, and establish a wireless communication link with the ground control station.

[0055] Step S102: Robot Start-up and Self-Test

[0056] Start the main control host 3, and the system loads the control program and the tank wall leak detection algorithm.

[0057] The robot performs a self-test procedure:

[0058] a. Vertical angle correction: The main control host 3 reads the data from the nine-axis attitude sensor 615 on each mechanical leg mechanism 6 and calculates the angle between the cylinder centerline of the servo electric cylinder 69 on each mechanical leg and the mechanical leg fixed base plate 61;

[0059] b. Adjust the extension and retraction of the servo electric cylinder 69 and the X / Y axis drive assembly of each mechanical leg mechanism 6 in sequence so that the above-mentioned included angle of each mechanical leg is corrected to 90°, ensuring that the robot's initial posture is perpendicular to the wall.

[0060] c. The self-test procedure sequentially verifies the extension / swing of the robotic leg, the movement of each joint of the robotic arm, the function of the camera pan-tilt unit 2 and the front / rear cameras, the attraction / release of the electromagnet, and whether the communication of each sensor is normal.

[0061] After passing the self-check, the robot enters the "ready" state.

[0062] Step S103: Wall Adhesion and Attitude Confirmation

[0063] The main control unit 3 controls the electromagnets 612 at the ends of all mechanical leg mechanisms 6 to be energized, so that the robot can be firmly attached to the wall at the starting point.

[0064] By fusing data from the nine-axis attitude sensor 615 and the laser rangefinder 613, the overall posture of the robot was confirmed to be stable and the force on each mechanical leg was uniform.

[0065] Phase 2: Autonomous Inspection and Leak Identification Phase

[0066] Step S201: Plan the inspection route and begin moving.

[0067] Operators can set the inspection area (such as a specified height range or a ring area) or select the automatic full inspection mode through the ground control station.

[0068] According to the path planning, the main control host 3 controls the mechanical leg mechanism 6 to perform coordinated gait movements (such as the "forward and backward movement" or "left and right movement" process mentioned above), driving the robot to start crawling on the wall.

[0069] Step S202: Real-time acquisition of multimodal data

[0070] During the movement, the camera pan-tilt 2 performs a wide-area scan, and the front camera 114 and the rear camera 116 simultaneously acquire high-definition image sequences of the wall from multiple perspectives and focal lengths.

[0071] The laser rangefinder 613 of each robotic leg collects local distance information of the wall surface in real time and calculates local curvature and roughness.

[0072] Step S203: Execute the tank wall leak detection algorithm

[0073] The main control host 3 calls and executes the preset tank wall leak detection method. The specific process is as follows:

[0074] The acquired images are decomposed into multi-scale Gaussian pyramids and then convolved in multiple directions using a biomimetic compound eye filter bank to generate multi-scale, multi-directional visual response maps. ;

[0075] According to the formula: By fusing multi-source visual response maps, a preliminary visual saliency map is generated. ;

[0076] Read the distance data acquired by the laser rangefinder 613, according to the formula: Generate tactile aid map ;

[0077] Visual salience map With tactile aids Weighted fusion is performed to obtain the final multimodal fusion feature map;

[0078] Candidate leak regions are extracted from the multimodal fusion feature map, and for each candidate region... Establish and maintain dynamic state vectors The dynamic state vector includes at least a confidence component, a size component, and a stability component.

[0079] According to the evolution equation: The dynamic state vector is updated by combining data from multiple consecutive frames.

[0080] Based on the calculation results of the evolution equation, candidate leak point regions are determined. The determination rule is: when the confidence component continuously exceeds the preset threshold and the stability component meets the preset requirements, the candidate leak point region is determined to be a confirmed leak point.

[0081] Record the precise image coordinates, estimated 3D spatial location (combining binocular vision or sensor fusion), size, shape features, and confidence level of each confirmed leak point to generate a leak list.

[0082] Phase 3: Autonomous Polishing Operation Phase

[0083] Step S301: Locating and Approaching the Target Leakage Point

[0084] The main control host 3 plans the optimal operation sequence (e.g., from near to far) based on the leak list.

[0085] Control the robot to move to the optimal working position near the first target leak point.

[0086] Adjust the robot's overall posture to ensure that the workspace of robotic arm 1 can cover the leak.

[0087] Step S302: Robotic arm positioning and grinding tool placement

[0088] The main control unit 3 controls the movement of the robotic arm 1:

[0089] a. Start the base drive motor 17 to rotate the robotic arm base 12, roughly aligning it with the direction of the leak point;

[0090] b. Coordinate and control the first-level servo cylinder 18 and the second-level servo cylinder 19 to drive the first-level robotic arm 13 and the second-level robotic arm 14 to move and aim the front camera 114 at the leak area for precise positioning.

[0091] c. Based on the precise positioning information, control the robotic arm to move further and transport the grinding wheel 115 installed at the front end of the fourth-level robotic arm 16 to the predetermined position directly above the leak point.

[0092] Step S303: Perform automatic polishing

[0093] The main control unit 3 starts the three-stage drive motor 110, which drives the grinding wheel 115 to rotate at high speed.

[0094] Control the robotic arm to perform fine movements so that the grinding wheel 115 contacts and grinds the leak area.

[0095] Grinding process monitoring: The grinding status is monitored in real time by the front camera 114, while the laser range sensor 613 monitors the changes in depth and surface roughness of the grinding area.

[0096] When the monitoring data (such as the grinding depth reaching the predetermined value and the surface roughness meeting the welding requirements) meet the preset grinding completion standards, the grinding operation is stopped.

[0097] The robotic arm retracts, and the grinding wheel 115 stops rotating.

[0098] Step S304: Grinding quality inspection and recording

[0099] Control the front camera 114 to take high-definition pictures of the polished area.

[0100] The main control unit 3 analyzes the captured images to check whether the polished area is uniform and whether there are any residual defects.

[0101] Record the images and data after polishing, and update the repair status of the leak to "polished".

[0102] Phase 4: Autonomous Welding Operation Phase

[0103] Step S401: Welding preparation and tool change

[0104] The main control unit 3 controls the robotic arm 1 to fully retract to a safe position.

[0105] If the end effector is a quick-change device, the grinding tool can be automatically replaced with the welding device 11. After remote confirmation by the operator, the robot moves to a position that is convenient for manual installation of the three-level robotic arm 15, the four-level robotic arm 16 and the welding device 11, and the replacement is completed manually.

[0106] After the replacement is completed, the robot re-attaches to the wall and adjusts its posture.

[0107] Step S402: Electrode clamping and positioning

[0108] The main control unit 3 starts the camera pan-tilt unit 2 to rotate 360° and identify the welding rod 4 in the welding rod storage cylinder 5.

[0109] Control the movement of the robotic arm 1 to accurately pick up a welding rod 4 through the welding device 11.

[0110] Based on the previously recorded precise location of the leak after grinding, control the robotic arm 1 to align the end of the welding rod 4 with the part to be welded.

[0111] Step S403: Perform automatic repair welding

[0112] Start the welding device 11 and execute the preset welding process parameters (current, voltage, wire feed speed, etc.).

[0113] Control the robotic arm 1 to move along a predetermined trajectory and speed to perform repair welding operations.

[0114] Welding process monitoring: The state of the molten pool is observed in real time through the rear camera 116 to ensure welding quality.

[0115] Step S404: Post-weld treatment and inspection

[0116] After welding is completed, welding device 11 stops working.

[0117] Control the robotic arm 1 to move away and take a picture of the weld.

[0118] The main control unit 3 can perform preliminary analysis of the weld image (such as checking continuity and whether there are obvious defects) and record the post-weld image.

[0119] Phase 5: Task Cycle and Assignment Completion Phase

[0120] Step S501: Task Loop

[0121] Return to step S301 and perform grinding and welding operations on the next leak in the list.

[0122] Repeat this cycle until all omissions in the list have been addressed.

[0123] Step S502: Task completion and robot retrieval

[0124] After all leaks have been repaired, the main control unit 3 generates a work report, summarizing the identification, grinding, welding records and process data of all leaks.

[0125] Control the robot to return to the starting point or designated recovery location along a safe path.

[0126] Release the electromagnets 612 of each mechanical leg in sequence, and the operator or auxiliary equipment will remove the robot from the wall.

[0127] Turn off the robot's power and perform necessary cleaning and maintenance.

[0128] Phase 6: Remote Monitoring and Manual Intervention

[0129] Throughout the implementation process, operators can monitor the robot's status in real time through the ground control station, view all camera feeds, receive algorithm recognition results and alarm information (such as abnormal posture, insufficient adhesion, low recognition confidence, etc.), and intervene when necessary to remotely control, correct operating parameters, or issue emergency stop commands.

[0130] The foregoing has shown and described 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 embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A multi-legged wall-crawling welding robot, characterized in that, include: Shell (7); Multiple sets of mechanical leg mechanisms (6) are evenly spaced at the bottom of the housing (7) to attract and drive the robot to move on the wall surface; A robotic arm (1) is located on one side of the housing (7), and its end is replaceably equipped with a welding device (11) or a grinding device; The visual perception module includes a camera gimbal (2) mounted on the housing (7) and a front camera (114) and a rear camera (116) mounted on the robotic arm (1) for acquiring wall images; The main control host (3) is located inside the housing (7) and is electrically connected to the mechanical leg mechanism (6), the mechanical arm (1) and the vision perception module; The main control host (3) is used to execute a preset method for identifying leaks in the tank wall and to control the robotic arm (1) to perform grinding or welding operations based on the identification results.

2. The multi-legged wall-crawling welding robot according to claim 1, characterized in that, The mechanical leg mechanism (6) includes: The mechanical leg fixing base plate (61) is fixed to the bottom of the housing (7); A servo electric cylinder (69) is connected at its top to the mechanical leg mounting base plate (61) via a universal joint (62); The mechanical leg foot plate (611) is fixed to the movable end of the bottom of the servo electric cylinder (69), and an electromagnet (612) for adsorbing the wall surface is provided on it. The X-axis drive assembly and the Y-axis drive assembly include X / Y axis tie rod drive motors (63, 64), X / Y axis drive push rods (67, 68), and X / Y axis tie rods (65, 66), respectively. A tie rod fixing ring (610) is sleeved and fixed to the outer wall of the cylinder body of the servo electric cylinder (69). One end of the X-axis tie rod (65) and the Y-axis tie rod (66) are respectively hinged to the tie rod fixing ring (610). One end of the X / Y axis drive push rod (67, 68) is connected to the drive shaft of the corresponding drive motor, and the other end is hinged to the corresponding pull rod. By the forward and reverse rotation of the drive motor, the servo electric cylinder (69) is driven to swing around the universal joint (62) in the X and Y axis directions.

3. The multi-legged wall-crawling welding robot according to claim 1, characterized in that, The robotic arm (1) includes: The robotic arm base (12) is fixed to the housing (7); The first-level robotic arm (13), the second-level robotic arm (14), the third-level robotic arm (15), and the fourth-level robotic arm (16) are hinged in sequence. The drive assembly includes servo electric cylinders (18, 19) for driving the first-level robotic arm (13) and the second-level robotic arm (14) to swing, a drive motor (110) for driving the third-level robotic arm (15) to rotate, and motors (112, 113) for driving the fourth-level robotic arm (16) to swing. The front camera (114) is mounted on the secondary robotic arm (14), and the rear camera (116) and the welding device (11) or grinding device are mounted on the fourth robotic arm (16).

4. The multi-legged wall-crawling welding robot according to claim 2, characterized in that, The mechanical leg mechanism (6) also includes a nine-axis attitude sensor (615) and a laser rangefinder (613); the nine-axis attitude sensor (615) is mounted on the servo cylinder (69) and is used to measure the tilt angle of the mechanical leg; the laser rangefinder (613) is mounted on the foot plate (611) of the mechanical leg and is used to measure the local distance and roughness information of the wall surface.

5. The multi-legged wall-crawling welding robot according to any one of claims 1-4, characterized in that, The robot performs a self-check process upon startup, which includes a vertical angle correction process, an extension movement of the mechanical leg mechanism (6), a forward swinging process of the mechanical leg mechanism (6), a backward swinging process of the mechanical leg mechanism (6), a left swinging process of the mechanical leg mechanism (6), a right swinging process of the mechanical leg mechanism (6), a forward and backward movement process of the robot, and a left and right movement process of the robot. The vertical angle correction process is as follows: the main control host (3) identifies the angle between the direction of the cylinder centerline of the servo electric cylinder (69) and the mechanical leg fixing base plate (61) based on the nine-axis attitude sensor (615) on any mechanical leg mechanism (6), and performs vertical angle correction on the mechanical leg mechanism (6) in sequence according to the identification result, so as to ensure that the angle between the direction of the cylinder centerline of the servo electric cylinder (69) on any mechanical leg mechanism (6) and the mechanical leg fixing base plate (61) is 90°.

6. A method for identifying leak points in the groove wall of a multi-legged wall-crawling welding robot according to any one of claims 1-5, characterized in that, Includes the following steps: S1: Obtain a multi-view image sequence of the wall surface through the visual perception module; S2: Perform multi-scale Gaussian pyramid decomposition on the acquired image to obtain image subbands of different scales; S3: A set of biomimetic compound eye filters with different directions are used to convolve the image sub-bands at each scale to generate a multi-scale, multi-directional visual response map. S4: Fuse the multi-scale, multi-directional visual response maps to generate a visual saliency map; S5: Based on the visual saliency map, identify and locate the leak area on the tank wall.

7. The method for identifying leaks in the tank wall according to claim 6, characterized in that, In step S4, the fused visual saliency map The calculation method is as follows: ,in Indicated as from camera The scale is , direction is Visual response diagram; This is for the operation of retrieving the maximum value; For scale Weighting coefficients; This represents the total number of scales.

8. The method for identifying leaks in the tank wall according to claim 6, characterized in that, It also includes the following steps: The local curvature of the wall surface measured by the laser rangefinder (613) is obtained. and roughness ; Based on the obtained local curvature and roughness Generate tactile aid map The calculation method is as follows: ,in, For the first The projected coordinates of the sensor in the image; The average roughness of the wall surface; , All are weighting coefficients; For the number of sensors; It is an exponential function; The visual saliency map With the tactile aid diagram Weighted fusion is performed to obtain the final multimodal fusion feature map.

9. The method for identifying leaks in the tank wall according to claim 8, characterized in that, It also includes the following steps: For any candidate leak region extracted from the multimodal fusion feature map Define its time The dynamic state vector on is The dynamic state vector It should include at least the confidence component, the size component, and the stability component; The dynamic state vector is obtained through evolutionary equations. In time The evolutionary process described above is expressed by the following equation: ,in, For time Image features extracted from frames; , Both are weight matrices; Represented as a region of the image With the region Interaction weights between them; , All are nonlinear functions; Based on the calculation results of the evolution equation, candidate leak point regions are determined according to the following rules: When the confidence component continuously exceeds a preset threshold and the stability component meets preset requirements, the candidate leak area is determined to be a confirmed leak.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for identifying leaks in the tank wall according to any one of claims 6 to 9, and generates control instructions based on the identification results for controlling the robotic arm of the multi-legged wall-crawling welding robot to perform grinding or welding operations.

Images