A flying robot system and detection method for pipeline facility contact detection

Abstract

This invention discloses a flying robot system and inspection method for contact inspection of pipeline facilities, including a drone flight platform, a pipeline crawling inspection body, a precision inspection robotic arm, and a smart terminal. The drone flight platform carries the pipeline crawling inspection body and moves it above the pipeline to be inspected. The pipeline crawling inspection body includes a circumferential clamping structure and a preliminary inspection mechanism. The circumferential clamping structure adaptively clamps the pipeline surface to achieve stable attachment. The preliminary inspection mechanism performs a large-area rapid scan of the pipeline's outer wall to locate suspected damaged areas. The precision inspection robotic arm, under the control of the smart terminal, performs contact-based precision inspection of the suspected areas. This invention combines flight maneuverability with crawling stability, enabling rapid deployment and stable inspection of pipelines in complex environments. Through a two-level inspection mode of preliminary screening and precise re-inspection, it balances inspection efficiency and accuracy, achieving full automation from defect discovery to diagnosis.

Description

Technical Field

[0001] This invention relates to the field of pipeline inspection and maintenance technology, and in particular to a flying robot system for contact inspection of pipeline facilities. Background Technology

[0002] As a crucial component of energy transmission, pipelines operate in complex environments for extended periods, making them susceptible to structural damage such as corrosion, cracks, and thinning. Failure to detect these damages promptly can lead to safety incidents like leaks and explosions. However, pipelines are often situated at high altitudes, crossing rivers, or in other complex terrains, and their surfaces are often covered by supports, valves, and other auxiliary structures, posing significant challenges to inspection.

[0003] Existing pipeline inspection technologies are mainly divided into three categories: manual inspection requires the use of lifting equipment to approach the pipeline, which is inefficient, costly and poses safety hazards; ground mobile robots can crawl along the pipeline, but it is difficult to cross obstacles, and deployment is extremely difficult for overhead pipelines or complex terrain; drone aerial inspection is flexible, but it is greatly affected by flight stability and lighting, and cannot perform precision inspections that require stable contact with the pipeline surface, such as ultrasonic thickness measurement and eddy current testing, making it difficult to detect minute defects.

[0004] It is evident that existing technologies suffer from a lack of balance between testing flexibility, adaptability to complex environments, and precision testing capabilities. Furthermore, they lack the intelligent ability to autonomously identify defects and plan precise testing actions, making it difficult to achieve full automation from defect discovery to diagnosis. Summary of the Invention

[0005] To address the challenges of existing pipeline inspection technologies in simultaneously achieving mobile deployment in complex environments, stable contact inspection, and efficient and accurate identification, this invention provides a flying robot system and inspection method for contact inspection of pipeline facilities.

[0006] On the one hand, a flying robot system for contact inspection of pipeline facilities is provided, including a drone flight platform, a pipeline crawling inspection body, a precision inspection robotic arm, and an intelligent terminal; The drone flight platform is a mobile carrier that carries the pipeline crawling detection body to the airspace above the pipeline equipment to be inspected. The pipeline crawling inspection body is installed below the UAV flight platform and includes a main body, a ring-shaped clamping structure, a preliminary inspection mechanism, and a drive wheel assembly. The ring-shaped clamping structure is a clamp-like structure composed of a multi-segment arm structure. The preliminary inspection mechanism is integrated at the end of the ring-shaped clamping structure. The precision inspection robotic arm is mounted on the UAV flight platform or the pipeline crawling inspection body; The intelligent terminal is connected to the UAV flight platform, the pipeline crawling detection body, and the precision detection robotic arm to control the coordinated operation of each component.

[0007] On the other hand, a detection method for a flying robot system for contact detection of pipeline facilities is provided, including: The drone flight platform carries the pipeline crawling detection body to the area above the preset detection point of the pipeline to be inspected; The ring-shaped clamping structure adaptively clamps the pipeline to be inspected, ensuring that the pipeline crawling inspection body is stably attached to the pipeline surface. The preliminary inspection agency conducts a large-scale, rapid scan of the outer wall of the pipeline to collect image data of the pipeline surface; The deep learning recognition module of the smart terminal analyzes the collected image data, identifies and locates suspected defects, and outputs the location information of the suspected defects. The coordinate transformation unit of the smart terminal converts the location information of the suspected defects into three-dimensional spatial coordinates in the base coordinate system of the precision inspection robot arm. The reinforcement learning control module of the smart terminal is activated to plan the motion trajectory of the precision inspection robot arm according to the three-dimensional spatial coordinates. The precision inspection robotic arm moves according to the planned motion trajectory, bringing the end effector to the suspected defect point for contact precision inspection and obtaining quantitative inspection data; Integrate the test data to generate a test report.

[0008] The above technical solution has the following advantages or beneficial effects: This invention organically combines a drone flight platform with a pipeline crawling inspection body, achieving a synergistic balance between flight maneuverability and crawling stability. The drone flight platform can quickly reach pipeline locations in complex terrain environments, overcoming the deployment difficulties of traditional ground robots. Upon arrival, it adaptively clamps onto the pipeline surface using a wraparound clamping structure, forming a stable inspection platform. This leverages the rapid deployment advantage of drones while compensating for their inability to perform stable contact inspections. Based on this, the invention constructs a two-tiered inspection mode: preliminary screening and precise re-inspection. The preliminary inspection mechanism performs a large-scale, rapid scan of the pipeline's outer wall, efficiently identifying and locating suspected damaged areas. The precision inspection robotic arm performs contact-based precision inspection on suspected defect points. This two-tiered screening avoids the inefficiency of blindly conducting full-coverage inspections while ensuring the accuracy of the inspection results. Attached Figure Description

[0009] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0010] Figure 1 This is a schematic diagram of the overall system structure of Embodiment 1; Figure 2 A schematic diagram showing the state of the system when it is clamped onto the pipeline and being inspected. Figure 3 This is a schematic diagram of the crawling detection mechanism in Example 1; Figure 4 This is a schematic diagram of the precision inspection robotic arm in Example 1; Figure 5 This is an overall architecture diagram of the deep learning recognition module for pipeline defect detection in Example 1; Figure 6 This is a flowchart of the system control logic.

[0011] Reference numerals: 1. Unmanned aerial vehicle (UAV) flight platform; 2. Precision inspection robotic arm; 3. Pipeline crawling inspection main body; 4. Intelligent terminal; 21. Base; 22. First arm segment drive motor; 23. First arm segment; 24. Second arm segment; 25. Third arm segment; 26. End effector; 27. Third arm segment drive motor; 28. Second arm segment drive motor; 31. First drive wheel set; 32. Flexible gripping arm; 33. Robotic arm servo motor; 34. Transmission link; 35. Second drive wheel set; 36. Preliminary inspection mechanism. Detailed Implementation

[0012] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0013] In this embodiment of the invention, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" do not necessarily imply that they are different.

[0014] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0015] All data acquisition in this embodiment is carried out in accordance with laws and regulations and with user consent, and the data is used legally.

[0016] Example 1 like Figure 1 As shown, the present invention provides a flying robot system for contact inspection of pipeline facilities, including a drone flight platform 1, a precision inspection robotic arm 2, a pipeline crawling inspection body 3, and an intelligent terminal.

[0017] The UAV flight platform 1 is the mobile carrier of this system, providing flight power and carrying the pipeline crawling detection body to move to the airspace above the pipeline equipment to be detected; The precision inspection robotic arm 2 is mounted on the UAV flight platform 1 or the pipeline crawling inspection device 3, and is used to perform contact precision inspection on suspected damaged areas on the pipeline surface; The pipeline crawling detection unit 3 is installed below the UAV flight platform and is accurately located to the work area through the vision module of the smart terminal 4, such as... Figure 3 As shown, the pipeline crawling detection body 3 includes a ring-shaped clamping structure, a preliminary detection mechanism 36, and a drive wheel set. The ring-shaped clamping structure is a clamp-like structure composed of a multi-segment arm structure, used to adaptively clamp the pipeline to be inspected. The preliminary detection mechanism is integrated on the ring-shaped clamping structure and is used to perform a large-scale rapid scan of the outer wall of the pipeline after the system clamps the pipeline to find and locate suspected damaged areas. The drive wheel set is used to control the clamping action of the ring-shaped clamping structure and the movement of the pipeline crawling detection body along the surface of the pipeline. The intelligent terminal 4 is connected to the UAV flight platform 1, the pipeline crawling detection body 3, and the precision detection robotic arm 2. The intelligent terminal 4 receives the data information collected by the preliminary detection mechanism and controls the precision detection robotic arm 2 to perform contact precision detection on the suspected damaged area.

[0018] Furthermore, the drone flight platform 1 is a multi-rotor drone with a quick-release interface at its bottom for detachable connection to the pipeline crawling detection body 3.

[0019] Furthermore, the ring-shaped clamping structure consists of four sets of clamp-like mechanical structures, which are symmetrically distributed. Each set of clamp-like mechanical structures includes a robotic arm servo motor 33, a transmission link 34, and a flexible clamping arm 32 connected in sequence. The end of the clamping arm is equipped with a flexible clamp and the preliminary detection mechanism 36.

[0020] The robotic arm servo motor 33 is an independently controlled high-torque digital servo motor, which overlaps with the first drive wheel set; the intelligent terminal controls the rotation angle of the four servo motors to achieve adaptive envelopment and clamping of pipes with different diameters and cross-sectional shapes.

[0021] The preliminary inspection unit 36 ​​includes a high-definition vision camera and a ring-shaped LED fill light. The high-definition vision camera is used to acquire high-definition images of the pipe surface, and the ring-shaped LED fill light is arranged around the outer periphery of the high-definition vision camera to provide uniform illumination when there is insufficient light.

[0022] The drive wheel set includes a first drive wheel set 31 and a second drive wheel set 35. The first drive wheel set 31 is located at the connection of the transmission linkage and is used to drive the opening and closing of the clamping arms. The second drive wheel set 35 is located inside each clamping arm of the circumferential clamping structure and is used to drive the pipe crawling detection body to move along the pipe surface in the clamping state.

[0023] When the robot is ready to perform a surround inspection, the servo motor of the robotic arm drives the flexible gripping arm, composed of multiple transmission links, to clamp the pipe to be tested. When the robot needs to move around the pipe 360 ​​degrees without blind spots, the motor of the first drive itself will drive the first drive wheel set to rotate, ultimately enabling the robot to perform a surround inspection around the pipe.

[0024] Furthermore, such as Figure 4 As shown, the precision inspection robotic arm 2 is a multi-degree-of-freedom lightweight collaborative robotic arm, including a base 21, a first arm segment drive motor 22, a first arm segment 23, a second arm segment 24, a third arm segment 25, an end effector 26, a second arm segment drive motor 28, and a third arm segment drive motor 27 connected in sequence.

[0025] The base is fixedly connected to the bottom of the UAV flight platform 1 or the side of the pipeline crawling detection body 3. The end effector is a replaceable detection sensor interface, which can be used to mount an ultrasonic thickness gauge, an eddy current probe, a magnetic field detection probe or a high-magnification optical microscope according to the detection scenario.

[0026] The first arm segment 23 is a straight plate structure. The first end of the first arm segment 23 is rotatably connected to the first drive motor 22. The first drive motor 22 is mounted on the base. The second end of the first arm segment 23 is clamped to the first end of the second arm segment 24 through a first joint. The first joint has a second arm segment drive motor 28 built in it, which is used to drive the second arm segment to translate relative to the first arm segment in the vertical plane. The second arm segment 24 has a Z-shaped bending structure. The second end of the second arm segment 24 is clamped and connected to the first end of the third arm segment 25 through a second joint. The second joint has a third arm segment drive motor 27 built in it, which is used to drive the third arm segment to translate relative to the second arm segment in the horizontal plane. The third arm segment 25 is a rotatable rod. The second end of the third arm segment 25 is fixedly connected to the end effector 26. An end drive motor is provided inside the third arm segment 25 to drive the end effector 26 to rotate around the axis of the third arm segment.

[0027] Furthermore, angle sensors are respectively installed inside the first arm segment, the second arm segment, and the third arm segment. The angle sensors are used to detect the rotation angle of each arm segment in real time and feed the angle information back to the smart terminal 4. The smart terminal 4 combines the angle information with the action commands output by the reinforcement learning control module.

[0028] Furthermore, the intelligent terminal 4 includes a deep learning recognition module, a reinforcement learning control module, and a coordinate transformation unit.

[0029] The deep learning recognition module is an image recognition model based on convolutional neural networks. It uses a pre-trained ResNet-50 model as the backbone network and performs transfer learning on thousands of pipe images labeled with categories such as "corrosion," "crack," and "normal." This module receives image data collected by the preliminary inspection agency and outputs prediction results with bounding boxes for "suspected defects," category confidence scores, and pixel coordinates in the image.

[0030] The overall architecture of the deep learning recognition module is as follows: Figure 5 As shown, from top to bottom, it includes: image preprocessing submodule, backbone feature extraction network, multi-scale feature fusion submodule, region proposal and feature alignment submodule, defect detection and classification submodule, and result processing and output submodule. The specific functions and connections of each submodule are as follows: The image preprocessing submodule is used to standardize the original image of the pipeline to be detected. Specifically, it includes: scaling pixel values ​​to the [0,1] range and normalizing them according to the mean and variance of the ImageNet dataset; using adaptive scaling to unify the image to a preset size, maintaining the original aspect ratio and adding black borders to the edges; and converting it to the tensor format required for model input. During the training phase, additional online data augmentation operations such as random flipping, rotation, and brightness / contrast adjustment are performed to improve the model's generalization ability.

[0031] The backbone feature extraction network, serving as the core feature extractor of the entire recognition module, is loaded with the weights of a ResNet-50 model pre-trained on the ImageNet large-scale general image dataset. A layered fine-tuning strategy is employed for transfer learning: the first three convolutional stages (conv1-conv3) are frozen to retain general visual feature extraction capabilities, while the last two convolutional stages (conv4-conv5) are fine-tuned to learn the specific texture features of pipe corrosion and cracks. This network simultaneously outputs high-level semantic features and low-level detail features. The high-level semantic features contain defect category information, while the low-level detail features contain edge and location information of the defects.

[0032] The multi-scale feature fusion submodule receives high-level semantic features and low-level detail features from the backbone feature extraction network. It performs multi-scale feature fusion using a combination of top-down upsampling and lateral connections to generate a fused feature map covering defects of different sizes. This submodule solves the problem of low detection accuracy of single-scale features caused by large differences in pipeline defect sizes (from small cracks to large-area corrosion).

[0033] The Region Proposal and Feature Alignment Submodule comprises two parts: the Region Proposal Generation Subunit (RPN) and the Feature Alignment Subunit (RoI-Align). Region proposal generates sub-units, generating anchor points of different scales and aspect ratios on the multi-scale fused feature map. Binary classification is used to determine whether the anchor points contain defects, and preliminary bounding box regression is performed on the anchor points containing defects to filter out several suspected defect candidate boxes. The feature alignment sub-unit maps the selected suspected defect candidate boxes onto the fused feature map of the corresponding scale. The bilinear interpolation method is used to extract the fixed-size feature vector corresponding to each candidate box, which avoids the quantization error of traditional RoI-Pooling and improves the localization accuracy of small defects.

[0034] The defect detection and classification submodule receives the fixed-size feature vector output by the feature alignment submodule. After feature mapping through a fully connected layer, it outputs two branch results: one is the classification branch, which outputs the confidence score of each candidate box belonging to the categories of "corrosion", "crack" and "normal"; the other is the bounding box regression branch, which outputs the precise coordinate correction amount of each candidate box.

[0035] The results processing output submodule performs post-processing on the output results of the defect detection and classification submodule, specifically including: filtering detection results with confidence scores below a preset threshold; using non-maximum suppression (NMS) to remove overlapping duplicate detection boxes; converting relative coordinates to absolute pixel coordinates of the original image; and finally outputting structured data containing the coordinates of suspected defect bounding boxes, defect categories, and category confidence scores.

[0036] The coordinate transformation unit is used to convert the pixel coordinates output by the deep learning recognition module into three-dimensional spatial coordinates in the base coordinate system of the precision inspection robot arm 2, based on the hand-eye calibration relationship between the camera and the precision inspection robot arm 2 and the geometric model of the circumferential clamping structure.

[0037] Furthermore, the hand-eye calibration relationship is a homogeneous transformation matrix from the camera coordinate system to the two-base coordinate system of the precision detection robot. To obtain it, follow these steps: A standard checkerboard calibration plate is used as the calibration target and fixed to the end of the precision inspection robotic arm 2. The precision inspection robotic arm 2 is controlled to move the calibration target to no less than 15 different spatial poses, and the camera is simultaneously controlled to acquire images of the calibration target in each pose. The pose matrix of the calibration target in the camera coordinate system in each pose is calculated by Zhang Zhengyou's calibration algorithm. Simultaneously, the pose matrix of the calibration target in the base coordinate system under the corresponding posture is obtained by precisely detecting the forward kinematics model of the robotic arm 2. Based on the classic hand-eye calibration equation AX=XB, multiple sets of pose data are substituted, and the homogeneous transformation matrix is ​​obtained by using the singular value decomposition algorithm. ,in 3 3. The rotation matrix represents the rotation relationship between the camera coordinate system and the base coordinate system. It is a 3×1 translation vector, representing the translation relationship of the camera optical center relative to the origin of the base coordinate system.

[0038] The geometric model of the encircling clamping structure includes the following core parameters. All parameters are determined through prior calibration and structural design parameters and stored in the parameter storage module of the coordinate transformation unit, including the camera intrinsic parameter matrix. ,in Let x be the focal length of the camera in the x and y directions. The coordinates of the principal point in the image; camera distortion coefficients include radial distortion coefficients. and tangential distortion coefficient The homogeneous transformation matrix obtained from the above hand-eye calibration Camera working distance That is, the vertical distance from the optical center of the camera to the surface of the pipe being inspected.

[0039] The specific process of converting the pixel coordinates into three-dimensional spatial coordinates is as follows: Let the defect pixel coordinates output by the deep learning recognition module be... First, the coordinates of the defective pixels After distortion correction, the coordinates are normalized using the inverse of the camera intrinsic matrix to obtain two-dimensional normalized coordinates in the camera coordinate system, which are then multiplied by the camera working distance. .

[0040] Obtain the three-dimensional coordinates of the defect point in the camera coordinate system. Subsequently, the homogeneous transformation matrix obtained by hand-eye calibration was used. After performing a rigid body transformation, the three-dimensional spatial coordinates of the defect point in the two-base coordinate system of the precision inspection robot are finally obtained. Let the three-dimensional coordinates of the defect point in the camera coordinate system be... ,in, , These are the horizontal and vertical coordinates of the defect point in the camera coordinate system. This refers to the aforementioned camera working distance. (Because the camera's optical axis is perpendicular to the surface being measured during inspection, the coordinates of the defect point's depth direction are always the working distance.) Its scalar calculation formula is:

[0041]

[0042]

[0043] in Rotation matrix elements, Translation vector Element.

[0044] The reinforcement learning control module is pre-trained using a proximal policy optimization algorithm. Its state space includes: target point coordinates, angles of each joint of the precision detection robot arm, and pipeline geometry information. Its action space is the angular velocity of each joint of the precision detection robot arm. Its reward function includes obtaining a positive reward for quickly reaching the target point and obtaining a negative reward for colliding or excessive energy consumption.

[0045] The target point coordinates are the three-dimensional spatial coordinates of the defect point output by the coordinate transformation unit in the base coordinate system of the precision inspection robot arm 2.

[0046] The reinforcement learning control module plans the trajectory motion as follows: it receives the target point coordinates output by the coordinate transformation unit, constructs the system state by combining the joint angles of the precision detection robotic arm 2 and the pipeline geometry information at the current moment, inputs the system state into the pre-trained proximal policy optimization network, and the network outputs the angular velocity commands of each joint of the precision detection robotic arm 2 to control the robotic arm to move towards the target point according to the planned collision-free trajectory; during the movement, the system state is updated in real time and control commands are continuously output until the detection probe at the end of the robotic arm accurately reaches the defect target point.

[0047] The pipeline geometry information includes the radius, axial direction vector, wall thickness, and surface curvature of the pipeline to be detected. This information is obtained through preliminary structural measurements, laser scanning, or system parameter input, and is pre-stored in the system's parameter storage module for use by the reinforcement learning control module. The pre-training convergence condition of the reinforcement learning control module is: the average reward value fluctuation over 100 consecutive training rounds is less than 5%, and the accuracy of the robotic arm successfully reaching the target point reaches over 95%. At this point, training stops, and the optimal network weights are saved for actual inference.

[0048] The control logic of the intelligent terminal 4 includes: controlling the UAV flight platform 1 to fly to the preset pipeline monitoring point, controlling the ring-shaped clamping structure to clamp the pipeline; controlling the preliminary inspection mechanism to collect pipeline surface data; starting the deep learning recognition module to analyze the collected data, identify suspected defects and output their locations; when a suspected defect is identified, starting the reinforcement learning control module to plan the motion trajectory of the precision inspection robotic arm 2; controlling the precision inspection robotic arm 2 to move according to the planned trajectory, bringing the end effector to the suspected defect point for precision inspection; receiving the precision inspection data and integrating it to form an inspection report.

[0049] Example 2 This embodiment provides a detection method for a flying robot system for contact detection of pipeline facilities, based on the flying robot system for contact detection of pipeline facilities described in Embodiment 1, and includes the following steps: The drone flight platform 1, carrying the pipeline crawling detection body 3, flies to the area above the preset detection point of the pipeline to be inspected; the vision module of the smart terminal 4 collects image data of the pipeline and its surrounding environment, and uses a visual recognition algorithm to accurately locate the pipeline, determine the hovering position of the drone flight platform 1, and place the pipeline crawling detection body 3 directly above the pipeline to be inspected.

[0050] Based on the pipe diameter information fed back by the vision module, the intelligent terminal 4 controls the four sets of clamp-like mechanical structures of the encircling clamping structure to move synchronously. Specifically, the intelligent terminal 4 controls the rotation angle of the four servo motors, driving the transmission linkage to open the clamping arm, so that the flexible clamp at the end of the clamping arm wraps around the outer wall of the pipe; when the flexible clamp contacts the pipe surface, the intelligent terminal 4 continues to control the servo motor to output a preset clamping torque, so that the second drive wheel set at the end of the clamping arm presses against the pipe surface with a set pressure, realizing the stable attachment of the system on the pipe, and finally shutting off the flight power of the UAV flight platform 1.

[0051] The intelligent terminal 4 controls the start of the preliminary inspection mechanism integrated on the ring-shaped clamping structure. The preliminary inspection mechanism includes a high-definition vision camera and a ring LED fill light. The intelligent terminal 4 controls the ring LED fill light to turn on, providing uniform illumination to the pipe surface, and controls the high-definition vision camera to perform a 360-degree surround scan of the outer wall of the pipe, acquiring high-definition images of the pipe surface.

[0052] The intelligent terminal 4 inputs the acquired high-definition images into the deep learning recognition module. The deep learning recognition module performs real-time analysis on the input images and outputs detection results with defect category, category confidence score, and bounding box pixel coordinates. When the category confidence score of any detection result is greater than a preset threshold, the detection result is judged as a suspected defect, and the bounding box pixel coordinates of the suspected defect are output.

[0053] The coordinate transformation unit of the intelligent terminal 4 receives the bounding box pixel coordinates output by the deep learning recognition module and converts them into three-dimensional spatial coordinates in the base coordinate system of the precision inspection robot arm 2, which are then used as the detection target points. The intelligent terminal 4 inputs the detection target points into the reinforcement learning control module to plan the motion trajectory of the precision inspection robot arm 2.

[0054] The reinforcement learning control module outputs the target angular velocity of each joint in real time according to the current state, driving the precision inspection robot arm 2 to move along the planned trajectory and accurately and stably contact the end effector to the target point; the intelligent terminal 4 controls the detection sensor on the end effector to start, perform contact precision detection on suspected defect points, and obtain quantitative detection data.

[0055] The intelligent terminal 4 integrates the pipeline surface images collected by the preliminary inspection agency, the suspected defect identification results output by the deep learning recognition module, the three-dimensional coordinates of the target point output by the coordinate transformation unit, and the quantitative detection data collected by the precision detection sensor to form a comprehensive inspection report.

[0056] The intelligent terminal 4 controls the precision inspection robotic arm 2 to reset and controls the ring-shaped clamping structure to loosen; restarts the UAV flight platform 1, flies to the next preset inspection point, and repeats the above steps until the inspection tasks of all preset inspection points are completed.

Claims

1. A flying robot system for contact inspection of pipeline facilities, characterized in that, This includes a drone flight platform, a pipeline crawling inspection unit, a precision inspection robotic arm, and a smart terminal; The drone flight platform is a mobile carrier that carries the pipeline crawling detection body to the airspace above the pipeline equipment to be inspected. The pipeline crawling inspection body is installed below the UAV flight platform and includes a main body, a ring-shaped clamping structure, a preliminary inspection mechanism, and a drive wheel assembly. The ring-shaped clamping structure is a clamp-like structure composed of a multi-segment arm structure. The preliminary inspection mechanism is integrated at the end of the ring-shaped clamping structure. The precision inspection robotic arm is mounted on the UAV flight platform or the pipeline crawling inspection body; The intelligent terminal is connected to the UAV flight platform, the pipeline crawling detection body, and the precision detection robotic arm to control the coordinated operation of each component.

2. The flying robot system for contact detection of pipeline facilities according to claim 1, characterized in that, The drone flight platform is a multi-rotor drone, and its bottom is equipped with a quick-release interface for detachable connection with the pipeline crawling detection body.

3. The flying robot system for contact detection of pipeline facilities according to claim 1, characterized in that, The ring-shaped clamping structure consists of four sets of clamp-like mechanical structures, which are symmetrically distributed. Each set of clamp-like mechanical structures includes a robotic arm servo motor, a transmission link, and a flexible clamping arm connected in sequence. The end of the clamping arm is equipped with a flexible clamp and the preliminary detection mechanism.

4. The flying robot system for contact detection of pipeline facilities according to claim 1, characterized in that, The robotic arm servo motor is an independently controlled high-torque digital servo motor. The intelligent terminal controls the rotation angle of the four servo motors to achieve adaptive envelopment and clamping of pipes with different diameters and cross-sectional shapes.

5. A flying robot system for contact detection of pipeline facilities according to claim 1, characterized in that, The preliminary inspection mechanism includes a high-definition vision camera and a ring-shaped LED supplementary light. The high-definition vision camera is used to acquire high-definition images of the pipe surface, and the ring-shaped LED supplementary light is arranged around the outer periphery of the high-definition vision camera to provide uniform illumination when the light is insufficient.

6. A flying robot system for contact detection of pipeline facilities according to claim 1, characterized in that, The drive wheel set includes a first drive wheel set and a second drive wheel set. The first drive wheel set is disposed at the connection between the transmission link and the clamping arm, and is used to drive the clamping arm to open and close. The second drive wheel set is equidistantly distributed on the inner side of the end of each clamping arm of the circumferential clamping structure, and is used to drive the pipe crawling detection body to move along the pipe surface in the clamping state.

7. A flying robot system for contact detection of pipeline facilities according to claim 1, characterized in that, The precision inspection robotic arm is a multi-degree-of-freedom lightweight collaborative robotic arm, comprising a base, a first arm segment, a second arm segment, a third arm segment, an end effector, a first arm segment drive motor, a second arm segment drive motor, and a third arm segment drive motor connected in sequence. The base is fixedly connected to the bottom of the UAV flight platform or the side of the pipeline crawling detection body. The end effector is a replaceable detection sensor interface, which can be used to mount an ultrasonic thickness gauge, an eddy current probe, a magnetic field detection probe or a high-magnification optical microscope according to the detection scenario. The first arm segment is a straight plate structure. The first end of the first arm segment is rotatably connected to the first drive motor. The first drive motor is mounted on the base. The second end of the first arm segment is clamped and connected to the first end of the second arm segment through a first joint. The first joint has a second arm segment drive motor built in it, which is used to drive the second arm segment to translate relative to the first arm segment in the vertical plane. The second arm segment has a Z-shaped bending structure. The second end of the second arm segment is clamped and connected to the first end of the third arm segment through a second joint. The second joint has a built-in drive motor for the third arm segment, which is used to drive the third arm segment to translate relative to the second arm segment in the horizontal plane. The third arm segment is a rotatable rod. The second end of the third arm segment is fixedly connected to the end effector. An end drive motor is installed inside the third arm segment to drive the end effector to rotate around the axis of the third arm segment.

8. A flying robot system for contact detection of pipeline facilities according to claim 7, characterized in that, Angle sensors are respectively installed inside the first arm segment, the second arm segment and the third arm segment. The angle sensors are used to detect the rotation angle of each arm segment in real time and feed the angle information back to the smart terminal.

9. A flying robot system for contact detection of pipeline facilities according to claim 1, characterized in that, The intelligent terminal includes a deep learning recognition module, a reinforcement learning control module, and a coordinate transformation unit; The deep learning recognition module is used to receive image data collected by the preliminary detection agency and output prediction results with suspected defect bounding boxes, category confidence scores and pixel coordinates in the image. The coordinate transformation unit is used to convert the position information output by the deep learning recognition module into three-dimensional spatial coordinates in the base coordinate system of the precision detection robot arm. The reinforcement learning control module is used to plan the motion path of the precision inspection robot arm after identifying a suspected defect.

10. A detection method for a flying robot system for contact detection of pipeline facilities, utilizing a flying robot system for contact detection of pipeline facilities as described in any one of claims 1-9, characterized in that, include: The drone flight platform carries the pipeline crawling detection body to the area above the preset detection point of the pipeline to be inspected; The ring-shaped clamping structure adaptively clamps the pipeline to be inspected, ensuring that the pipeline crawling inspection body is stably attached to the pipeline surface. The preliminary inspection agency conducts a large-scale, rapid scan of the outer wall of the pipeline to collect image data of the pipeline surface; The deep learning recognition module of the smart terminal analyzes the collected image data, identifies and locates suspected defects, and outputs the location information of the suspected defects. The coordinate transformation unit of the smart terminal converts the location information of the suspected defects into three-dimensional spatial coordinates in the base coordinate system of the precision inspection robot arm. The reinforcement learning control module of the smart terminal is activated to plan the motion trajectory of the precision inspection robot arm according to the three-dimensional spatial coordinates. The precision inspection robotic arm moves according to the planned motion trajectory, bringing the end effector to the suspected defect point for contact precision inspection and obtaining quantitative inspection data; Integrate the test data to generate a test report.

Images