A full-closed-loop intelligent inspection and repair method for trenchless pipeline detection and repair

Abstract

The application discloses a kind of trenchless pipeline detection-repair full closed loop intelligent maintenance system and method, for internal defect detection and accurate repair of urban underground pipeline.The system uses the integration architecture of mobile platform embedded edge computing and repair execution mechanism, forms edge end decision trigger condition through the real-time fusion of multi-source sensing data in feature level, mechanical arm trajectory planning and directional deposition repair responds defect coordinates in space-time, and after repair, secondary scanning verification forms quality closed loop, and the three collaborative constitute the technical closed loop of space-time continuity, complete the full chain operation from defect discovery to repair verification in situ in pipeline interior.It is suitable for trenchless maintenance scene such as municipal drainage, water supply and industrial pipe network.

Description

Technical Field

[0001] This invention relates to the field of trenchless repair technology for municipal pipelines, and in particular to a fully closed-loop intelligent inspection and repair system and method for trenchless pipeline inspection and repair, which is applicable to the detection, intelligent diagnosis and precise repair of internal defects in urban underground drainage pipelines, water supply pipelines and industrial pipe networks. Background Technology

[0002] After long-term operation, urban underground pipeline systems commonly suffer from structural defects such as corrosion, cracks, deformation, misaligned joints, sediment buildup, and tree root intrusion. Traditional pipeline repair methods require road excavation, resulting in high construction costs, significant social impact, and long repair cycles, which are no longer sufficient to meet the demands of modern cities for low-disturbance, high-efficiency maintenance. Trenchless repair technology, by performing inspection and repair within the pipeline itself, avoids road excavation and has become an important development direction for municipal pipeline network maintenance.

[0003] Existing trenchless pipeline maintenance technologies have the following main shortcomings: Firstly, existing pipeline inspection devices mostly use a single visual sensor (such as a CCTV camera) to observe the inside of the pipeline. They rely solely on image information to judge defects and cannot obtain the geometric contour of the defects, the degree of physical thinning of the pipe wall, or the precise spatial pose. In complex pipeline environments (such as low light, high humidity, and sediment coverage), the perception blind zone is large, and the accuracy of defect identification is limited.

[0004] Secondly, existing pipeline defect identification relies heavily on manual interpretation or offline image processing. Detection data needs to be transmitted back to the ground center for analysis, resulting in long decision-making delays and making it impossible to achieve real-time diagnosis and immediate repair response within the pipeline. Furthermore, traditional image processing algorithms rely on manually designed feature extractors, which are difficult to adapt to the complex and ever-changing internal environment of pipelines.

[0005] Third, although existing pipeline repair robots integrate detection and repair functions into a single platform, they mostly adopt extensive repair methods such as circumferential multi-nozzle spraying or mechanical filling, without performing point-to-point directional deposition based on the defect coordinates. This results in significant waste of repair materials and low coverage accuracy. Furthermore, the lack of secondary scanning verification after repair makes it impossible to form a quality closed loop, and the repair effect is difficult to guarantee.

[0006] Fourth, while existing technologies employ convolutional neural networks for pipeline image defect recognition, they only involve offline image processing devices and do not form an integrated architecture with mobile detection platforms and repair execution mechanisms. Furthermore, they do not involve real-time fusion of multi-source sensor data at the feature level and in-situ edge decision-making. There are also solutions that use robotic arms for pipeline repair, but these do not achieve spatiotemporal coordination with multi-source sensing and intelligent decision-making, and the repair actions lack a precise correspondence with the defect locations.

[0007] Therefore, existing technologies still lack a trenchless maintenance system capable of completing the entire chain of operations—"multi-source real-time sensing, edge intelligent decision-making, precise directional repair, and quality closed-loop verification"—in situ inside pipelines. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a trenchless pipeline inspection and repair closed-loop intelligent maintenance system and method, aiming to solve the problems of insufficient adaptability, low defect identification accuracy, poor repair precision, and low operation efficiency of existing pipeline maintenance devices.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: A trenchless pipeline inspection and repair closed-loop intelligent maintenance method adopts an integrated architecture of mobile platform with embedded edge computing and repair execution mechanism. It achieves edge-end decision triggering conditions through real-time fusion of multi-source sensor data at the feature level, spatiotemporal response of the robotic arm trajectory planning and directional deposition repair to defect coordinates, and a quality closed loop formed by post-repair secondary scanning verification. These three elements work together to form a spatiotemporally continuous technical closed loop, completing the following steps in situ inside the pipeline: S1: Multi-source real-time acquisition The mobile inspection platform is controlled to move along the pipeline axis, and visual images, 3D point clouds, wall thickness measurements and platform pose data are collected simultaneously through a multi-source sensor fusion module.

[0010] The multi-source sensor fusion module is integrated into the front end of the mobile inspection platform and includes a high-resolution industrial camera, a 3D laser scanner, an ultrasonic thickness probe, and an inertial measurement unit. Visual images acquire surface texture information of defects, 3D point clouds acquire geometric contour information of defects, the ultrasonic thickness probe acquires information on physical thinning of the pipe wall, and the inertial measurement unit acquires spatial positioning information of defects. The four data sources are complementary and enhanced at the feature level, eliminating the perception blind spots of a single sensor in the complex environment of a pipeline, and providing multimodal complementary input for edge defect identification.

[0011] S2: Edge fusion recognition and decision triggering In the edge computing and decision-making unit, the collected multimodal data is fused and defect identified in real time at the feature level, and the defect type, coordinates, size and severity score are output. When the score exceeds the preset threshold, a repair decision is triggered.

[0012] The edge computing and decision-making unit is located inside the mobile detection platform, equipped with a multi-core processor and a dedicated AI acceleration chip. It receives raw data from the multi-source sensor fusion module via a gigabit Ethernet interface. This unit performs data preprocessing, feature extraction, and defect identification algorithms in situ inside the pipeline. It can output defect type, coordinates, size, and severity scores in real time without transmitting the raw data back to the ground center, and trigger repair decisions in situ based on the scoring results, shortening the decision delay from defect discovery to repair response.

[0013] S3: Robotic Arm Trajectory Planning Based on the defect identification results and the platform's current pose, a preset search algorithm is used to plan the robotic arm's motion trajectory from its current position to the defect point.

[0014] The preset search algorithm comprehensively evaluates path length, joint motion smoothness, and obstacle avoidance constraints, enabling the robotic arm to move to the defect point with continuous position, speed, and acceleration within the narrow space of the pipe, ensuring that the repair actuator can accurately respond to the defect coordinates.

[0015] S4: Targeted Depositional Remediation The control and repair actuator moves to the defect point according to the planned trajectory, and deposits repair material in a directional manner with preset flow rate and pressure.

[0016] The repair execution mechanism, installed at the rear of the mobile inspection platform, includes a robotic arm, a material storage bin, and a spraying device. The robotic arm is a 6-DOF (degrees of freedom) serial structure, and the spraying device uses a precision metering pump and nozzles to directionally deposit repair material onto the defect area. The robotic arm and the spraying device work together to ensure that the nozzle axis is always aligned with the normal direction of the defect center during the repair process, achieving a precise point-to-point response to the defect coordinates, which differs from the rough repair method of circumferential spraying.

[0017] S5: Secondary Scan Verification After the repair is completed, the repaired area is scanned and verified a second time using a multi-source sensor fusion module to confirm the integrity of the repair material coverage and the uniformity of its thickness.

[0018] The secondary scanning verification verifies whether the repair material completely covers the defect area and meets the design requirements by comparing visual images and matching three-dimensional point clouds, thus forming a complete closed loop from defect discovery, repair execution to quality verification.

[0019] S6: Data Upload The entire process data and results are uploaded to the ground monitoring center via a wireless communication module.

[0020] It is important to note that the aforementioned multi-source feature fusion, in-situ edge decision triggering, robotic arm spatiotemporal response, and secondary verification are not simply parallel setups within the pipeline, but rather form a spatiotemporally continuous technical closed loop. Real-time fusion of multi-source sensor data at the feature level provides decision triggering conditions at the edge, enabling repair decisions to be triggered in situ without waiting for data transmission back to the ground center. Robotic arm trajectory planning and directional deposition repair directly respond to defect coordinates in spatiotemporally, establishing a spatiotemporal correspondence between repair actions and defect locations within the pipeline. Post-repair secondary scanning verification feeds the repair results back to the edge, forming a quality closed loop. All three are indispensable. Without in-situ edge decision-making, real-time repair triggering is impossible; without trajectory planning and directional deposition, accurate response to defect coordinates is impossible; and without secondary verification, repair quality cannot be guaranteed. Together, they constitute a systematic synergy that distinguishes them from existing step-by-step or extensive detection and repair schemes.

[0021] As described above, this invention eliminates the blind spots of a single visual sensor in the complex environment of pipelines by real-time fusion of multi-source sensor data at the feature level, significantly improving the accuracy and robustness of defect identification. Through in-situ decision-making using edge computing, the decision-making process from defect discovery to repair response is compressed into the pipeline, avoiding the time delay of data transmission back to the ground center for analysis in traditional methods. Through the spatiotemporal coordination of robotic arm trajectory planning and directional deposition repair, point-to-point precise response to defect coordinates is achieved, resulting in high utilization of repair materials and good coverage accuracy. Post-repair secondary scanning verification forms a complete closed loop from detection to repair to quality confirmation, ensuring traceability and verifiability of the repair effect. These four elements, through an integrated architecture embedded in a mobile platform, form a spatiotemporally continuous technical closed loop, whose overall operational efficiency and repair quality are significantly superior to the simple summation of the individual effects of each technical feature. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the overall architecture of the present invention, showing the overall configuration relationship and data flow of the mobile detection platform, multi-source sensor fusion module, edge computing and decision-making unit, repair execution mechanism and central control and communication system.

[0023] Figure 2 This is a schematic diagram illustrating the principle of multi-source sensor fusion and defect identification of the present invention. It shows the process of multi-source acquisition, preprocessing, feature extraction, and deep convolutional neural network defect identification of visual images, 3D point clouds, wall thickness measurements, and platform pose data.

[0024] Figure 3 The flowchart illustrates the closed-loop operation of detection, decision-making, and repair in this invention, showing six closed-loop steps in sequence: multi-source real-time acquisition, edge fusion recognition and decision triggering, robotic arm trajectory planning, directional deposition repair, secondary scanning verification, and data upload.

[0025] Figure 4 This is a schematic diagram of the composition of the mobile detection platform of the present invention, showing the arrangement of the multi-section articulated structure, active flexible joint, adaptive drive module and support wheel assembly.

[0026] Figure 5 This is a schematic diagram of the multi-source sensing fusion module of the present invention, showing the integrated configuration of a high-resolution industrial camera, a 3D laser scanner, an ultrasonic thickness probe, and an inertial measurement unit.

[0027] Figure 6 This is a schematic diagram of the composition of the repair actuator of the present invention, showing the arrangement and connection relationship of the 6-DOF serial robotic arm, material storage bin, precision metering pump and nozzle. Detailed Implementation

[0028] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0029] Example 1: Municipal drainage pipeline maintenance under standard operating conditions like Figures 1 to 6 As shown, this embodiment is applied to the internal maintenance of a reinforced concrete municipal drainage pipe with a diameter of 500mm in urban areas.

[0030] Step 1: System Initialization and Platform Management The system is initialized and its path loaded at the ground monitoring center. The mobile detection platform is inserted into the pipeline through the pipeline inlet. For example... Figure 4 As shown, the mobile detection platform adopts a multi-section articulated structure, with each segment connected by active flexible joints. Each joint incorporates a servo motor and a harmonic reducer, providing pitch and yaw motion capabilities. The active flexible joints employ dual-encoder feedback control: an absolute encoder detects the absolute joint angle, and an incremental encoder detects the motor speed. The joint control law uses a proportional-integral-derivative controller, with the following transfer function: G(s) = Kp + Ki / s + Kd·s The proportional coefficient Kp ranges from 2.0 to 5.0, the integral coefficient Ki ranges from 0.1 to 0.5, and the derivative coefficient Kd ranges from 0.05 to 0.2. In a preferred embodiment, Kp is set to 3.5, Ki is set to 0.3, and Kd is set to 0.1, so that the overshoot of the joint step response is controlled within 5% and the adjustment time does not exceed 0.5 seconds.

[0031] The platform is equipped with an adaptively adjustable diameter drive module (2). The drive module (2) consists of a support wheel assembly, which is driven by three radially distributed electric push rods with a stroke range of 50mm to 200mm. When the platform enters the pipeline, the electric push rods of the support wheel assembly automatically adjust their extension length according to the pipe diameter data (500mm) detected by the system, so that the wheel assembly is pressed tightly against the pipe wall with constant pressure, preventing slippage or jamming on the deposit surface. When the platform enters the pipe bend area, the active flexible joint dynamically adjusts the relative angle of each segment according to the pose data fed back by the inertial measurement unit (14), ensuring that the central axis of the platform is always parallel to the pipeline axis.

[0032] Step 2: Real-time data acquisition from multiple sources During the platform's movement, the multi-source sensor fusion module (1) continues to operate. For example... Figure 5 As shown, the multi-source sensor fusion module (1) is integrated at the front end of the mobile inspection platform and includes a high-resolution industrial camera (11), a 3D laser scanner (13), an ultrasonic thickness probe (12), and an inertial measurement unit (14). The high-resolution industrial camera (11) is equipped with a ring-shaped LED fill light strip and acquires images of the inner wall of the pipe at a speed of 30 frames / second; the 3D laser scanner (13) acquires three-dimensional point cloud data of the inner wall of the pipe based on the time-of-flight principle, with a scanning frequency of 10Hz and a point cloud density of not less than 10,000 points per square meter; the ultrasonic thickness probe (12) operates at a frequency of 5MHz and has a measurement accuracy of ±0.1mm; the inertial measurement unit (14) outputs the six-degree-of-freedom pose data of the platform in the pipe. The data acquisition interval is set to 0.1 seconds.

[0033] The data preprocessing process of the multi-source sensor fusion module (1) includes: performing gamma correction and white balance adjustment on the RGB images acquired by the high-resolution industrial camera (11); performing statistical outlier filtering on the point cloud data acquired by the 3D laser scanner (13) to remove noise points with a neighborhood number of k=50 and a standard deviation multiple of threshold=2.0; and performing temperature compensation on the readings of the ultrasonic thickness probe (12). The compensation coefficient is dynamically calculated based on the thermal expansion coefficient of the pipe material. For example, for concrete materials, the compensation coefficient is based on a thermal expansion coefficient of 10×10. -6 / ℃ Dynamic calculation. The preprocessed multimodal data is transmitted to the edge computing and decision-making unit via a gigabit Ethernet interface, providing standardized input for real-time feature fusion.

[0034] Step 3: Edge fusion recognition and decision triggering The edge computing and decision unit is equipped with a multi-core processor and a dedicated AI acceleration chip. After receiving the raw data from the multi-source sensing fusion module (1), it performs data preprocessing, feature extraction and defect identification algorithms.

[0035] Feature extraction includes: edge detection based on the Canny operator, with high and low thresholds set to 100 and 200 respectively; surface defect segmentation based on region growing, with a similarity threshold set to 15 gray levels; and geometric feature descriptor calculation based on Fast Point Feature Histogram (FPFH) for 3D point cloud data, with a neighborhood search radius set to 5mm. FPFH can effectively capture local geometric features of the point cloud, such as curvature and normal direction, providing rich geometric information for subsequent defect classification. The above feature extraction process is processed in parallel on the AI ​​acceleration chip of the edge computing unit to ensure real-time performance.

[0036] The defect identification algorithm is based on a deep convolutional neural network architecture. The network input is a fused multimodal data tensor, constructing a deep neural network with 12 convolutional layers: the first three convolutional layers use 3×3 kernels with a stride of 1, followed by a max-pooling layer with a pooling size of 2×2; the middle six convolutional layers introduce residual connections to avoid gradient vanishing; the last three convolutional layers use dilated convolutions with dilation rates of 2, 4, and 6 to increase the receptive field. The network ends with a global average pooling layer and three fully connected layers, with 512, 256, and 8 neurons respectively, corresponding to eight common pipeline defect categories: corrosion, cracks, perforations, scaling, deformation, interface misalignment, deposition, and root intrusion. The network is trained using a cross-entropy loss function and the Adam optimizer, with an initial learning rate of 0.001 and a batch size of 32. After training on a dataset containing 100,000 pipeline defect images, the test set accuracy reaches 96.3%. A Dropout rate of 0.15 is set before the fully connected layers to improve generalization ability.

[0037] The network output includes the defect type, precise coordinates, size parameters, and severity score, with a score threshold set between 0 and 1. When the score exceeds 0.7, the system triggers a repair decision.

[0038] Step 4: Robotic Arm Trajectory Planning Based on the defect identification results and the current pose of the platform, the motion trajectory of the robotic arm (32) from the current position to the defect point is planned by a preset search algorithm. The trajectory evaluation function comprehensively considers the path length, joint motion smoothness and obstacle avoidance constraints. In a preferred embodiment, the weight coefficients of the three factors are set to 0.6, 0.3 and 0.1, respectively.

[0039] The trajectory planning of the robotic arm (32) is based on a fifth-order polynomial interpolation algorithm to ensure the continuity of joint position, velocity and acceleration. Its joint space trajectory function is expressed as: q(t) = a0 + a1t + a2t 2 + a3t 3 + a4t 4 + a5t 5 Among them, the coefficients a0 to a5 are uniquely determined by the position, velocity, and acceleration boundary conditions of the starting and ending points. This algorithm ensures that the robotic arm (32) moves to the defect point without impact or vibration in the narrow space of the pipe.

[0040] Step 5: Targeted Deposition Remediation The repair actuator (3) moves to the defect point according to the planned trajectory. For example... Figure 6 As shown, the repair actuator (3) is installed at the rear of the mobile inspection platform, including a robotic arm (32), a material storage bin (31), and a spraying device (33). The robotic arm (32) is a 6-DOF serial structure with a repeatability of ±0.05mm and a load capacity of 2kg. The material storage bin (31) contains fast-curing repair material. The spraying device (33) uses a precision metering pump and nozzle to directionally deposit the repair material onto the defect area. In a preferred embodiment, the spraying device (33) applies the repair material at a preset flow rate of 2ml / s and a pressure of 0.5MPa, and the material curing time is controlled at 60 seconds. The robotic arm (32) and the spraying device (33) work together to ensure that the nozzle axis is always aligned with the normal direction of the defect center during the repair process, ensuring uniform thickness of the repair material deposition and achieving point-to-point precise response of the defect coordinates.

[0041] Step Six: Secondary Scan Verification and Data Upload After the repair is completed, the multi-source sensor fusion module (1) performs a second scan on the repaired area. By comparing the visual image acquired by the high-resolution industrial camera (11) with the three-dimensional point cloud obtained by the 3D laser scanner (13), it is verified whether the repair material completely covers the defect area and the thickness meets the design requirements, thus forming a complete closed loop from defect discovery, repair execution to quality verification.

[0042] The central control and communication system's main computer runs a real-time operating system, defining five operating modes through state machine logic: initialization mode, detection mode, decision-making mode, repair mode, and emergency stop mode. Switching between modes is triggered by sensor data and internal timers, ensuring high reliability of the workflow. The wireless communication module supports 4G / 5G and Wi-Fi 6 protocols, enabling bidirectional data exchange with the ground monitoring center. The power management module monitors the lithium-ion battery pack status and provides overcurrent and overheat protection; the overcurrent threshold is set at 20A, and the overheat protection temperature is 85℃.

[0043] All process data and results, including original inspection images, defect identification reports, repair process videos, and repair verification data, are uploaded to the ground monitoring center in real time via a wireless communication module.

[0044] Example 2: Comparison with existing distributed architectures Taking a 500mm diameter municipal drainage pipe in a city as an example, if the existing step-by-step inspection model is adopted: First, a CCTV inspection vehicle is dispatched into the pipe to collect video. After the video is transmitted back to the ground, defects are analyzed manually or by offline software. After the analysis is completed, a repair team is dispatched with repair equipment to re-enter the pipe. After the repair is completed, an inspection vehicle is dispatched again for verification. Under this model, a single 100-meter section of pipe requires three pipe entry operations, with a total construction period of about 3 days. The defect omission rate of manual interpretation is about 15%, the waste rate of circumferential spray repair materials is about 40%, and there is no in-situ verification method after repair.

[0045] After adopting the trenchless pipeline inspection-repair closed-loop intelligent maintenance system described in this invention, the mobile inspection platform can complete all operations from defect discovery to repair verification in a single pipeline entry. The maintenance cycle of a single 100-meter pipeline is shortened to 4 hours, the defect identification accuracy reaches 96.3%, the utilization rate of directional deposition repair materials is increased by more than 60%, and the coverage integrity and thickness uniformity are ensured by secondary scanning verification through the multi-source sensor fusion module (1) after repair. The above effects are due to the systematic synergy of the multi-source sensor fusion module (1) in eliminating perception blind spots, eliminating backhaul delay through edge computing in-situ decision-making, eliminating material waste through trajectory planning of the robotic arm (32) and the spraying device (33) through directional deposition, and forming a quality closed loop through secondary verification, rather than the isolated improvement of a single technical module.

[0046] Compared with the prior art, the present invention has the following specific beneficial effects: (1) Significantly enhanced perception capability. By complementing multiple sources of visual images, 3D point clouds, wall thickness measurements and platform pose data, the blind spots of a single CCTV vision sensor in complex pipeline environments are eliminated, and the defect identification accuracy reaches 96.3%, which is significantly better than a single vision detection scheme.

[0047] (2) Decision response time is greatly shortened. By using edge computing for in-situ real-time reasoning, the decision-making process from defect discovery to repair response is compressed to be completed inside the pipeline. There is no need to wait for data to be sent back to the ground center for analysis. The maintenance cycle of a single pipeline section is shortened from 3 days in the step-by-step mode to 4 hours.

[0048] (3) Significantly improved repair accuracy. Through the spatiotemporal coordination of robotic arm trajectory planning and directional deposition repair, the nozzle axis is always aligned with the normal direction of the defect center, achieving point-to-point precise response of defect coordinates, and the utilization rate of repair materials is increased by more than 60% compared with the circumferential spraying method.

[0049] (4) A complete quality closed loop is formed. Through secondary scanning verification after repair, the integrity of repair coverage and thickness uniformity are confirmed by visual image comparison and 3D point cloud matching, forming a complete closed loop from defect discovery, repair execution to quality verification, which solves the defect of existing technology that has no in-situ verification after repair.

[0050] (5) Significant system synergy effect. Multi-source feature fusion, edge in-situ decision-making, trajectory planning directional deposition and secondary scanning verification form a spatiotemporally continuous technical closed loop through the integrated architecture embedded in the mobile platform. Its overall operation efficiency and repair quality are significantly better than the simple superposition of the effects of each technical feature alone, and it is suitable for the high-efficiency and low-disturbance maintenance needs of urban underground pipe networks.

[0051] As a further explanation of the principle of the technical solution of the present invention: The aforementioned technical effects of this invention are not a simple superposition of the isolated effects of each technical module, but rather a systematic enhancement effect inevitably produced by the spatiotemporal continuous closed loop formed within the pipeline by multi-source sensor fusion, edge in-situ decision-making, trajectory planning and directional deposition, and secondary scanning verification. Specifically: Regarding the principle behind the improved accuracy of multi-source fusion identification: In environments with low light, high humidity, and sediment coverage in pipelines, a single visual sensor can only acquire surface texture information of defects, failing to perceive the degree of thinning within the pipe wall and the three-dimensional geometric contours of the defects. This leads to the missed detection of numerous hidden defects (such as internal wall corrosion and layered peeling). This invention utilizes visual images to acquire texture, three-dimensional point clouds to acquire geometric contours, ultrasonic thickness measurements to acquire physical thinning, and inertial measurements to acquire spatial pose. These four data sources complement each other at the feature level: when the visual image suffers from blurred defect boundaries due to uneven lighting, the geometric abrupt changes in the three-dimensional point cloud can assist in locating the defect contour; when the point cloud data generates noise due to reflections from the inner wall of the pipe, the physical measurement value of the ultrasonic thickness can verify the authenticity of the defect; and the pose data from the inertial measurement unit provides a global positioning reference for the defect coordinates. This four-source collaboration results in complementary gains in the information entropy of defect identification, significantly improving accuracy compared to single-vision detection—a technical effect that a single sensor, regardless of image algorithm optimization, cannot achieve.

[0052] Regarding the principle of shortening response latency through in-situ edge decision-making: In existing distributed architectures, detection data needs to be collected inside the pipeline, transmitted back via cable, analyzed at the ground center, issued commands, and executed for repair. Data travels back and forth between the pipeline and the ground, and latency is limited by communication bandwidth and pipeline length. This invention embeds the edge computing unit within a mobile platform. Multi-source data is transmitted in a closed loop within the platform via gigabit Ethernet. Defect identification and repair triggering are both completed in-situ within the pipeline, eliminating the physical transmission latency of data exiting and re-entering the pipeline. Simultaneously, the edge AI acceleration chip performs parallel computations on Canny edge detection, region growing and segmentation, FPFH feature description, and CNN inference, compressing the time complexity of defect identification and decision triggering from "communication round-trip time + ground analysis time" to "single-frame inference time." This is an architectural-level efficiency improvement brought about by sinking ground center computing power into the platform, rather than something that can be achieved through simple algorithm optimization.

[0053] Regarding the principle of trajectory planning and directional deposition to improve material utilization: Existing circumferential spraying methods involve rotating nozzles around the pipe axis or arranging multiple fixed nozzles circumferentially, spraying repair material onto the inner wall of the pipe in a surface-covering manner. The material deposition area is much larger than the defect area, resulting in a large amount of material adhering to the normal pipe wall and causing waste. This invention plans the motion trajectory of a robotic arm using an A* search algorithm, enabling the nozzle to move precisely to the defect coordinates in a point-to-point manner. Fifth-order polynomial interpolation ensures continuous and impact-free joint movement, allowing the nozzle to maintain a constant posture aligned with the normal direction of the defect center after reaching the defect point. A precision metering pump, with preset flow rate and pressure, directionally deposits the material into the defect area, ensuring a high degree of overlap between the material deposition area and the defect area. This causal chain of "defect coordinates, trajectory planning, normal alignment, and fixed-point deposition" increases the utilization rate of repair material by more than 60% compared to the extensive circumferential spraying method. This is a direct result of the synergistic effect of robotic arm trajectory planning and precision metering; the absence of any one of these links would prevent accurate deposition.

[0054] Regarding the principle of forming a quality closed loop through secondary scanning verification: Existing technologies typically remove the repaired area directly from the pipeline after repair, relying solely on construction experience to judge the repair effect. This lacks the ability to quantitatively verify whether the repair material completely covers the defects, whether the thickness is uniform, and whether there are air bubbles or detachment. This invention, after repair, performs a secondary scan of the repaired area using the same multi-source sensor fusion module. It utilizes visual image comparison to examine texture changes before and after repair, 3D point cloud matching to the geometric shape of the repaired area, and ultrasonic thickness verification to check the repair layer thickness. The repair results are then fed back to the edge in a data-driven manner. This in-situ closed-loop verification of "detection, repair, and re-detection" upgrades repair quality from "qualitative experience" to "quantitative data," ensuring that the delivery standard is met on the first pipe insertion and avoiding rework due to incomplete repair.

[0055] Regarding the principle of systemic synergy: A clear causal dependency exists among the four components mentioned above—multi-source fusion provides reliable input for edge decision-making, edge decision-making provides triggering conditions for trajectory planning, trajectory planning provides an execution path for directional deposition, and secondary verification provides quality confirmation for the overall closed loop. These four components form a spatiotemporally continuous technical closed loop within the pipeline through an integrated architecture embedded in the mobile platform. The overall effect exhibits non-linear enhancement: without multi-source fusion, edge decision-making may be falsely triggered or missed due to insufficient input information; without edge decision-making, trajectory planning may delay response due to the inability to obtain defect coordinates in real time; without trajectory planning, directional deposition may revert to extensive spraying due to inaccurate positioning; without secondary verification, the quality closed loop breaks, and the repair effect cannot be confirmed. Therefore, the technical effect of this invention stems from the synergistic binding of the four components in the spatiotemporal dimension within the pipeline, rather than a simple juxtaposition of independent functions of each module.

[0056] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A trenchless pipeline inspection and repair fully closed-loop intelligent maintenance method, characterized in that, An integrated architecture is adopted, incorporating edge computing and repair execution mechanisms within a mobile platform. This architecture utilizes real-time fusion of multi-source sensor data at the feature level to form edge-end decision triggering conditions, robotic arm trajectory planning and directional deposition repair to respond spatiotemporally to defect coordinates, and post-repair secondary scanning verification to form a quality closed loop. These three elements work together to constitute a spatiotemporally continuous technical closed loop, completing the following steps in situ inside the pipeline: S1: Control the mobile inspection platform to move along the pipeline axis and simultaneously collect visual images, 3D point clouds, wall thickness measurements and platform pose data through the multi-source sensor fusion module; S2: In the edge computing and decision-making unit, the collected multimodal data is fused and defect identified in real time at the feature level, and the defect type, coordinates, size and severity score are output. When the score exceeds the preset threshold, a repair decision is triggered. S3: Based on the defect identification results and the platform's current pose, the robot arm's motion trajectory from its current position to the defect point is planned using a preset search algorithm; S4: Control the repair actuator to move to the defect point according to the planned trajectory, and deposit repair material in a directional manner with preset flow rate and pressure; S5: After the repair is completed, the repair area is scanned and verified a second time using a multi-source sensor fusion module to confirm the integrity of the repair material coverage and the uniformity of its thickness. S6: Upload all process data and results to the ground monitoring center via wireless communication module.

2. A trenchless pipeline inspection and repair fully closed-loop intelligent maintenance system, characterized in that, include: The mobile inspection platform is configured to move autonomously inside the pipeline and carry inspection and repair loads. It adopts a multi-section articulated structure, with each section connected by an active flexible joint. The platform's outer contour is equipped with a drive module that can adaptively adjust the diameter. The multi-source sensor fusion module, integrated at the front end of the mobile detection platform, includes a high-resolution industrial camera, a 3D laser scanner, an ultrasonic thickness probe, and an inertial measurement unit. The edge computing and decision-making unit is located inside the mobile detection platform. It is equipped with a multi-core processor and a dedicated AI acceleration chip. It receives the raw data from the multi-source sensing fusion module through a gigabit Ethernet interface and performs data preprocessing, feature extraction and defect identification algorithms. The repair actuator, installed at the rear of the mobile inspection platform, includes a robotic arm, a material storage bin, and a spraying device. The robotic arm is a 6-DOF serial structure, and the spraying device uses a precision metering pump and nozzles to directionally deposit material onto the defect area. The central control and communication system integrates a main control computer, a wireless communication module, and a power management module.

3. The method or system according to claim 1 or 2, characterized in that, The multi-source sensor fusion module synchronously acquires visual images, 3D point clouds, wall thickness measurements, and platform pose data, which are then complementarily enhanced at the feature level. Visual images are used to obtain surface texture information of defects, 3D point clouds are used to obtain geometric contour information of defects, ultrasonic thickness probes are used to obtain physical thinning information of pipe walls, and inertial measurement units are used to obtain spatial positioning information of defects. This eliminates the blind spots of a single sensor in the complex environment of a pipeline and provides multimodal complementary input for edge defect identification.

4. The method or system according to claim 1 or 2, characterized in that, The edge computing and decision-making unit performs data preprocessing, feature extraction, and defect identification algorithms in situ inside the pipeline. It can output defect type, coordinates, size, and severity scores in real time without transmitting the original data back to the ground center. Based on the score results, it can trigger repair decisions in situ, shortening the decision delay from defect discovery to repair response.

5. The method or system according to claim 1 or 2, characterized in that, The active flexible joint of the mobile detection platform adopts dual encoder feedback control. The absolute encoder detects the absolute angle of the joint and the incremental encoder detects the motor speed to form a closed-loop control. The drive module consists of a support wheel assembly. The extension length of the support wheel assembly is adjusted by radially distributed electric push rods, so that the wheel assembly is in close contact with the pipe wall with constant pressure, keeping the central axis of the platform parallel to the pipe axis in pipe bends and sediment environments.

6. The method or system according to claim 1 or 2, characterized in that, In step S3, when the preset search algorithm plans the movement trajectory of the robotic arm, it comprehensively evaluates the path length, joint movement smoothness and obstacle avoidance constraints, so that the robotic arm moves to the defect point with continuous position, speed and acceleration in the narrow space of the pipe; in step S4, the spraying device uses a precision metering pump and nozzle to deposit the repair material into the defect area at a preset flow rate and pressure, so as to achieve a point-to-point precise response to the defect coordinates, which is different from the rough repair of circumferential spraying.

7. The method or system according to claim 1 or 2, characterized in that, The robotic arm of the repair actuator is a 6-DOF serial structure, the material storage bin contains rapidly curing repair material, and the spraying device uses a precision metering pump and nozzle to directionally deposit the material onto the defect area; The robotic arm works in conjunction with the spraying device to ensure that the nozzle axis is always aligned with the normal direction of the defect center during the repair process, thus ensuring that the thickness of the repair material deposition is uniform.

8. The method or system according to claim 1 or 2, characterized in that, In step S5, the secondary scanning verification uses a multi-source sensor fusion module to compare visual images and match three-dimensional point clouds of the repair area to verify whether the repair material completely covers the defect area and whether the thickness meets the design requirements, thus forming a complete closed loop from defect discovery, repair execution to quality verification.

9. The method or system according to claim 1 or 2, characterized in that, The main control computer of the central control and communication system runs a real-time operating system. It defines initialization mode, detection mode, decision mode, repair mode and emergency stop mode through state machine logic. The switching conditions between modes are triggered by sensor data and internal timers. The wireless communication module supports 4G / 5G and Wi-Fi 6 protocols to realize two-way data interaction with the ground monitoring center. The power management module monitors the status of the lithium-ion battery pack and provides overcurrent and overheat protection.

10. The method or system according to claim 3, characterized in that, The data preprocessing of the multi-source sensing fusion module includes: performing gamma correction and white balance adjustment on RGB images acquired by industrial cameras; performing statistical outlier filtering on point cloud data acquired by 3D laser scanners to remove noise points; performing temperature compensation on ultrasonic thickness probe readings, with the compensation coefficient dynamically calculated based on the thermal expansion coefficient of the pipe material; and transmitting the preprocessed multimodal data to the edge computing and decision unit via a gigabit Ethernet interface to provide standardized input for real-time feature-level fusion.

Images