A pipeline facility automatic inspection system and method based on distributed optical fiber sensing and unmanned inspection feedback

Abstract

The present application belongs to the technical field of pipeline automatic inspection, and particularly relates to a pipeline facility automatic inspection system and method based on distributed optical fiber sensing and unmanned inspection feedback. The sensing subsystem is deployed at multiple monitoring nodes, collects vibration signals along the optical fiber and generates risk event information. The first management platform generates an alarm level according to a risk assessment strategy, and the second management platform generates an inspection task accordingly, dispatches a drone or unmanned vehicle to the scene to perform inspection and collect on-site state information. The second management platform disposes according to the on-site state information, and pushes the information to the first management platform to store in the training database, which is used to iteratively update the risk assessment strategy and issue to the sensing subsystem. The present application uses the optical fiber sensing system to detect the construction vibration signals along the pipeline, automatically triggers the drone or unmanned vehicle to the scene for verification, and returns the verification result to correct the judgment standard of the sensing system, thereby reducing false alarms and invalid dispatches.

Description

Technical Field

[0001] This invention belongs to the field of pipeline automated inspection technology, specifically relating to an automated pipeline facility inspection system and method based on distributed optical fiber sensing and unmanned inspection feedback. Background Technology

[0002] Currently, the inspection of long-distance pipeline facilities such as communication optical cables, oil and gas pipelines, and municipal pipe networks mainly relies on manual visual inspections along the pipeline routes. Since pipelines are usually buried underground or laid overhead along roads, the area that inspectors can cover each day is limited. In particular, for destructive hazards such as external construction excavation and mechanical impacts along the pipeline route, manual inspections often can only discover them after an accident has occurred. The pre-detection rate of safety risk events is low, making it impossible to achieve the safety management goal of early detection and early response.

[0003] Existing technologies have proposed some automated inspection solutions. For example, patent CN202510322226.1 provides a method for inspecting hazardous sources. It uses distributed fiber optic acoustic wave sensors to collect acoustic wave information and performs spectrum analysis through a client to identify hazardous sources. However, acoustic wave propagation is easily affected by environmental noise interference and has a limited distance. Furthermore, the sensing devices and drone base stations are all converged at a local switch, failing to form a provincial-level network. Another example is patent CN202411355588.2, which discloses an intelligent monitoring method for optical cables based on digital twins. This method filters transmission fault markers and predicts fault levels by analyzing optical cable performance data. However, its detection target is the transmission performance faults of the optical cable itself, rather than environmental hazards such as external construction vibrations along the pipeline, and it lacks on-site inspection execution capabilities. For example, patent CN201910251745.8 proposes an automatic inspection system for power distribution automation systems. However, its inspection terminal consists of a communication module, a data acquisition unit, and an alarm module. Its actual function is limited to issuing alarms, lacking on-site inspection execution entities such as drones or unmanned vehicles, thus failing to achieve on-site verification of dangerous events. Furthermore, the distributed fiber optic sensing devices in the aforementioned existing solutions are all limited to independent deployment on a single pipeline or in a single equipment room. The sensing results are disconnected from the inspection execution, failing to achieve multi-node-scale networking and unified scheduling within a province. Inspection resources are difficult to automatically allocate according to risk levels, and the sensing detection rate does not match the actual handling efficiency. Summary of the Invention

[0004] The purpose of this invention is to provide an automatic pipeline facility inspection system and method based on distributed optical fiber sensing and unmanned inspection feedback, so as to solve the technical problems of insufficient inspection triggering effectiveness caused by detection and positioning deviations in existing distributed optical fiber sensing systems under different pipeline scenarios, and the lack of a scenario adaptive calibration mechanism based on on-site feedback.

[0005] The present invention achieves the above objectives through the following technical solutions: Firstly, this invention proposes an automatic pipeline facility inspection system based on distributed optical fiber sensing and unmanned inspection feedback, applicable to multi-domain networked distributed optical fiber sensing devices and joint inspection scenarios involving drones and / or unmanned vehicles. The system includes: The sensing subsystem is deployed at multiple monitoring nodes. It collects vibration signals and generates risk event information by connecting to optical fibers laid along the pipeline under test. The first management platform is used to receive the risk event information and generate alarm levels according to the preset risk assessment strategy. The second management platform is used to generate inspection tasks based on the alarm level. The scheduling subsystem is used to receive the inspection tasks and schedule the inspection terminals. Inspection terminals, including drones and / or unmanned vehicles, are controlled by the scheduling subsystem to go to the target site to perform automatic inspections, collect on-site status information and provide feedback; The second management platform handles the event based on the on-site status information, pushes the on-site status information to the first management platform, stores it in the training database for iteratively updating the risk assessment strategy, and distributes the updated strategy to the perception subsystem.

[0006] Furthermore, the sensing subsystem includes distributed optical fiber sensing hosts deployed at multiple monitoring nodes. Each distributed optical fiber sensing host is connected to an idle optical fiber core in the optical cable associated with the pipeline under test at the corresponding monitoring node to independently collect vibration signals. The distributed optical fiber sensing host has a built-in event judgment and decision model. The event judgment and decision model is used to identify whether there are external hazards along the pipeline under test and the radial position of the external hazards relative to the pipeline under test based on the vibration signal, and to generate the risk event information. The training database stores radial position data confirmed by the on-site status information; The first management platform iteratively updates the event judgment and decision model based on the radial position data, and sends the updated event judgment and decision model to the distributed optical fiber sensing host.

[0007] Furthermore, the risk assessment strategy includes: obtaining the radial distance, the trend of the situation, and the line grade of the pipeline under test from the risk event information; assigning values ​​to the radial distance, the trend of the situation, and the line grade respectively to obtain distance assignment, trend assignment, and grade assignment; and calculating the cumulative risk assignment using the following formula: ; in, Assigning cumulative values ​​to risks, Assign a value to the distance. Assign a value to the trend. Assign a value to the level; Based on the threshold range into which the accumulated risk value falls, a corresponding alarm level is generated.

[0008] Furthermore, the distance assignment is determined according to the following rules: When the radial distance is less than the first distance threshold, the distance is assigned the first distance value; When the radial distance is greater than or equal to the first distance threshold and less than or equal to the second distance threshold, the distance is assigned the second distance value; When the radial distance is greater than the second distance threshold, the distance is assigned a third distance value; Wherein, the first distance threshold is less than the second distance threshold, the first distance assignment is greater than the second distance assignment, and the second distance assignment is greater than the third distance assignment; The trend assignment is determined according to the following rules: When the radial distance decreases over time, it is determined to be an asymptotic trend, and the trend is assigned the first trend value. When the radial distance remains constant over time, it is determined to be a sustained trend, and the trend is assigned a second trend value. When the radial distance increases over time, it is determined to be a gradually increasing trend, and the trend is assigned the value of the third trend. Wherein, the first trend assignment is greater than the second trend assignment, and the second trend assignment is greater than the third trend assignment; The grade assignment is determined according to the business importance of the pipeline under test; the higher the business importance, the larger the grade assignment.

[0009] Furthermore, the distributed optical fiber sensing host extracts the phase change information of the signal transmitted in the optical fiber through coherent optical detection technology, so as to identify the vibration signal based on the phase change information; the vibration signal is the vibration signal caused by external destructive hazard sources along the pipeline under test.

[0010] Furthermore, the training database categorizes and stores inspection feedback data according to the scene types along the pipeline route; the scene types include at least two of the following: urban areas, farmland, hills, and mountains. The first management platform iteratively updates the assignment range and / or assignment of the distance assignment in the risk assessment strategy based on the newly added inspection feedback data in the training database, and after the update is completed, it sends the updated risk assessment strategy to the perception subsystem. The perception subsystem adjusts the judgment boundary of the radial distance in the event judgment decision model according to the updated risk assessment strategy.

[0011] Furthermore, the iterative update employs a clustering algorithm, including: Obtain inspection feedback data from the training database that is in the same scenario type as when the perception subsystem generates the risk event information, and use the radial position in the inspection feedback data confirmed by the on-site status information as a sample; the sample corresponds to the radial distance in the risk event information generated by the perception subsystem. The sample is divided into a corresponding number of data families according to a preset number of initial family cores; An iterative optimization algorithm is used to perform alternating iterative optimization on the data family. The allocation and update steps of the iterative optimization algorithm are as follows: Allocation steps: Assign each sample to the family to which the nearest family core belongs; Update steps: For each family, update the family core of the family with the mean of all samples in the family; Repeat the allocation and update steps until the preset convergence condition is met. Use the updated cores of each family as the interval boundary for distance assignment in the risk assessment strategy, so that the radial distance subsequently generated by the perception subsystem converges to the radial position confirmed by the inspection feedback data.

[0012] Furthermore, the second management platform is also used to push the risk event information and / or the on-site status information to third-party pipeline ownership units according to business agreements; the third-party pipeline ownership units include at least one of gas companies, water supply companies, and power supply companies.

[0013] Secondly, this invention proposes an automatic inspection method for pipeline facilities based on distributed optical fiber sensing and unmanned inspection feedback, which employs the system described above to perform the following steps: The sensing subsystem collects vibration signals by connecting to optical fibers laid along the pipeline under test, generates risk event information, and uploads the risk event information to the first management platform through a communication network. The first management platform receives the risk event information and generates an alarm level according to a preset risk assessment strategy; The second management platform generates inspection tasks based on the alarm levels. The scheduling subsystem receives the inspection task and schedules the inspection terminal. The inspection terminal goes to the target site to perform automatic inspection, collects on-site status information and feeds it back. The second management platform handles the event based on the on-site status information and pushes the on-site status information to the first management platform for storage in the training database; The first management platform iteratively updates the risk assessment strategy based on the inspection feedback data in the training database, and then sends the updated strategy to the perception subsystem.

[0014] Furthermore, the first management platform iteratively updates the risk assessment strategy based on the inspection feedback data in the training database, including: The inspection feedback data is categorized and stored according to the scene type along the pipeline route; the scene type includes at least two of the following: urban areas, farmland, hills, and mountains. Obtain inspection feedback data from the training database that is in the same scenario type as when the current perception subsystem generates risk event information, and use the radial position in the inspection feedback data confirmed by the on-site status information as a sample; The samples are clustered iteratively using a clustering algorithm. The updated cores of each cluster are used as the interval boundaries for distance assignment in the risk assessment strategy, so that the radial distances subsequently generated by the perception subsystem converge to the radial positions confirmed by the inspection feedback data.

[0015] The beneficial effects of this invention are as follows: 1. This application achieves province-wide network connectivity for distributed fiber optic sensing devices by deploying distributed fiber optic sensing hosts in communication equipment rooms across multiple cities and counties within a province. Each sensing host aggregates and uploads risk event information to a primary management platform via the operator's network. Compared to existing solutions where sensing devices are limited to independent deployments on single pipelines or in individual equipment rooms, this application enables unified monitoring and management of pipeline safety risk events across the entire province. Inspection resources can be uniformly scheduled according to maintenance responsibility areas, reducing the time overhead of manual hierarchical reporting and coordination.

[0016] 2. This application establishes a risk assessment strategy within a first management platform. It acquires three factors from risk event information: radial distance, situation progression trend, and line level. These factors are assigned values, and the sum is used to obtain a cumulative risk value. An alarm level is generated based on the threshold range into which the cumulative value falls. Compared to directly using sensing and detection results as the alarm criterion, multi-factor comprehensive assessment reduces the impact of single-factor measurement errors on alarm results, making the generated alarm level more closely reflect the actual security risk situation.

[0017] 3. This application establishes a differentiated inspection triggering strategy in the second management platform. For Level 1 alarms, an inspection task is immediately generated; for Level 2 alarms, a continuous observation state is established, and the cumulative risk value is recalculated every first preset time interval; for Level 3 alarms, an observation state is established for a second preset time interval, and whether to generate an inspection task is determined based on the subsequent alarm level. This strategy enables differentiated scheduling of inspection resources according to risk level, avoiding the indiscriminate dispatch of inspection tasks for all alarm events and reducing the ineffective use of inspection resources.

[0018] 4. This application categorizes and stores inspection feedback data in the training database according to the scenario type along the pipeline. The first management platform acquires inspection feedback data of the same scenario type as when the perception subsystem generates risk event information. Using the radial position confirmed by the on-site status information in the inspection feedback data as a sample, a clustering algorithm is used to iteratively cluster the samples. The updated core of each cluster is used as the interval boundary for distance assignment in the risk assessment strategy, so that the radial distance subsequently generated by the perception subsystem converges to the radial position confirmed by the inspection feedback data. This mechanism gradually reduces the positioning deviation of the perception host in each scenario as inspection data accumulates, improving the alarm accuracy of the system in actual operation.

[0019] 5. This application replaces the traditional manual inspection method by setting the inspection terminal as a drone or unmanned vehicle, and having the dispatch subsystem automatically dispatch the inspection terminal to the site to perform inspections according to the inspection task. The drone can take off to perform low-altitude shooting when the unmanned vehicle cannot obtain ground images due to obstacles, and the unmanned vehicle can provide a long-distance mobile carrying platform for the drone, enabling the system to complete the collection of ground images along the pipeline and the identification of hazard sources in different scenarios such as urban roads and suburbs.

[0020] 6. This application uses the second management platform to push risk event information and on-site confirmation information to third-party pipeline owners such as gas companies, water supply companies, and power supply companies according to business agreements, so that other pipeline owners along the pipeline can obtain the safety risk status of their pipeline surroundings in a timely manner, thus expanding the application scope of sensing and detection data. Attached Figure Description

[0021] Figure 1 This is an overall architecture diagram of the automatic pipeline facility inspection system provided in an embodiment of the present invention; Figure 2 This is a deployment architecture diagram of the distributed optical fiber sensing subsystem provided in this embodiment of the invention within a provincial area; Figure 3 A flowchart illustrating the alarm level generation and differentiated inspection triggering strategy provided in this embodiment of the invention; Figure 4 This is a flowchart illustrating the iterative updating of risk assessment strategies based on inspection feedback data, provided in an embodiment of the present invention. Figure 5 This is a diagram illustrating the configuration of an automated pipeline facility inspection system that combines distributed fiber optic sensing with an unmanned terminal in an example application scenario of an embodiment of the present invention. Detailed Implementation

[0022] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0023] Before describing the specific implementation methods of this application, for ease of understanding, the relevant terms involved in this application will be explained as follows: Distributed Fiber Optic Sensing Host: Also known as a DAS / DVS host, it is a device that uses optical fiber as the sensing medium. By injecting laser pulses into the fiber and detecting phase changes in the backscattered light, it enables distributed monitoring of vibration / vibration events along the fiber. A single host can monitor distances up to 100 kilometers and achieve event location accuracy down to the meter level.

[0024] The pipeline to be tested: In this application, it specifically refers to long-distance linear infrastructure such as communication optical cables, oil and gas pipelines, and municipal pipe networks that are buried underground or laid overhead along roads. Communication optical cables are usually laid along these lines, and the idle optical fiber cores in these cables can be used as sensing media for distributed optical fiber sensing.

[0025] Idle fiber cores: refer to fiber cores in communication optical cables that are not carrying service transmission, which can be used to connect distributed optical fiber sensing hosts as sensing channels for vibration signals.

[0026] Radial distance: refers to the vertical distance, measured in meters, between an external hazard source (such as vibration points from construction machinery) detected by the distributed fiber optic sensing host and the axis of the pipeline under test. This distance is used to assess the degree of threat posed by the hazard source to the pipeline.

[0027] Situation development trend: refers to the direction of change of the position of the external hazard source relative to the pipeline under test over time, including three states: approaching (distance decreasing), maintaining (distance unchanged), and moving away (distance increasing), used to determine whether the hazard source is approaching the pipeline.

[0028] Line grade: refers to the classification based on the importance of the services carried by the pipeline under test. The higher the grade, the more important the pipeline is, and the greater the impact and loss if it is damaged.

[0029] Risk cumulative assessment: The comprehensive score obtained by summing three factors, namely radial distance assessment, situation development trend assessment, and line level assessment, is used to quantitatively assess the severity of safety risk events. The higher the score, the greater the risk.

[0030] Alarm Level: The alarm level is generated based on the threshold range into which the cumulative risk value falls, including Level 1 alarm (high risk), Level 2 alarm (medium risk) and Level 3 alarm (low risk), which are used to determine the triggering method and response strategy of the inspection task.

[0031] Training database: Used to store risk event information collected and confirmed on-site by each inspection terminal, especially radial location data of hazardous sources confirmed during inspections. It is stored in categories according to the scenario type along the pipeline and serves as the data basis for iterative updates of risk assessment strategies.

[0032] Example 1

[0033] Please see Figure 1 An automated pipeline facility inspection system based on distributed fiber optic sensing and unmanned inspection feedback is applied to multi-domain networked scenarios of distributed sensing subsystems and joint inspection scenarios using drones and / or unmanned vehicles. The system includes: The sensing subsystem is deployed at multiple monitoring nodes. It collects vibration signals and generates risk event information by connecting to optical fibers laid along the pipeline under test. The first management platform is used to receive risk event information and generate alarm levels based on preset risk assessment strategies. The second management platform is used to generate inspection tasks based on alarm levels. The scheduling subsystem is used to receive inspection tasks and schedule inspection terminals. Inspection terminals, including drones and / or unmanned vehicles, are controlled by the dispatch subsystem to go to the target site to perform automatic inspections, collect on-site status information and provide feedback; The second management platform handles events based on on-site status information and pushes the on-site status information to the first management platform, storing it in the training database for iterative updates to the risk assessment strategy. The updated strategy is then distributed to the perception subsystem.

[0034] See Figure 1 The sensing subsystem is connected to the first management platform through the DCN network, the first management platform is connected to the second management platform through a communication network, the second management platform is connected to the dispatch subsystem through a communication network, and the dispatch subsystem is connected to the inspection terminal through a wireless public network.

[0035] See Figure 1The system also includes a business platform. The second management platform pushes risk event information and / or on-site status information to the business platform, which in turn pushes early warning information to third-party pipeline ownership units according to business agreements. The system also includes a maintenance personnel APP, which receives inspection task work orders via the public wireless network. This APP is used to monitor and track the execution status of inspection terminals and to participate in on-site verification as a backup inspection method in scenarios where inspection terminals cannot cover the area or where the risk level is the highest.

[0036] See Figure 1 In this embodiment, the communication network adopts a provincial-level hierarchical network provided by the operator. The sensing subsystem is deployed in the communication equipment rooms of various cities, districts, and counties. The distributed fiber optic sensing hosts in various cities, districts, and counties aggregate and upload risk event information to the provincial first management platform through the network management DCN network. This provincial-level networking method enables multiple sensing hosts distributed throughout the province to access the same management platform, facilitating centralized monitoring and unified scheduling of security risk events of pipeline resources throughout the province.

[0037] As a preferred option, please refer to Figure 2 The sensing subsystem includes distributed optical fiber sensing hosts deployed at multiple monitoring nodes. Each distributed optical fiber sensing host is connected to an idle optical fiber core in the optical cable associated with the pipeline under test at the corresponding monitoring node to independently collect vibration signals. The distributed optical fiber sensing host has a built-in event judgment and decision model. The event judgment and decision model is used to identify whether there are external hazards along the pipeline under test and the radial position of the external hazards relative to the pipeline under test based on vibration signals, and to generate risk event information. The training database stores radial position data confirmed by on-site status information; The first management platform iteratively updates the event judgment and decision model based on radial position data, and then distributes the updated event judgment and decision model to the distributed optical fiber sensing host.

[0038] In one optional implementation, each distributed fiber optic sensing host independently completes vibration signal acquisition and event judgment, uploading only the risk event information from the judgment results to the first management platform; the raw vibration signal data is not uploaded. Since the distributed fiber optic sensing hosts collect a large amount of data, transmitting all raw signals back to the central platform for unified processing would consume significant network bandwidth and processing resources. This implementation adopts a distributed computing model, where each sensing host completes AI event judgment locally, uploading only the judgment results, effectively reducing network transmission pressure and the load on the central platform.

[0039] The event judgment and decision-making model built into the sensing host is a classification model built on machine learning algorithms. Its input is the time-domain and / or frequency-domain characteristics of the fiber optic vibration signal, and its output is whether there is an external hazard source and the radial location of the hazard source. The initial version of this model can be distributed by the first management platform when the sensing host is deployed, and iterative updates can be made based on the inspection feedback data accumulated in the training database.

[0040] As a preferred option, see [reference] Figure 3 The risk assessment strategy includes: obtaining radial distance, event progression trend, and pipeline classification from risk event information; assigning values ​​to the radial distance, event progression trend, and pipeline classification respectively to obtain distance assignment, trend assignment, and classification assignment; and calculating the cumulative risk assignment using the following formula: ; in, Assigning cumulative values ​​to risks, Assign a value to the distance. Assigning values ​​to trends Assign a value to the level; Based on the threshold range into which the accumulated risk value falls, a corresponding alarm level is generated.

[0041] Because there is a certain measurement error between the radial distance detected by the distributed fiber optic sensing host and the actual radial distance, directly using the original detected distance as the alarm criterion may lead to false alarms or missed alarms. This implementation reduces the impact of measurement error on alarm results by segmenting and normalizing the radial distance, mapping continuous detected distance values ​​to discrete assigned scores. Simultaneously, it introduces two dimensions—the trend of the situation's progression and the line level—for comprehensive evaluation, making the generation of alarm levels more aligned with actual security management needs.

[0042] As a preferred method, the distance assignment is determined according to the following rules: When the radial distance is less than the first distance threshold, the distance is assigned the first distance value; when the radial distance is greater than or equal to the first distance threshold and less than or equal to the second distance threshold, the distance is assigned the second distance value; when the radial distance is greater than the second distance threshold, the distance is assigned the third distance value; wherein, the first distance threshold is less than the second distance threshold, the first distance value is greater than the second distance value, and the second distance value is greater than the third distance value.

[0043] For example, in one specific embodiment, the first distance threshold is set to 15 meters, the second distance threshold is set to 25 meters, the first distance is assigned a value of 2, the second distance is assigned a value of 1, and the third distance is assigned a value of 0. That is, when the radial distance of the hazard detected by the sensing host is less than 15 meters, the distance is assigned a value of 2; when the radial distance is between 15 and 25 meters, the distance is assigned a value of 1; and when the radial distance is greater than 25 meters, the distance is assigned a value of 0. This segmented assignment method transforms the continuous quantity of radial distance into three discrete levels, which matches the actual needs of inspection work. That is, the closer the distance, the greater the threat of the hazard to the pipeline, and the higher the assigned value.

[0044] As a preferred approach, trend assignment is determined according to the following rules: When the radial distance decreases over time, it is determined to be an asymptotic trend, and the trend value is assigned as the first trend value; when the radial distance remains constant over time, it is determined to be a maintaining trend, and the trend value is assigned as the second trend value; when the radial distance increases over time, it is determined to be an asymptotic trend, and the trend value is assigned as the third trend value; wherein, the first trend value is greater than the second trend value, and the second trend value is greater than the third trend value.

[0045] For example, in one specific embodiment, the first trend is assigned a value of 2, the second trend is assigned a value of 1, and the third trend is assigned a value of 0. That is, when the hazard source is approaching the pipeline (radial distance decreasing), the trend is assigned a value of 2; when the location of the hazard source is relatively stable (radial distance unchanged), the trend is assigned a value of 1; and when the hazard source is moving away from the pipeline (radial distance increasing), the trend is assigned a value of 0. The trend of the situation reflects the development trend of the hazard source, and hazard sources approaching the pipeline should receive higher attention than hazard sources moving away from the pipeline.

[0046] The sensing host detects the same hazard source multiple times, obtains the radial distance value of the hazard source at different times, arranges the radial distance values ​​at each time in chronological order, and compares the direction of change of the distance values ​​at adjacent times. This allows it to determine whether the radial distance decreases, remains unchanged, or increases over time, thereby determining the trend of the situation.

[0047] As a preferred approach, the grade assignment is determined according to the business importance of the pipeline under test, with higher grades being assigned larger grades as the business importance increases.

[0048] For example, in one specific embodiment, the pipelines under test are divided into Level 1, Level 2, and Level 3 lines according to the importance of the services they carry, with corresponding level values ​​of 2, 1, and 0, respectively. Among them, Level 1 lines are the backbone lines carrying core services, and their impact is the greatest if damaged; Level 2 lines are lines carrying general services; and Level 3 lines are lines carrying backup or non-critical services.

[0049] In one optional implementation, the generation of alarm levels specifically includes: setting a first threshold S1, a second threshold S2, and a third threshold S3, where S1 > S2 > S3. When the cumulative risk value S ≥ S1, a Level 1 alarm is generated; when S2 ≤ S < S1, a Level 2 alarm is generated; and when S3 ≤ S < S2, a Level 3 alarm is generated. For example, in a specific embodiment, S1 is 5, S2 is 3, and S3 is 1. Then, when S is 5 or 6, a Level 1 alarm is generated; when S is 3 or 4, a Level 2 alarm is generated; and when S is 1 or 2, a Level 3 alarm is generated.

[0050] Furthermore, the second management platform generates inspection tasks based on the alarm level, specifically including: When the alarm level is Level 1, the second management platform immediately generates an inspection task and pushes it to the scheduling subsystem. The scheduling subsystem then immediately dispatches the inspection terminal to the site to perform the inspection. When the alarm level is Level 2, the second management platform does not immediately generate an inspection task but enters a continuous observation state. Every first preset time interval, it recalculates the cumulative risk value based on the updated risk event information. If the cumulative risk value subsequently rises to the Level 1 alarm condition, an inspection task is generated. If the alarm level remains at Level 2 without any upward trend, observation continues. When the alarm level is Level 3, the second management platform enters an observation state for the second preset time interval. It determines whether to generate an inspection task based on the subsequent alarm level. If the alarm level subsequently upgrades to Level 2 or Level 1, the corresponding processing procedure is initiated.

[0051] For example, in one specific embodiment, the first preset time interval is 10 minutes and the second preset time interval is 60 minutes.

[0052] This differentiated inspection triggering strategy ensures the fastest possible response to high-risk events while avoiding resource waste caused by assigning inspection tasks for every detection. For low- and medium-risk events, the decision to inspect is made based on continuous observation of their situation, maximizing the utilization efficiency of inspection resources while ensuring safety.

[0053] In one optional implementation, the second management platform is also used to trigger the generation of daily inspection tasks based on a preset inspection plan and push them to the scheduling subsystem. Daily inspection tasks and alarm-triggered inspection tasks are independent of each other, together forming a complete pipeline inspection system. The preset inspection plan can be set according to the line level and maintenance requirements, such as monthly inspections for Level 1 lines and quarterly inspections for Level 2 lines.

[0054] In one optional implementation, the inspection terminal includes a vehicle-mounted drone system comprised of an unmanned vehicle and a drone. The unmanned vehicle carries the drone along the pipeline's ground route, and the drone takes off to take pictures when the unmanned vehicle is unable to acquire ground images due to obstacles. Since road construction work in urban areas is often obstructed by roadside trees or other obstacles, drones cannot clearly capture the ground conditions from the air. The unmanned vehicle can leverage its advantage of close-range ground photography. Simultaneously, rotary-wing drones typically have a flight time of around one hour, which is insufficient for the approximately 100-kilometer pipeline inspection requirements. The unmanned vehicle, with its longer flight time, can compensate for the drone's limited range. The vehicle-mounted drone system complements the advantages of both inspection methods, enabling long-distance travel along pipelines and low-altitude photography when aerial visibility is limited, making it suitable for pipeline inspection tasks in various urban and suburban scenarios.

[0055] As a preferred embodiment, the distributed optical fiber sensing host extracts the phase change information of the signal transmitted in the optical fiber through coherent optical detection technology, so as to identify the vibration signal based on the phase change information; the vibration signal is the vibration signal caused by external destructive hazard sources along the pipeline under test.

[0056] The basic principle of coherent optical sensing technology is as follows: a distributed optical fiber sensing host injects laser pulses into an optical fiber. As the pulses propagate through the fiber, they generate backscattered light (including Brillouin scattering and Raman scattering). When the fiber is subjected to external vibration at a certain point along its path, the refractive index and length of the fiber at that location undergo slight changes, resulting in a phase change in the backscattered light. The sensing host extracts the phase change information of the backscattered light using coherent detection technology. Based on the location of the phase change, it determines the distance between the vibration point and the host, and identifies the type and intensity of the vibration based on the amplitude and frequency characteristics of the phase change. Compared to acoustic sensing technology, coherent optical sensing technology is unaffected by environmental noise, a single host can monitor distances up to 100 kilometers, and event location accuracy can reach the meter level, making it suitable for detecting hazardous sources along long-distance pipelines.

[0057] In this embodiment, the vibration signal is caused by external destructive hazards along the pipeline under test. These external destructive hazards include, but are not limited to, periodic vibrations generated by construction machinery excavation operations and transient impact vibrations generated by mechanical impacts. These vibration signals propagate through the soil or air to the optical cable laid in the same trench as or adjacent to the pipeline, causing minor mechanical deformation of the optical cable, which is then detected by the distributed optical fiber sensing host.

[0058] Preferably, the training database categorizes and stores inspection feedback data according to the scenario type along the pipeline route; the scenario type includes at least two of the following: urban areas, farmland, hills, and mountains. See Figure 4The first management platform iteratively updates the assignment range and / or assignment of distance values ​​in the risk assessment strategy based on the newly added inspection feedback data in the training database. After the update is completed, the updated risk assessment strategy is sent to the perception subsystem. The perception subsystem adjusts the judgment boundary of radial distance in the event judgment decision model according to the updated risk assessment strategy.

[0059] It is worth noting that the propagation characteristics and positioning accuracy of fiber optic sensing signals vary across different scenario types. For example, in urban scenarios where the road surface is highly hardened, the attenuation characteristics of vibration signals propagating along the ground are significantly different from those in soft farmland soil scenarios. In hilly and mountainous scenarios, the fiber optic cable may be laid according to the terrain's undulations, making the propagation path of the vibration signal more complex. Therefore, using the same set of distance assignment intervals to assess the risk of detection results across all scenarios may lead to a higher false alarm rate in some scenarios and a higher false negative rate in others. This implementation method categorizes and stores inspection feedback data according to scenario type, enabling subsequent iterative updates to be performed separately for different scenarios, thereby improving the accuracy of risk assessment in each scenario.

[0060] Correspondingly, the newly added inspection feedback data in the training database refers to the actual radial location data of the hazard source confirmed by on-site images after each on-site inspection by the inspection terminal. These data, together with the radial distance data originally reported by the sensing subsystem, form a correspondence between the detected values ​​and the actual values, providing a supervisory signal for subsequent iterative updates.

[0061] Preferably, the iterative update employs a clustering algorithm, including: Obtain inspection feedback data from the training database that is in the same scenario type as when the perception subsystem generates risk event information. Use the radial position in the inspection feedback data confirmed by the on-site status information as a sample. The sample corresponds to the radial distance in the risk event information generated by the perception subsystem. The sample is divided into a corresponding number of data families according to a preset number of initial family cores; An iterative optimization algorithm is used to perform alternating iterative optimization on the data family. The allocation and update steps of the iterative optimization algorithm are as follows: Allocation steps: Assign each sample to the family to which the nearest family core belongs; Update steps: For each family, update the family core of the family with the mean of all samples in the family; Repeat the assignment and update steps until the preset convergence condition is met. Use the updated cores of each family as the interval boundary for distance assignment in the risk assessment strategy, so that the radial distance subsequently generated by the perception subsystem converges to the radial position confirmed by the inspection feedback data.

[0062] The basic idea of ​​the clustering algorithm used in this application is as follows: For the radial position samples accumulated under the same scene type, since the samples are distributed in different radial distance intervals, such as near, medium, and far distances, the clustering algorithm can automatically divide these samples into several data clusters. The cluster core of each cluster represents the average position of the samples in that cluster, and the boundary between the cluster cores is the boundary of the updated distance assignment interval. For example, if a large number of radial positions confirmed by inspections in a certain scene are concentrated around 10 meters, 20 meters, and 35 meters, the clustering algorithm will automatically form three data clusters with cluster cores of 10 meters, 20 meters, and 35 meters, respectively. Based on this, the boundary of the distance assignment interval is adjusted to the area centered on these three values, making the subsequent alarm judgment more consistent with the actual distribution in the scene.

[0063] In one optional implementation, the clustering algorithm uses K-means clustering, and the iterative optimization algorithm uses the expectation-maximization (EM) iterative algorithm. The essence of the K-means algorithm is to minimize the sum of the squares of the Euclidean distances from each sample to its cluster core, achieved through EM iteration, such as alternating minimization in the allocation and update steps. A preset convergence condition is that the offset of the cluster cores is less than a preset convergence threshold; that is, when the positional change of each cluster core is less than the preset convergence threshold in two consecutive iterations, the iteration is considered converged, and the iteration stops.

[0064] Through the aforementioned iterative update process, the radial distance values ​​subsequently detected by the sensing subsystem will gradually converge towards the actual radial position confirmed by on-site inspections, meaning that the detection accuracy of the sensing host continuously improves during actual operation. This mechanism enables the system's alarm accuracy to continuously improve with the increase in operating time and the accumulation of inspection data, effectively reducing false alarms and invalid inspection scheduling caused by detection and positioning deviations.

[0065] In one optional implementation, the trigger condition for iterative updates is: when the number of newly added vectors of the same data family in the training database for a certain scenario type reaches a preset trigger threshold, such as 10, the risk assessment strategy iteration for that scenario type is triggered. This iterative strategy, which triggers separately according to scenario type and adaptively starts based on the amount of accumulated data, ensures both the timeliness of model updates and avoids the waste of computing resources caused by frequent iterations.

[0066] Preferably, the second management platform is also used to push risk event information and / or on-site status information to third-party pipeline owners in accordance with business agreements; the third-party pipeline owners include at least one of gas companies, water supply companies, and power supply companies.

[0067] In one alternative implementation, optical cable lines often run parallel to other municipal infrastructure pipelines such as gas pipelines, water supply pipelines, and power supply pipelines. When the distributed fiber optic sensing system detects a hazardous event such as construction or excavation, this construction activity may not only damage the communication optical cable but also threaten other pipelines along the route. The second management platform, based on pre-signed business agreements with third-party pipeline owners, pushes the detected risk event information and on-site inspection confirmation information to the relevant owners, enabling them to promptly grasp the safety risk status around their pipelines and take appropriate measures.

[0068] According to the above embodiments, this application deploys distributed optical fiber sensing devices at multiple monitoring nodes within a province, using idle optical fiber cores in the optical cables associated with the pipeline under test as sensing media to collect vibration signals along the line. Each sensing host performs event judgment locally and then uploads the results to a first management platform for unified management, thus achieving province-wide grid connection of the sensing devices. The first management platform calculates the cumulative risk value and generates an alarm level based on three factors in the risk event information: radial distance, event progression trend, and line level. The second management platform generates inspection tasks based on the alarm level, dispatching drones or unmanned vehicles to the site to perform automatic inspections and collect on-site status information, thus directly linking the inspection execution with the sensing results. The second management platform pushes the radial location data confirmed by the inspection to the first management platform for storage in the training database. The first management platform classifies and stores the feedback data according to scene types such as urban areas, farmland, hills, and mountains. It uses a clustering algorithm to iterate the clustering of radial location samples confirmed by the inspection in the same scene, using the updated core of each cluster as the boundary of the distance assignment interval. This allows the radial distances subsequently generated by the sensing host to converge to the actual confirmed location on site, thereby gradually correcting the positioning deviation of fiber optic sensing in different scenes. This solution uses on-site inspection feedback data to back-optimize the sensing judgment criteria, reducing false alarms and invalid scheduling caused by positioning deviations. It is suitable for safety inspection scenarios of long-distance pipeline facilities such as communication optical cables, oil and gas pipelines, and municipal pipe networks.

[0069] Example 2

[0070] An automatic inspection method for pipeline facilities based on distributed optical fiber sensing and unmanned inspection feedback is characterized by performing the following steps using the system in Example 1: The sensing subsystem collects vibration signals by connecting to optical fibers laid along the pipeline under test, generates risk event information, and uploads the risk event information to the first management platform through the communication network. The first management platform receives risk event information and generates alarm levels based on preset risk assessment strategies. The second management platform generates inspection tasks based on alarm levels; The scheduling subsystem receives the inspection task and schedules the inspection terminal. The inspection terminal goes to the target site to perform automatic inspection, collects on-site status information and feeds it back. The second management platform handles events based on on-site status information and pushes the on-site status information to the first management platform for storage in the training database; The first management platform iteratively updates the risk assessment strategy based on the inspection feedback data in the training database, and then distributes the updated strategy to the perception subsystem.

[0071] In one optional implementation, the specific steps for the sensing subsystem to generate risk event information include: a distributed optical fiber sensing host continuously collects phase change data of the transmitted signal in the optical fiber at a preset sampling frequency by connecting an idle optical fiber core in the optical cable corresponding to the pipeline under test; when the amplitude of the phase change data exceeds a preset detection threshold, it is determined that a vibration event has been detected, and the time-domain and frequency-domain features of the vibration event are extracted; the extracted features are input into an event judgment decision model, and the model outputs whether the vibration event is an external hazard source event and the radial distance of the hazard source relative to the pipeline; risk event information is generated based on the model output results.

[0072] In one optional implementation, the specific steps for the first management platform to generate an alarm level include: parsing three factors—radial distance, situation progression trend, and line level—from the received risk event information; assigning values ​​to the three factors respectively to obtain distance assignment, trend assignment, and level assignment; calculating the sum of the three assignments to obtain a cumulative risk assignment; comparing the cumulative risk assignment with multiple preset thresholds to determine the threshold range to which it belongs, and generating the corresponding alarm level.

[0073] In one optional implementation, the specific steps for the second management platform to generate inspection tasks include: querying a preset inspection triggering strategy table based on the received alarm level to determine the triggering method and response strategy corresponding to the alarm level; when the alarm level is a Level 1 alarm, immediately generating an inspection task work order; when the alarm level is a Level 2 alarm, starting a continuous observation timer and re-acquiring the cumulative risk value of the event every first preset time interval; when the alarm level is a Level 3 alarm, starting an interval observation timer and reassessing the risk status of the event after a second preset time interval; the generated inspection task work order contains the following information: security risk event information (including event location, event type, alarm level, etc.), line resource information (including line level, line route latitude and longitude coordinates, key node locations, etc.), and maintenance responsibility unit information, and pushing the inspection task work order to the scheduling subsystem.

[0074] In one optional implementation, the specific steps of the scheduling subsystem in scheduling the inspection terminal include: determining the maintenance responsibility area to which the task belongs based on the attribution information of the line section in the inspection task work order; sending the inspection task to the drone / unmanned vehicle scheduling platform corresponding to the maintenance responsibility area; the scheduling platform selecting the inspection terminal closest to the event site and in an idle state based on the current location and status of the inspection terminal; generating an inspection path plan and sending it to the selected inspection terminal; and the inspection terminal proceeding to the site to perform the inspection according to the planned path.

[0075] In one optional implementation, the specific steps for the inspection terminal to perform automatic inspection include: after the inspection terminal arrives at the target site according to the planned path, the unmanned vehicle travels along the pipeline ground route and collects ground image data along the pipeline through the vehicle-mounted camera; when the unmanned vehicle encounters an obstacle and cannot obtain ground images along the pipeline, the drone takes off from the unmanned vehicle and flies at low altitude to the target location to take pictures; the inspection terminal transmits the collected image data back to the dispatch platform and / or the second management platform in real time through the wireless communication network; the second management platform confirms the risk event information based on the transmitted image data and determines whether there is a real external hazard source and the actual radial location of the hazard source.

[0076] In one implementation, the first management platform iteratively updates the risk assessment strategy based on inspection feedback data in the training database, including: Inspection feedback data is categorized and stored according to the scenario type along the pipeline route; the scenario type includes at least two of the following: urban areas, farmland, hills, and mountains. Obtain inspection feedback data from the training database that is in the same scenario type as when the current perception subsystem generates risk event information, and use the radial position confirmed by the on-site status information in the inspection feedback data as a sample; Clustering algorithms are used to iterate the clustering of samples. The updated core of each cluster is used as the interval boundary for distance assignment in the risk assessment strategy, so that the radial distance subsequently generated by the sensing subsystem converges to the radial position confirmed by the inspection feedback data.

[0077] In one optional implementation, an initial cluster core is determined before performing clustering iterations using a clustering algorithm. The initial cluster core can be derived from one of the following methods: during the initial system construction phase, the initial cluster core is set based on preset empirical values; or after the system has accumulated a certain amount of inspection feedback data, the cluster core obtained from pre-clustering of stored samples is used as the initial cluster core. When using the expectation-maximization iterative algorithm for clustering optimization, the allocation step and the update step are executed alternately until the preset convergence condition is met.

[0078] Please see Figure 5In a specific application scenario, a provincial operator deploys distributed fiber optic sensing hosts in computer rooms in various cities and prefectures throughout the province. Each host is connected to an idle fiber core in the trunk optical cable to form a provincial sensing network.

[0079] Along the trunk optical cable in City A, there is municipal pipe jacking construction. Construction vibrations propagate through the soil to the adjacent optical cable, causing phase changes in the transmitted light in the optical fiber. The distributed optical fiber sensing host deployed in the City A data center extracts phase change information in real time using coherent optical detection technology. The built-in event judgment and decision model identifies it as a construction excavation event, determining that the radial distance is approximately 12 meters and shows a decreasing trend. Risk event information is generated and uploaded to the provincial centralized sensing event management platform (i.e., the first management platform) through the DCN network.

[0080] The provincial centralized incident management platform, combining the pipeline resource information (line level 1), calculates the cumulative risk value S=6, generates a level 1 alarm, and pushes it to the operation and maintenance management platform (i.e., the second management platform). The operation and maintenance management platform immediately generates an inspection work order and pushes it to the drone flight control platform and unmanned vehicle dispatch platform of city A.

[0081] The dispatch platform selects an idle unmanned vehicle-mounted drone system and issues inspection commands via public wireless network. After the unmanned vehicle travels along the ground road to the vicinity of the site, due to obstructions from roadside trees and construction barriers, the drone takes off from the unmanned vehicle platform, flies at low altitude to the target location to take pictures, and transmits the construction site images back to the dispatch platform and operation and maintenance management platform via public wireless network.

[0082] The operation and maintenance management platform confirms the existence of the construction excavation event and pushes the confirmed radial location to the provincial centralized sensing event management platform, storing it in the training database. Using this data as a sample, the provincial centralized sensing event management platform iterates through the inspection feedback data in urban scenarios using a clustering algorithm, updating the interval boundaries of distance assignments in the risk assessment strategy. This ensures that subsequent sensing and positioning gradually converges to the confirmed location, reducing false alarms and ineffective scheduling in similar scenarios. Simultaneously, the operation and maintenance management platform pushes this risk information to the business platform, which, according to the service agreement, sends a warning to the local gas company. The gas company then arranges personnel for on-site monitoring. After the construction is completed and the vibration signal disappears, the operation and maintenance management platform shuts down the alarm event, completing the inspection loop.

[0083] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0084] In addition, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0085] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. An automated pipeline facility inspection system based on distributed fiber optic sensing and unmanned inspection feedback, characterized in that, A system applicable to multi-domain networking and joint inspection scenarios of distributed sensing subsystems and UAVs and / or unmanned vehicles, the system comprising: The sensing subsystem is deployed at multiple monitoring nodes. It collects vibration signals and generates risk event information by connecting to optical fibers laid along the pipeline under test. The first management platform is used to receive the risk event information and generate alarm levels according to the preset risk assessment strategy. The second management platform is used to generate inspection tasks based on the alarm level. The scheduling subsystem is used to receive the inspection tasks and schedule the inspection terminals. Inspection terminals, including drones and / or unmanned vehicles, are controlled by the scheduling subsystem to go to the target site to perform automatic inspections, collect on-site status information and provide feedback; The second management platform handles the event based on the on-site status information, pushes the on-site status information to the first management platform, stores it in the training database for iteratively updating the risk assessment strategy, and distributes the updated strategy to the perception subsystem.

2. The automatic pipeline facility inspection system according to claim 1, characterized in that, The sensing subsystem includes distributed optical fiber sensing hosts deployed at multiple monitoring nodes. Each distributed optical fiber sensing host is connected to an idle optical fiber core in the optical cable associated with the pipeline under test at the corresponding monitoring node to independently collect vibration signals. The distributed optical fiber sensing host has a built-in event judgment and decision model. The event judgment and decision model is used to identify whether there are external hazards along the pipeline under test and the radial position of the external hazards relative to the pipeline under test based on the vibration signal, and to generate the risk event information. The training database stores radial position data confirmed by the on-site status information; The first management platform iteratively updates the event judgment and decision model based on the radial position data, and sends the updated event judgment and decision model to the distributed optical fiber sensing host.

3. The automatic pipeline facility inspection system according to claim 1, characterized in that, The risk assessment strategy includes: acquiring the radial distance, event progression trend, and line grade of the pipeline under test from the risk event information; assigning values ​​to the radial distance, event progression trend, and line grade respectively to obtain distance assignment, trend assignment, and grade assignment; and calculating the cumulative risk assignment using the following formula: ； in, Assigning cumulative values ​​to risks, Assign a value to the distance. Assign a value to the trend. Assign a value to the level; Based on the threshold range into which the accumulated risk value falls, a corresponding alarm level is generated.

4. The automatic pipeline facility inspection system according to claim 3, characterized in that, The distance assignment is determined according to the following rules: When the radial distance is less than the first distance threshold, the distance is assigned the first distance value; When the radial distance is greater than or equal to the first distance threshold and less than or equal to the second distance threshold, the distance is assigned the second distance value; When the radial distance is greater than the second distance threshold, the distance is assigned a third distance value; Wherein, the first distance threshold is less than the second distance threshold, the first distance assignment is greater than the second distance assignment, and the second distance assignment is greater than the third distance assignment; The trend assignment is determined according to the following rules: When the radial distance decreases over time, it is determined to be an asymptotic trend, and the trend is assigned the first trend value. When the radial distance remains constant over time, it is determined to be a sustained trend, and the trend is assigned a second trend value. When the radial distance increases over time, it is determined to be a gradually increasing trend, and the trend is assigned the value of the third trend. Wherein, the first trend assignment is greater than the second trend assignment, and the second trend assignment is greater than the third trend assignment; The grade assignment is determined according to the business importance of the pipeline under test; the higher the business importance, the larger the grade assignment.

5. The automatic pipeline facility inspection system according to claim 2, characterized in that, The distributed optical fiber sensing host extracts the phase change information of the signal transmitted in the optical fiber through coherent optical detection technology, so as to identify the vibration signal based on the phase change information; the vibration signal is the vibration signal caused by external destructive hazard sources along the pipeline under test.

6. The automatic pipeline facility inspection system according to claim 3, characterized in that, The training database categorizes and stores inspection feedback data according to the scenario types along the pipeline route; the scenario types include at least two of the following: urban areas, farmland, hills, and mountains. The first management platform iteratively updates the assignment range and / or assignment of the distance assignment in the risk assessment strategy based on the newly added inspection feedback data in the training database, and after the update is completed, it sends the updated risk assessment strategy to the perception subsystem. The perception subsystem adjusts the judgment boundary of the radial distance in the event judgment decision model according to the updated risk assessment strategy.

7. The automatic pipeline facility inspection system according to claim 6, characterized in that, The iterative update employs a clustering algorithm, including: Obtain inspection feedback data from the training database that is in the same scenario type as when the perception subsystem generates the risk event information, and use the radial position in the inspection feedback data confirmed by the on-site status information as a sample; the sample corresponds to the radial distance in the risk event information generated by the perception subsystem. The sample is divided into a corresponding number of data families according to a preset number of initial family cores; An iterative optimization algorithm is used to perform alternating iterative optimization on the data family. The allocation and update steps of the iterative optimization algorithm are as follows: Allocation steps: Assign each sample to the family to which the nearest family core belongs; Update steps: For each family, update the family core of the family with the mean of all samples in the family; Repeat the allocation and update steps until the preset convergence condition is met. Use the updated cores of each family as the interval boundary for distance assignment in the risk assessment strategy, so that the radial distance subsequently generated by the perception subsystem converges to the radial position confirmed by the inspection feedback data.

8. The automatic pipeline facility inspection system according to claim 1, characterized in that, The second management platform is also used to push the risk event information and / or the on-site status information to third-party pipeline ownership units according to business agreements; the third-party pipeline ownership units include at least one of gas companies, water supply companies, and power supply companies.

9. An automatic inspection method for pipeline facilities based on distributed optical fiber sensing and unmanned inspection feedback, characterized in that, The following steps are performed using the system according to any one of claims 1-8: The sensing subsystem collects vibration signals by connecting to optical fibers laid along the pipeline under test, generates risk event information, and uploads the risk event information to the first management platform through a communication network. The first management platform receives the risk event information and generates an alarm level according to a preset risk assessment strategy; The second management platform generates inspection tasks based on the alarm levels. The scheduling subsystem receives the inspection task and schedules the inspection terminal. The inspection terminal goes to the target site to perform automatic inspection, collects on-site status information and feeds it back. The second management platform handles the event based on the on-site status information and pushes the on-site status information to the first management platform for storage in the training database; The first management platform iteratively updates the risk assessment strategy based on the inspection feedback data in the training database, and then sends the updated strategy to the perception subsystem.

10. The automatic inspection method for pipeline facilities according to claim 9, characterized in that, The first management platform iteratively updates the risk assessment strategy based on the inspection feedback data in the training database, including: The inspection feedback data is categorized and stored according to the scene type along the pipeline route; the scene type includes at least two of the following: urban areas, farmland, hills, and mountains. Obtain inspection feedback data from the training database that is in the same scenario type as when the current perception subsystem generates risk event information, and use the radial position in the inspection feedback data confirmed by the on-site status information as a sample; The samples are clustered iteratively using a clustering algorithm. The updated cores of each cluster are used as the interval boundaries for distance assignment in the risk assessment strategy, so that the radial distances subsequently generated by the perception subsystem converge to the radial positions confirmed by the inspection feedback data.

Images