Intelligent inspection system and method for sealing performance of power cable pipeline

Abstract

The invention provides an intelligent inspection system and method for the sealing performance of a power cable pipeline, and belongs to the technical field of power cable pipeline operation and maintaining.The intelligent inspection system comprises a dynamic barometric disturbance unit for injecting periodic barometric signals into a closed pipe section; the distributed sensing network unit is arranged along the closed pipe section and is used for collecting dynamic pressure waveform data, a static pressure stable value, a soil resistivity value and environment temperature and humidity data; the blind source separation unit is used for preprocessing the collected data and extracting leakage characteristic data; the three-dimensional positioning unit is used for calculating three-dimensional coordinate data of a leakage point in combination with the leakage characteristic data and satellite navigation system positioning data; and the AI prediction unit outputs future leakage probability and maintenance suggestions through the leakage risk prediction model. According to the intelligent inspection system for the sealing performance of the power cable pipeline provided by the invention, the accuracy, the anti-interference capability and the operation and maintenance perspectiveness of the sealing performance detection of the cable pipeline are comprehensively improved, and the operation and maintenance efficiency is remarkably optimized.

Description

Technical Field

[0001] This invention belongs to the field of power cable pipeline operation and maintenance technology, and more specifically, it relates to an intelligent inspection system and method for the sealing performance of power cable pipelines. Background Technology

[0002] With the increasing density of urban power grids and the widespread construction of underground utility tunnels, power cable pipelines are often laid in clusters, resulting in a large number of pipelines, close spacing, and complex geological environments. Currently, most existing technologies for inspecting the sealing of power cable pipelines use static air pressure monitoring (injecting a fixed pressure and observing pressure changes). However, traditional static air pressure monitoring methods can only detect continuous leaks, with low detection rates for intermittent dynamic leaks caused by soil settlement, temperature changes, etc. Furthermore, issues such as signal interference when multiple pipelines are running in parallel, pressure attenuation errors over long distances, and the inability to pinpoint leak locations only in planar terms (lacking burial depth information) severely impact detection accuracy and maintenance efficiency. In addition, existing methods are mostly "post-incident detection," unable to predict pipeline leakage risks in advance, leading to high maintenance costs and delayed fault response. Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent inspection system and method for the sealing performance of power cable ducts, aiming to solve the problems of existing technologies in the background art for detecting the sealing performance of cable ducts, so as to improve the detection accuracy and operation and maintenance efficiency.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is: to provide an intelligent inspection system and method for the sealing performance of power cable ducts, comprising: The dynamic air pressure disturbance unit is used to inject periodically changing air pressure signals into the closed pipe section to form a dynamic excitation field and output real-time excitation parameters. A distributed sensing network unit is arranged along the closed pipe section to collect the original physical field data of the closed pipe section under the dynamic excitation field. The original physical field data includes dynamic pressure waveform data, static pressure stability value, soil resistivity value and environmental temperature and humidity data. The blind source separation unit is used to preprocess the original physical field data through blind source separation and extract leakage characteristic data of the closed pipe section; A three-dimensional positioning unit is used to calculate the three-dimensional coordinate data of the leak point based on the leak characteristic data and the positioning data of the satellite navigation system. The AI ​​prediction unit is used to output the future leakage probability and maintenance suggestions of the closed pipe section based on the real-time excitation parameters, the environmental temperature and humidity data, the three-dimensional coordinate data and the historical operation and maintenance data of the closed pipe section, through a leakage risk prediction model.

[0005] In one possible implementation, the dynamic pressure disturbance unit includes a programmable pressure generator and an adaptive flow controller. The programmable pressure generator is used to generate a sine wave or square wave pressure signal with a frequency of 0.1-10Hz and an amplitude of 0.01-0.1MPa. The adaptive flow controller is used to dynamically adjust the gas injection rate according to the pipeline volume.

[0006] In one possible implementation, the distributed sensor network unit includes multiple sets of sensor groups and a data acquisition unit. Each set of sensor groups includes a dynamic pressure sensor, a static pressure sensor, a soil resistivity sensor, and an ambient temperature and humidity sensor. The sensor groups are connected to the data acquisition unit via a wireless communication module, and the arrangement density of the sensor groups is 10-50 meters per group.

[0007] In one possible implementation, the blind source separation unit employs an improved blind source separation model based on the FastICA algorithm, which preprocesses the original physical field data by introducing the natural frequency of the closed pipe segment as prior knowledge.

[0008] In one possible implementation, the three-dimensional positioning unit includes a pressure gradient calculation module and a resistivity inversion module. The pressure gradient calculation module generates a two-dimensional pressure gradient distribution map based on the pressure difference and distance between adjacent static pressure sensors. The resistivity inversion module infers the burial depth information of the leakage point based on the change in soil resistivity value and outputs three-dimensional coordinates in combination with satellite navigation system data.

[0009] In one possible implementation, the AI ​​prediction unit employs a gradient boosting decision tree model, the training parameters of which include: a learning rate of 0.01-0.1, a number of decision trees of 100-500, a depth of 3-8 layers per tree, a minimum number of sample splits of 5-10, and an L2 regularization coefficient of 0.01-0.1.

[0010] The beneficial effects of the intelligent inspection system for the sealing performance of power cable ducts provided by this invention are as follows: Compared with the prior art, this intelligent inspection system for the sealing performance of power cable ducts constructs a dynamic excitation field by injecting periodically changing air pressure signals through a dynamic air pressure disturbance unit, effectively capturing dynamic leakage characteristics; the distributed sensor network unit synchronously collects multiple parameters and rationally arranges sensor groups to reduce the impact of signal attenuation; the blind source separation unit introduces the inherent frequency of the pipeline as prior knowledge to remove interference and improve data purity; the three-dimensional positioning unit combines air pressure gradient and resistivity inversion to achieve accurate output of the three-dimensional coordinates of the leakage point, making up for the lack of burial depth information; the AI ​​prediction unit integrates multi-dimensional data based on a gradient boosting decision tree model to achieve leakage probability prediction and maintenance suggestions in the future time period, shifting from passive detection to proactive prediction, comprehensively improving the accuracy, anti-interference ability, and operation and maintenance foresight of cable duct sealing performance inspection, and significantly optimizing operation and maintenance efficiency.

[0011] This invention also provides an intelligent inspection method for the sealing performance of power cable ducts, which uses the aforementioned intelligent inspection system for the sealing performance of power cable ducts, characterized by comprising the following steps: S1. Pipeline pretreatment and dynamic plugging: Clean the manholes at both ends of the pipeline, and use an adaptive plugger with pressure feedback to plug the pipe and construct a closed pipe section. The adaptive plugger has a built-in miniature pressure sensor to monitor the sealing performance of the plugging gap in real time. S2. Acquisition of raw physical field data: Periodic air pressure signals are injected into the closed pipe section through the dynamic air pressure disturbance unit, and the distributed sensor network unit is activated at the same time to synchronously acquire the raw physical field data of the closed pipe section. The raw physical field data includes dynamic pressure data, static pressure data, soil resistivity data and environmental data. S3. Data processing: The collected original physical field data is transmitted to the blind source separation unit to remove environmental noise and interference signals from adjacent pipes, and to extract the leakage characteristic data of the closed pipe section. S4. Leakage detection and 3D location: Static pressure data analysis is used to determine whether there is a leak in the pipeline. Combined with the phase and amplitude changes of dynamic pressure signals, the dynamic characteristics of the leak are determined. The 3D coordinates of the leak point are calculated using a 3D location unit. S5. Leakage risk prediction and report generation: The original physical field data collected in step S2, the leakage feature data extracted in step S3, the leakage judgment results output in step S4, and the three-dimensional coordinates are input into the AI ​​prediction unit. Combined with the pipeline's historical operation and maintenance data and basic parameters, the leakage probability and maintenance suggestions for the pipeline in the next 1-3 months are output.

[0012] In one possible implementation, the adaptive plugging device in step S1 includes an inflatable airbag, a pressure feedback sensor, and an electric regulating valve. The pressure feedback sensor monitors the contact pressure between the airbag and the inner wall of the pipe in real time. When the pressure is lower than a preset threshold, the electric regulating valve automatically replenishes the air.

[0013] In one possible implementation, the dynamic pressure signal in step S2 adopts a composite waveform of static bias and dynamic disturbance. The static bias pressure is 0.05-0.3MPa, and the frequency of the dynamic disturbance signal is dynamically adjusted according to the pipeline length. When the pipeline length is ≤100 meters, the frequency is 5-10Hz, and when the pipeline length is >100 meters, the frequency is 0.1-5Hz.

[0014] In one possible implementation, the leakage determination in step S4 adopts a dual threshold method. When the static pressure drop rate exceeds 0.002 MPa / h and the harmonic distortion rate of the dynamic pressure signal exceeds 5%, it is determined that there is a leak. The harmonic distortion rate is obtained by calculating the ratio of the amplitude of each harmonic of the dynamic pressure signal to the amplitude of the fundamental wave through fast Fourier transform.

[0015] The beneficial effects of the intelligent inspection method for the sealing performance of power cable ducts provided by this invention are as follows: Compared with the prior art, the intelligent inspection method for the sealing performance of power cable ducts of this invention ensures the sealing performance of the closed pipe section through pressure feedback adaptive sealing in step S1, avoiding the influence of sealing gaps on the detection results; step S2 adopts a composite waveform of static bias and dynamic disturbance, dynamically adjusting the signal frequency according to the pipe length to adapt to the detection requirements in different scenarios; step S3 effectively filters environmental noise and interference from adjacent pipes through blind source separation technology, accurately extracting leakage characteristic data; step S4 uses a dual threshold method combined with fast Fourier transform to calculate harmonic distortion rate, achieving accurate leakage determination, and then completes three-dimensional positioning of the leakage point through air pressure gradient calculation and resistivity inversion, improving positioning accuracy; step S5 integrates multi-dimensional data and historical operation and maintenance information for risk prediction, generating an inspection report containing detailed parameters and targeted suggestions, which not only solves the problems of many blind spots and large errors in traditional methods, but also realizes the optimization of the entire process from detection to prediction, from positioning to operation and maintenance guidance, reducing operation and maintenance costs and the risk of delayed fault response. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1A flowchart of an intelligent inspection method for the sealing performance of power cable ducts provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0019] The present invention provides an intelligent inspection system and method for the sealing performance of power cable ducts. The intelligent inspection system for the sealing performance of power cable ducts includes a dynamic air pressure disturbance unit, a distributed sensor network unit, a blind source separation unit, a three-dimensional positioning unit, and an AI prediction unit.

[0020] The dynamic pressure disturbance unit is used to inject periodically changing pressure signals into the closed pipe section to form a dynamic excitation field and output real-time excitation parameters. It includes a programmable pressure generator and an adaptive flow controller. The programmable pressure generator uses a servo-controlled pressure generation device, which adjusts the frequency and amplitude of the output signal through pulse width modulation (PWM) technology. It supports switching between sine and square wave waveforms and is used to generate sine or square wave pressure signals with frequencies of 0.1-10Hz and amplitudes of 0.01-0.1MPa. In applications, when the pressure signal frequency is below 0.1Hz, the dynamic signal period is too long to capture short-cycle dynamic leaks; when it is above 10Hz, the signal attenuates too quickly in the pipeline, affecting the long-distance detection effect; when the pressure signal amplitude is below 0.01MPa, the signal strength is insufficient and easily drowned out by noise; when it is above 0.1MPa, it may cause additional stress damage to the pipeline. The adaptive flow controller adopts a differential pressure flow regulation structure with a built-in flow sensor and PID controller. By monitoring the pressure change rate (dP / dt) and volume of the closed pipe section in real time, it dynamically adjusts the gas injection rate to avoid pressure overshoot that could damage the closed pipe section, while ensuring stable transmission of the dynamic signal.

[0021] A distributed sensor network unit, arranged at intervals along a closed pipe section, is used to collect raw physical field data of the closed pipe section under the dynamic excitation field. The raw physical field data includes dynamic pressure waveform data, static pressure stability value, soil resistivity value, and environmental temperature and humidity data. In application, the distributed sensor network unit includes multiple sensor groups and a data acquisition unit. Each sensor group includes at least one dynamic pressure sensor, one static pressure sensor, one soil resistivity sensor, and one environmental temperature and humidity sensor. The arrangement density of the sensor groups is 10-50 meters per group. All types of sensors in the sensor group are connected to the data acquisition unit through a wireless communication module (such as a LoRaWAN module). The data acquisition unit uses an industrial-grade edge computing gateway with built-in filtering and noise reduction algorithms (such as moving average filtering and wavelet noise reduction) to preprocess the raw data before transmission, reducing the computing pressure on the backend system.

[0022] The blind source separation unit is used to preprocess the original physical field data through blind source separation to extract leakage characteristic data of the closed pipe section. In application, the blind source separation unit adopts an improved blind source separation model based on the FastICA algorithm. First, the original physical field data of the closed pipe section is collected by sensors. This original physical field data is a mixed signal, including leakage signals from the closed pipe section, interference signals from adjacent pipes, and environmental noise. Second, the natural frequency of the closed pipe section is introduced as prior knowledge (pipes of different diameters and materials have unique acoustic natural frequencies) to preprocess the mixed signal and enhance the identification of the target signal. Finally, the FastICA algorithm is used to separate the mixed signal into independent source signals to extract the leakage characteristic data of the closed pipe section.

[0023] The 3D positioning unit is used to calculate the 3D coordinates of the leak point based on leak characteristic data and positioning data from a satellite navigation system. In application, the 3D positioning unit includes a pressure gradient calculation module and a resistivity inversion module. The pressure gradient calculation module generates a 2D pressure gradient distribution map based on the pressure difference and distance between adjacent static pressure sensors. The resistivity inversion module infers the burial depth of the leak point based on changes in soil resistivity, and then outputs the 3D coordinates of the leak point by combining this with data from the satellite navigation system.

[0024] The AI ​​prediction unit uses real-time excitation parameters from the dynamic pressure disturbance unit, ambient temperature and humidity data collected by the distributed sensor network unit, and 3D coordinate data output by the 3D positioning unit, combined with historical operation and maintenance data of the closed pipe section, to output the future leakage probability and maintenance recommendations for the closed pipe section through a leakage risk prediction model. In application, the AI ​​prediction unit employs a gradient boosting decision tree model, including a data preprocessing module, a feature engineering module, a model training module, and a prediction output module. The data preprocessing module cleans and standardizes historical monitoring data, pipeline basic information (material, pipe diameter, service life), environmental data (soil corrosivity, groundwater level, temperature and humidity), and maintenance records. The feature engineering module extracts key features, such as static pressure change rate, dynamic pressure harmonic distortion rate, soil resistivity change, and pipeline aging coefficient. The model training module uses the Gradient Boosting Decision Tree (GBDT) algorithm, with training parameters including: learning rate 0.01-0.1, number of decision trees 100-500, tree depth 3-8 layers, minimum number of sample splits 5-10, and L2 regularization coefficient 0.01-0.1. The prediction output module, based on real-time monitoring data, outputs the leakage probability (0-100%) and risk level (low: <30%; medium: 30%-60%; high: 60%-80%; extremely high: >80%) of the closed pipe section in the next 1-3 months, and recommends maintenance time windows based on the risk level (e.g., extremely high risk: maintenance within 1 week; high risk: maintenance within 1 month).

[0025] This invention provides an intelligent inspection system and method for the sealing performance of power cable ducts. Compared with existing technologies, it constructs a dynamic excitation field by injecting periodically changing air pressure signals through a dynamic air pressure disturbance unit, effectively capturing dynamic leakage characteristics; a distributed sensor network unit synchronously collects multiple parameters and rationally arranges sensor groups to reduce the impact of signal attenuation; a blind source separation unit introduces the pipeline's inherent frequency as prior knowledge to remove interference and improve data purity; a three-dimensional positioning unit combines air pressure gradient and resistivity inversion to achieve accurate output of the three-dimensional coordinates of the leakage point, compensating for the lack of burial depth information; and an AI prediction unit integrates multi-dimensional data based on a gradient boosting decision tree model to predict the leakage probability and provide maintenance suggestions for the next 1-3 months, shifting from passive detection to proactive prediction. This comprehensively improves the accuracy, anti-interference capability, and operational foresight of cable duct sealing performance inspection, significantly optimizing operational efficiency.

[0026] Please see Figure 1 The present invention also provides an intelligent inspection method for the sealing performance of power cable ducts, which uses the above-mentioned intelligent inspection system for the sealing performance of power cable ducts and includes the following steps: S1. Pipeline pretreatment and dynamic plugging: Clean the manholes at both ends of the pipeline, and use an adaptive plugger with pressure feedback to plug the pipe and construct a closed pipe section. The adaptive plugger has a built-in miniature pressure sensor to monitor the sealing performance of the plugging gap in real time.

[0027] This step specifically includes cleaning debris and accumulated water from the manholes at both ends of the pipeline before testing, checking for damage or deformation at the pipeline ports, and grinding down any sharp protrusions. An adaptive plugging device with pressure feedback is used for sealing. This plugging device includes an inflatable airbag, a pressure feedback sensor, and an electric regulating valve. During installation, the plugging device is placed at the pipeline port, and the airbag is inflated using an air pump. The pressure feedback sensor monitors the contact pressure between the airbag and the inner wall of the pipeline in real time (preset threshold is 0.1-0.3 MPa). When the pressure falls below the threshold, the electric regulating valve automatically replenishes air to ensure a tight seal. After sealing, a pre-test is performed: 0.05 MPa of compressed air is injected into the sealing gap and maintained for 5 minutes. If the pressure does not drop, the sealing is considered successful.

[0028] S2. Acquisition of raw physical field data: Periodic air pressure signals are injected into the closed pipe section through the dynamic air pressure disturbance unit, and the distributed sensor network unit is activated at the same time to synchronously acquire the raw physical field data of the closed pipe section. The raw physical field data includes dynamic pressure data, static pressure data, soil resistivity data and environmental data.

[0029] In this step, the dynamic pressure disturbance module is activated to inject a composite pressure signal of static bias and dynamic disturbance into the closed pipe section. The static bias pressure is set according to the pipe material and diameter (e.g., 0.05-0.15 MPa for PVC pipes; 0.15-0.3 MPa for steel pipes). The frequency of the dynamic disturbance signal is adjusted according to the length of the closed pipe section (e.g., 5-10 Hz for pipes less than or equal to 100 meters; 0.1-5 Hz for pipes longer than 100 meters) to prevent signal attenuation during long-distance transmission. Simultaneously, the distributed sensor network unit is activated to synchronously collect dynamic pressure, static pressure, soil resistivity, and ambient temperature and humidity data. The sampling time is 5-30 minutes (extended to 15-30 minutes for dynamic leak detection to ensure complete capture of the dynamic signal cycle). After low-pass filtering (cutoff frequency 50 Hz) and noise reduction processing, the data acquisition unit transmits the raw data to the backend system via a 4G / 5G network.

[0030] S3. Data processing: The collected original physical field data is transmitted to the blind source separation unit to remove environmental noise and interference signals from adjacent pipes, and to extract the leakage characteristic data of the closed pipe section.

[0031] Specifically, the raw physical field data acquired in step S2 is transmitted to the blind source separation unit. First, the mixed signal is converted to the frequency domain using Fourier transform. The frequency band range of the target signal is extracted based on the inherent frequency of the closed pipe section (e.g., the inherent frequency of DN150 PVC pipe is 2-5Hz), and interference signals in non-target frequency bands are filtered out. Second, the filtered signal is input into the improved FastICA algorithm, and the independence criterion of the target signal (maximizing negative entropy) is set to separate the leakage characteristic signal of the closed pipe section. Finally, the signal is converted back to the time domain using inverse Fourier transform to obtain the interference-free dynamic and static pressure signals.

[0032] S4. Leakage detection and 3D location: Static pressure data analysis is used to determine whether there is a leak in the pipeline. Combined with the phase and amplitude changes of dynamic pressure signals, the dynamic characteristics of the leak are determined. The 3D coordinates of the leak point are calculated using a 3D location unit.

[0033] In this step, a dual-threshold method is used to determine leakage: if the static pressure drop rate is >0.002 MPa / h (the threshold for persistent leakage) and the harmonic distortion rate of the dynamic pressure signal is >5% (the threshold for dynamic leakage), then the pipeline is determined to have a leak. The harmonic distortion rate is calculated using Fast Fourier Transform (FFT) to determine the ratio of each harmonic amplitude to the fundamental amplitude of the dynamic pressure signal. Dynamic leakage leads to an increase in harmonic components and a higher distortion rate.

[0034] Leak location involves two steps: Step 1: Planar positioning. Using the pressure gradient calculation module, the pressure gradient is calculated based on the pressure difference (ΔP) and installation distance (ΔL) between adjacent static pressure sensors. P=ΔP / ΔL), the location corresponding to the maximum gradient is the plane projection of the leak point, and the longitude and latitude coordinates are obtained by combining GPS / BeiDou data; Step 2: Burial Depth Location. Using the resistivity inversion module, a resistivity distribution cloud map is generated based on the measurement data from the soil resistivity sensor. The center location of the resistivity depression is identified, and the burial depth corresponding to this location is the burial depth of the leak point. Finally, the three-dimensional coordinates of the leak point are output.

[0035] S5. Leakage risk prediction and report generation: The original physical field data collected in step S2, the leakage feature data extracted in step S3, the leakage judgment results output in step S4, and the three-dimensional coordinates are input into the AI ​​prediction unit. Combined with the pipeline's historical operation and maintenance data and basic parameters, the leakage probability and maintenance suggestions for the pipeline in the next 1-3 months are output.

[0036] In this step, the leak determination results, 3D location data, detection data (static pressure change rate, dynamic pressure harmonic distortion rate, soil resistivity change), pipeline basic information (material, pipe diameter, service life), and environmental data (soil corrosivity, groundwater level, temperature and humidity) are input into the AI ​​prediction unit. The AI ​​prediction unit calls the trained gradient boosting decision tree model to calculate the probability and risk level of pipeline leakage in the next 1-3 months, and generates maintenance suggestions based on the risk level (e.g., extremely high risk: immediately arrange excavation and repair; medium risk: handle in conjunction with the annual maintenance plan).

[0037] This step also automatically generates an inspection report, which includes basic pipeline information (pipeline number, material, diameter, length, laying time), detection parameters (dynamic air pressure signal parameters, sampling duration, sensor density), detection data curves (static pressure change curve, dynamic pressure waveform, soil resistivity distribution cloud map), leak determination results, three-dimensional coordinates of the leak point, risk level, maintenance recommendations, and information such as inspection personnel and inspection time. The inspection report can be uploaded to the power operation and maintenance management platform via a wireless communication module to achieve data sharing and archiving.

[0038] This invention provides an intelligent inspection method for the sealing performance of power cable ducts. Compared with existing technologies, step S1, with its pressure feedback adaptive sealing, ensures the sealing performance of the closed pipe section and avoids the impact of sealing gaps on the detection results. Step S2 uses a composite waveform of static bias and dynamic disturbance to dynamically adjust the signal frequency according to the pipe length, adapting to the detection needs of different scenarios. Step S3 effectively filters environmental noise and interference from adjacent pipes through blind source separation technology, accurately extracting leakage characteristic data. Step S4 uses a dual threshold method combined with fast Fourier transform to calculate the harmonic distortion rate, achieving accurate leakage determination. Then, it completes the three-dimensional location of the leakage point through air pressure gradient calculation and resistivity inversion, improving the location accuracy. Step S5 integrates multi-dimensional data and historical operation and maintenance information for risk prediction, generating an inspection report containing detailed parameters and targeted suggestions. This not only solves the problems of many blind spots and large errors in traditional methods, but also achieves full-process optimization from detection to prediction, and from location to operation and maintenance guidance, reducing operation and maintenance costs and the risk of delayed fault response.

[0039] Example This embodiment focuses on a scenario of laying a cluster of 10kV underground power cables in an urban area. A DN150 PVC cable duct (5-year installation period, 120 meters long, 1.8 meters deep, surrounded by silty clay soil, with a groundwater level 3.5 meters deep) is selected beneath a main road in a certain urban area. This duct is laid parallel to three adjacent cable ducts with a spacing of only 1.2 meters, making it susceptible to signal interference from adjacent ducts. Furthermore, there has been a prior risk of intermittent leakage due to soil subsidence.

[0040] System Configuration 1. Dynamic air pressure disturbance unit Programmable air pressure generator: generates sinusoidal composite air pressure signal with a static bias pressure of 0.1MPa, a dynamic disturbance frequency of 0.5Hz (due to pipeline length > 100 meters), and an amplitude of 0.05MPa.

[0041] Adaptive flow controller: Based on the pipeline volume (approximately 2.12 m³), ​​dynamically adjusts the gas injection rate to 0.03 m³ / min to avoid gas pressure overshoot.

[0042] 2. Distributed sensor network unit Sensor group configuration: Each group includes a dynamic pressure sensor (range 0-0.5MPa, accuracy ±0.001MPa), a static pressure sensor (range 0-0.5MPa, accuracy ±0.0005MPa), and a soil resistivity sensor (measurement range 1-1000Ω). m), ambient temperature and humidity sensor (temperature -20-60℃, humidity 0-100%RH).

[0043] Deployment method: One set is deployed every 30 meters along a 120-meter pipeline, for a total of 4 sets. It communicates with the industrial-grade edge computing gateway through the LoRaWAN module, with a sampling frequency of 10Hz.

[0044] 3. Blind Source Separation Unit An improved model based on FastICA is adopted, and the natural frequency (2-5Hz) of DN150 PVC pipe is introduced as prior knowledge to perform frequency domain filtering and separation of mixed signals.

[0045] 4. Three-dimensional positioning unit The system integrates a GPS / BeiDou dual-mode positioning module, a pressure gradient calculation module that generates a two-dimensional gradient map based on the pressure difference between adjacent sensors and a 30-meter spacing, and a resistivity inversion module that infers the burial depth by analyzing changes in soil resistivity.

[0046] 5. AI Prediction Unit Gradient boosting decision tree model parameters: learning rate 0.05, number of decision trees 300, tree depth 5 layers, minimum number of sample splits 8, L2 regularization coefficient 0.05; input data includes basic information such as pipeline historical operation and maintenance records (3 routine maintenance without major faults) and soil corrosivity level (medium).

[0047] Implementation steps 1. Pipeline pretreatment and dynamic plugging (S1) Clean debris and water from the manholes at both ends of the pipeline, grind down any sharp protrusions at the ends, and install an adaptive plug with pressure feedback.

[0048] Set the airbag contact pressure preset threshold to 0.2 MPa, inject 0.05 MPa of compressed air into the sealing gap and maintain the pressure for 5 minutes. If the pressure does not drop, the sealing is qualified.

[0049] 2. Acquisition of raw physical field data (S2) The dynamic pressure disturbance unit is activated to inject a composite signal of static bias (0.1MPa) + dynamic disturbance (0.5Hz, 0.05MPa) into the closed pipe section.

[0050] The distributed sensor network is started simultaneously, and data is continuously collected for 30 minutes: dynamic pressure waveform, static pressure stability value, soil resistivity (average 85Ωm), and ambient temperature and humidity (temperature 25℃, humidity 60%RH). The data is transmitted to the backend after being filtered by the gateway through moving average.

[0051] 3. Data Processing (S3) The raw data is transmitted to the blind source separation unit, where the 2-5Hz target frequency band signal is extracted by Fourier transform, and the interference from adjacent pipes (1.8Hz signal) and environmental noise (>50Hz signal) are filtered out.

[0052] The improved FastICA algorithm separates independent source signals and extracts leakage characteristic data (dynamic pressure phase offset 0.12 rad, static pressure fluctuation 0.0015 MPa).

[0053] 4. Leakage detection and three-dimensional localization (S4) Leakage determination: The static pressure drop rate is 0.003 MPa / h (exceeding the 0.002 MPa / h threshold), and the dynamic pressure harmonic distortion rate calculated by FFT is 7.2% (exceeding the 5% threshold), indicating that a leak exists.

[0054] 3D positioning: The pressure difference between adjacent sensors is 0.008 MPa, and the plane coordinates (116°30′25″E, 39°54′18″N) are calculated from the air pressure gradient; resistivity inversion shows that the resistivity at the leak point drops to 62Ω. m, burial depth 1.75 m, output three-dimensional coordinates (116°30′25″, 39°54′18″, 1.75 m).

[0055] 5. Leakage Risk Prediction and Report Generation (S5) The detection data, leakage characteristics, and 3D coordinates are input into the AI ​​prediction unit, and combined with historical operation and maintenance data, the probability of leakage in the next 2 months is 78% (high risk level).

[0056] An inspection report is generated, and the recommended maintenance time window is within one month. It is suggested to use partial excavation to repair the leak point and simultaneously reinforce the surrounding soil to prevent further settlement.

[0057] Implementation Results Detection accuracy: The deviation between the actual excavation location of the leak point and the three-dimensional positioning coordinates is ≤0.3 meters, the dynamic leak detection rate is 100%, and there is no misjudgment due to interference from adjacent pipeline signals.

[0058] Operation and maintenance optimization: By using AI to predict and formulate maintenance plans in advance, the power outage losses caused by the escalation of faults can be avoided. Compared with traditional static detection methods, the operation and maintenance response efficiency is improved by 60% and the detection cost is reduced by 40%.

[0059] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent inspection system for the sealing performance of power cable ducts, characterized in that, include: The dynamic air pressure disturbance unit injects periodically changing air pressure signals into the closed pipe section to form a dynamic excitation field and outputs real-time excitation parameters. A distributed sensor network unit, arranged along the closed pipe section, collects raw physical field data of the closed pipe section under the dynamic excitation field. The raw physical field data includes dynamic pressure waveform data, static pressure stability value, soil resistivity value, and ambient temperature and humidity data. A blind source separation unit preprocesses the raw physical field data through blind source separation to extract leakage characteristic data of the closed pipe section. A three-dimensional positioning unit calculates the three-dimensional coordinate data of the leakage point based on the leakage characteristic data and the positioning data of the satellite navigation system. An AI prediction unit outputs the future leakage probability of the closed pipe section and maintenance suggestions based on the real-time excitation parameters, the ambient temperature and humidity data, the three-dimensional coordinate data, and the historical operation and maintenance data of the closed pipe section through a leakage risk prediction model.

2. The intelligent inspection system for the sealing performance of power cable ducts as described in claim 1, characterized in that, The dynamic pressure disturbance unit includes a programmable pressure generator and an adaptive flow controller. The programmable pressure generator is used to generate a sine wave or square wave pressure signal with a frequency of 0.1-10Hz and an amplitude of 0.01-0.1MPa. The adaptive flow controller is used to dynamically adjust the gas injection rate according to the volume of the closed pipe section.

3. The intelligent inspection system for the sealing performance of power cable ducts as described in claim 1, characterized in that, The distributed sensor network unit includes multiple sensor groups and a data acquisition unit. Each sensor group includes a dynamic pressure sensor, a static pressure sensor, a soil resistivity sensor, and an ambient temperature and humidity sensor. The sensor groups are connected to the data acquisition unit through a wireless communication module. The arrangement density of the sensor groups is 10-50 meters per group.

4. The intelligent inspection system for the sealing performance of power cable ducts as described in claim 1, characterized in that, The blind source separation unit adopts an improved blind source separation model based on the FastICA algorithm, which preprocesses the original physical field data by introducing the natural frequency of the closed pipe section as prior knowledge.

5. The intelligent inspection system for the sealing performance of power cable ducts as described in claim 3, characterized in that, The three-dimensional positioning unit includes a pressure gradient calculation module and a resistivity inversion module. The pressure gradient calculation module generates a two-dimensional pressure gradient distribution map based on the pressure difference and distance between adjacent static pressure sensors. The resistivity inversion module infers the burial depth information of the leakage point based on the change in soil resistivity value and outputs three-dimensional coordinates in combination with satellite navigation system data.

6. The intelligent inspection system for the sealing performance of power cable ducts as described in claim 1, characterized in that, The AI ​​prediction unit adopts a gradient boosting decision tree model. The training parameters of the gradient boosting decision tree model include: learning rate 0.01-0.1, number of decision trees 100-500, depth of a single tree 3-8 layers, minimum number of sample splits 5-10, and L2 regularization coefficient 0.01-0.

1.

7. A method for intelligent inspection of the sealing performance of power cable ducts, using the intelligent inspection system for the sealing performance of power cable ducts as described in any one of claims 1-6, characterized in that, Includes the following steps: S1. Pipeline pretreatment and dynamic plugging: Clean the manholes at both ends of the pipeline, and use an adaptive plugger with pressure feedback to plug the pipe and construct a closed pipe section. The adaptive plugger has a built-in miniature pressure sensor to monitor the sealing performance of the plugging gap in real time. S2. Acquisition of raw physical field data: Periodic air pressure signals are injected into the closed pipe section through the dynamic air pressure disturbance unit, and the distributed sensor network unit is activated at the same time to synchronously acquire the raw physical field data of the closed pipe section. The raw physical field data includes dynamic pressure data, static pressure data, soil resistivity data and environmental data. S3. Data processing: The collected original physical field data is transmitted to the blind source separation unit to remove environmental noise and interference signals from adjacent pipes, and to extract the leakage characteristic data of the closed pipe section. S4. Leakage detection and 3D location: Static pressure data analysis is used to determine whether there is a leak in the pipeline. Combined with the phase and amplitude changes of dynamic pressure signals, the dynamic characteristics of the leak are determined. The 3D coordinates of the leak point are calculated using a 3D location unit. S5. Leakage risk prediction and report generation: The original physical field data collected in step S2, the leakage feature data extracted in step S3, the leakage judgment results output in step S4, and the three-dimensional coordinates are input into the AI ​​prediction unit. Combined with the pipeline's historical operation and maintenance data and basic parameters, the leakage probability and maintenance suggestions for the pipeline in the next 1-3 months are output.

8. The intelligent inspection method for the sealing performance of power cable ducts as described in claim 7, characterized in that, The adaptive plugging device in step S1 includes an inflatable airbag, a pressure feedback sensor, and an electric regulating valve. The pressure feedback sensor monitors the contact pressure between the airbag and the inner wall of the pipe in real time. When the pressure is lower than a preset threshold, the electric regulating valve automatically replenishes the air.

9. The intelligent inspection method for the sealing performance of power cable ducts as described in claim 7, characterized in that, The dynamic pressure signal in step S2 adopts a composite waveform of static bias and dynamic disturbance. The static bias pressure is 0.05-0.3MPa, and the frequency of the dynamic disturbance signal is dynamically adjusted according to the pipeline length. When the pipeline length is ≤100 meters, the frequency is 5-10Hz, and when the pipeline length is >100 meters, the frequency is 0.1-5Hz.

10. The intelligent inspection method for the sealing performance of power cable ducts as described in claim 7, characterized in that, The leakage determination in step S4 adopts the dual threshold method. When the static pressure drop rate exceeds 0.002 MPa / h and the harmonic distortion rate of the dynamic pressure signal exceeds 5%, it is determined that there is a leak. The harmonic distortion rate is obtained by calculating the ratio of the amplitude of each harmonic of the dynamic pressure signal to the amplitude of the fundamental wave through fast Fourier transform.

Images