Underground pipe network leakage monitoring method and system based on big data analysis

Abstract

The application discloses a leakage monitoring method and system for underground pipe network based on big data analysis, relates to the technical field of intelligent operation and maintenance of urban water supply pipe network, and comprises the following steps: collecting multi-source original sensing data of target underground pipe network operation; performing space-time standardization management and standardization preprocessing on the multi-source original sensing data to obtain a standardized data set; performing leakage research and judgment on the standardized data set according to a pipe network digital twin model and a multi-source data fusion analysis algorithm to obtain a suspected leakage pipe section; performing step-by-step accurate positioning verification on the suspected leakage pipe section, and returning field verification data to iteratively optimize the leakage research and judgment capability. The application realizes a full-process closed-loop leakage detection system from sensing collection, data management, intelligent research and judgment to field verification, significantly improves the automation level and identification accuracy of underground pipe network leakage detection, effectively reduces the pipe network leakage rate, and guarantees the safe and stable operation of the urban water supply pipe network.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for urban water supply networks, specifically to a method and system for monitoring leakage in underground pipe networks based on big data analysis. Background Technology

[0002] With the rapid advancement of urbanization, the operational safety and leakage control of underground water supply networks, as critical urban infrastructure, have become core issues in water resource management. Affected by factors such as pipe aging, ground subsidence, and construction disturbance, network leakage is frequent, causing not only significant water waste but also potential secondary disasters such as road collapses. To promptly detect and locate leaks, existing technologies have evolved from traditional manual auditory inspections to intelligent detection systems based on DMA (District Metering) metering and integrating online sensing and data analysis. Representative solutions include: using acoustic sensors deployed at network nodes for leak sound signature identification; cross-analysis of multi-source data including flow, pressure, acoustics, and soil data; and large-scale surveys combining satellite remote sensing and UAV thermal imaging. These solutions have all made attempts at enhancing single-point perception or expanding data coverage, laying the technological foundation for online and automated detection of network leakage.

[0003] However, the aforementioned existing technologies still have the following significant drawbacks: First, the acquisition frequency, spatial reference, and time synchronization mechanism of multi-source sensor data are not unified, making it difficult for heterogeneous data to achieve accurate correlation and matching within a unified spatiotemporal framework. The data fusion depth is insufficient, making it difficult to support high-precision leakage analysis and judgment. Second, cloud-based big data analysis results are mostly presented in the form of regional anomaly alerts, lacking a standardized connection process with traditional on-site acoustic correlators, ground-penetrating radar, and other precise positioning methods. The judgment conclusions are difficult to directly guide excavation and repair, resulting in low efficiency in implementation. These problems severely restrict the closed-loop connection of pipeline leakage detection technology from "regional early warning" to "precise positioning." Therefore, how to overcome the problems of spatiotemporal fragmentation of multi-source data, shallow fusion analysis, and disconnect between detection and treatment in existing technologies, and achieve high-precision, highly generalized, and fully closed-loop detection of underground pipeline leakage points, is a problem we need to solve. To this end, we now provide a method and system for underground pipeline leakage monitoring based on big data analysis. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for monitoring leakage in underground pipe networks based on big data analysis.

[0005] The objective of this invention can be achieved through the following technical solution: a method for monitoring leakage in underground pipe networks based on big data analysis, comprising: Collect multi-source raw sensing data on the operation of the target underground pipeline network; The multi-source raw sensing data is subjected to spatiotemporal standardization governance and standardization preprocessing to obtain a standardized dataset; A digital twin model of the pipeline network is constructed, and leakage is assessed on the standardized dataset based on the digital twin model of the pipeline network and a multi-source data fusion analysis algorithm to identify suspected leaking pipe sections. The suspected leaking pipe sections were precisely located and verified step by step, and the on-site verification data was transmitted back to iteratively optimize the leak assessment capability.

[0006] Furthermore, the process of collecting multi-source raw sensing data from the pipeline network includes: The target underground pipe network is divided into several independent DMA zones. Electromagnetic flow meters and pressure transmitters are installed at the main inlet pipes of each DMA zone to obtain zone-level flow and pressure data. Acoustic vibration sensors were deployed at the network nodes within each DMA partition to obtain acoustic data of pipe wall vibration at the pipe segment level. Portable ground-penetrating radar, infrared thermal imager, and tracer gas detection device are configured as precision positioning auxiliary sensing equipment, and on-site verification data can be collected and retrieved as needed.

[0007] Furthermore, the spatiotemporal standardization governance includes: Edge computing gateways are deployed in each pipeline area. The BeiDou high-precision timing module built into the edge computing gateway is used to stamp all multi-source raw sensing data with millisecond-level timestamps. At the same time, the spatial coordinates of all multi-source raw sensing data are uniformly mapped to the standard coordinate system to complete the spatiotemporal benchmark alignment.

[0008] Furthermore, the standardization preprocessing includes: The standardized dataset is obtained by performing preprocessing on the multi-source raw sensing data aligned with the spatiotemporal reference through an edge computing gateway. The preprocessing includes: using the 3σ criterion to remove outliers, using linear interpolation combined with historical data fitting to complete missing data caused by communication interruption, and converting data of different formats and protocols into a structured standard format containing timestamps, spatial coordinates, data values, and device ID fields.

[0009] Furthermore, this also includes tiered data transmission and storage: Standardized datasets are uploaded to the cloud big data platform according to priority; among them, the zone-level flow and pressure condition data are uploaded in real time at the minute level, and the pipe section-level pipe wall vibration acoustic data are uploaded in batches at the hour level. The cloud uses a time-series database to store runtime sequence data and a GIS spatial database to store pipeline attribute data, and establishes a spatiotemporal correlation index.

[0010] Furthermore, the process of constructing the digital twin model of the pipeline network includes: Using pipeline as-built drawings, GIS spatial data, pipe diameter parameters, laying year, soil geological parameters, and ground load data as inputs, and combining computational fluid dynamics tools and finite element vibration simulation, a high-fidelity digital twin model that maps one-to-one with the real pipeline network is constructed.

[0011] Furthermore, the process of assessing leakage in the standardized dataset based on the pipeline network digital twin model and the multi-source data fusion analysis algorithm includes: The leak samples are generated by simulating the digital twin model of the pipeline network, and combined with the standardized dataset, suspected leaking pipe sections are identified through a three-level progressive fusion analysis at the data level, feature level, and decision level.

[0012] Furthermore, the process of identifying suspected leaky pipe sections through a three-tiered progressive fusion analysis at the data level, feature level, and decision level includes: After aligning the spatiotemporal reference, the partition-level flow rate, pressure condition data, and acoustic vibration data are mapped to a unified feature space and spliced ​​together to form a multi-source feature matrix. Principal component analysis was used to reduce the dimensionality of the multi-source feature matrix, remove redundant and interfering features, and obtain the core feature vector of leakage. For the core feature vector of the leakage, the leakage probability results are independently output by the minimum flow rate algorithm at night, the pressure gradient algorithm, and the acoustic feature matching algorithm. The three leakage probability results are weighted and fused using the DS evidence theory to obtain the leakage confidence of each pipe segment. Pipe segments with leakage confidence exceeding the preset threshold are marked as suspected leakage pipe segments and the preliminary location range is output.

[0013] Furthermore, the process of performing step-by-step precise location verification of the suspected leaking pipe section includes: Acoustic correlators were installed at the valves at both ends of the suspected leaking pipe section to collect dual-channel leakage vibration signals; The arrival time difference of the dual-channel leakage vibration signal is calculated by cross-correlation algorithm, and the location of the leak is calculated by combining the propagation speed of the vibration signal in the pipeline, thus narrowing the range of the leak to a preset interval. In the narrowed area, ground-penetrating radar is used to perform lateral scanning to identify areas with high water content in the underground soil and abnormal cavities, and to eliminate signal interference caused by geological heterogeneity and intersecting pipelines. An electronic listening rod was used to verify each suspected leak point to confirm the precise location of the leak.

[0014] Furthermore, the present invention also provides an underground pipe network leakage monitoring system based on big data analysis, characterized in that it includes: Data acquisition module: Acquires multi-source raw sensing data of pipeline network operation; Data processing module: Performs spatiotemporal standardization governance and standardization preprocessing on the multi-source raw sensing data to obtain a standardized dataset; Data analysis module: Constructs a digital twin model of the pipeline network, and uses the digital twin model of the pipeline network and a multi-source data fusion analysis algorithm to perform leakage assessment on the standardized dataset to obtain suspected leaking pipe sections; Detection and positioning module: performs step-by-step precise positioning and verification of the suspected leaking pipe section, and transmits the on-site verification data back to iteratively optimize the leak assessment capability.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention obtains a standardized dataset under a unified spatiotemporal benchmark by performing spatiotemporal standardization and preprocessing on the multi-source raw sensing data. This achieves dual alignment of the temporal and spatial benchmarks of the multi-source heterogeneous data, effectively eliminating spatiotemporal deviations and data format differences between different sensing devices. Data quality and consistency are improved through anomaly removal and missing data completion, providing reliable standardized input for subsequent leakage fusion analysis. Based on the pipeline network digital twin model and multi-source data fusion analysis algorithm, leakage assessment is performed on the standardized dataset to obtain suspected leakage pipe sections and preliminary leak detection. By supplementing gaps in real-world leakage samples with digital twin simulation, and achieving multi-dimensional cross-validation through progressive fusion of multi-source data, this invention effectively improves the anti-interference capability and identification accuracy of leakage assessment, reducing the risk of misjudgment caused by environmental noise and operating condition fluctuations. It also links with on-site detection equipment to perform step-by-step precise location verification of suspected leaking pipe sections, and transmits on-site verification data back to iteratively optimize leakage assessment capabilities. This establishes a closed-loop linkage between cloud-based assessment and on-site operations. Gradual, coarse-to-fine positioning ensures leak location accuracy, while continuous optimization of the assessment model using real-world data enables continuous iterative improvement of detection capabilities. This invention realizes a closed-loop leakage detection system covering the entire process from sensing and data acquisition, data governance, intelligent assessment to on-site verification. It significantly improves the automation level, identification accuracy, and engineering practicality of underground pipe network leakage detection, effectively reducing pipe network leakage rates and ensuring the safe and stable operation of urban water supply networks. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a schematic diagram of the method of the present invention.

[0018] Figure 2 This is a system schematic diagram of the present invention. Detailed Implementation

[0019] like Figure 1 As shown, the method for monitoring leakage in underground pipe networks based on big data analysis includes: A hierarchical multimodal sensing network is deployed in the target underground pipeline network to collect multi-source raw sensing data on pipeline operation; The multi-source raw sensing data is subjected to spatiotemporal standardization governance and standardization preprocessing to obtain a standardized dataset under a unified spatiotemporal benchmark. A digital twin model of the pipeline network is constructed. Based on the digital twin model of the pipeline network and the multi-source data fusion analysis algorithm, leakage is assessed on the standardized dataset to obtain suspected leakage pipe sections and preliminary location intervals. The suspected leaking pipe sections are precisely located and verified step by step using the on-site detection equipment, and the on-site verification data is transmitted back to iteratively optimize the leak assessment capability.

[0020] It should be further explained that, in the specific implementation process, the process of deploying a hierarchical multimodal sensing network in the target underground pipeline network and collecting multi-source raw sensing data of the pipeline network operation includes: The target pipeline network is divided into several independent DMA zones. Electromagnetic flow meters and pressure transmitters are installed at the inlet main pipe of each DMA zone to obtain zone-level flow and pressure operating data. Specifically, based on the topology, hydraulic zoning characteristics, and user water distribution patterns of the target pipeline network, the target pipeline network is first divided into several DMA zones with independent boundaries and hydraulic non-interference. At the inlet main pipe of each DMA zone, an electromagnetic flow meter with an accuracy class of 0.2 and a pressure transmitter with an accuracy class of 0.1 are installed, and the acquisition frequency is set to 1 time / minute. The acquired data is transmitted through NB-IoT wireless communication to obtain the zone-level flow and pressure condition data. Subsequently, piezoelectric acoustic vibration sensors were installed at pipe network nodes such as valves, fire hydrants, and branch pipe interfaces in each DMA zone, with a sampling frequency of not less than 1024Hz, to collect pipe section-level pipe wall vibration acoustic data. In addition, portable ground-penetrating radar, infrared thermal imager and helium tracer gas detection device are equipped as auxiliary sensing devices for precise positioning. They are only used as needed during the precise positioning stage on site and are not fixedly deployed throughout the area. They are only carried to the site for precise positioning after a suspected leaking pipe section is located to collect on-site verification data. It should be noted that the installation process of acoustic vibration sensors at the pipeline nodes in each DMA zone includes: using rigid coupling fasteners made of 304 stainless steel to tightly attach and fix the acoustic vibration sensors to the outer wall of the pipe; uniformly applying thermal grease to the contact surface between the sensor and the pipe wall to ensure the efficiency of vibration signal transmission; using Beidou RTK positioning equipment to measure the coordinates of the installation positions of all sensing devices, uniformly calibrating them to the 2000 National Geodetic Coordinate System and then entering them into the pipeline GIS system, controlling the spatial coordinate error of all devices to be no more than 0.3 meters, thereby ensuring the accuracy of subsequent spatiotemporal correlation.

[0021] It should be further explained that, in the specific implementation process, the process of performing spatiotemporal standardization governance and standardization preprocessing on the multi-source raw sensing data to obtain a standardized dataset under a unified spatiotemporal benchmark includes: First, spatiotemporal standardization is implemented for the multi-source raw sensing data: edge computing gateways are deployed in each pipeline area to access data from all sensing devices within the area; the BeiDou high-precision time synchronization module built into the gateway adds millisecond-level timestamps to all multi-source raw sensing data, unifying the global time reference; simultaneously, the spatial coordinates of all sensing data are uniformly mapped to the 2000 National Geodetic Coordinate System, achieving dual alignment of the time and spatial references; in this embodiment, the edge computing gateway adopts an industrial-grade embedded gateway, supporting multiple communication access methods such as NB-IoT, LoRa, and Ethernet; the BeiDou time synchronization module has a timing accuracy better than 10ms, ensuring that the time deviation of the data collected by different sensors is controlled within the millisecond level; Subsequently, the multi-source raw sensing data is standardized and preprocessed: the multi-source raw sensing data aligned with the spatiotemporal reference is preprocessed through the edge computing gateway to obtain the standardized dataset; the preprocessing includes: using the 3σ criterion to remove outliers caused by electromagnetic interference and instantaneous equipment failure, and using linear interpolation combined with historical data fitting to fill in missing data caused by communication interruption; The outlier determination logic of the 3σ criterion is as follows: calculate the mean of data within N consecutive acquisition cycles of the same device. with standard deviation The value falls within Data outside the specified interval is identified as outliers and removed; when filling in missing data, for short-term missing data (less than 5 minutes), linear interpolation is used. ; in, Here are the data values ​​for time t to be completed. , These are the adjacent times before and after the missing interval. , The measured data values; for long-term missing data longer than 5 minutes, the average historical data from the same device over the same period in the past 7 days was used for fitting and completion; Finally, data of different formats and protocols are uniformly converted into the JSON standard structured format and encapsulated into standardized data containing fields such as "timestamp-spatial coordinates-data value-device ID", which is denoted as the standardized dataset.

[0022] Furthermore, after standardized preprocessing, the data is also transmitted and stored in a hierarchical manner: the standardized preprocessed multi-source raw sensing data is uploaded to the cloud big data platform according to priority, wherein the zone-level flow and pressure condition data is uploaded in real time at the minute level to meet the timeliness requirements of DMA zone metering; the pipe wall vibration acoustic full waveform data is uploaded in batches at the hour level; the cloud uses the time-series database InfluxDB to store runtime sequence data and the GIS spatial database to store pipeline attribute data, and establishes a spatiotemporal correlation index based on timestamps and spatial coordinates to support the rapid retrieval of corresponding multi-source sensing data by time and location.

[0023] It should be further explained that, in the specific implementation process, the process of judging leakage based on the standardized dataset using the pipeline network digital twin model and multi-source data fusion analysis algorithm to obtain suspected leaking pipe sections and preliminary location intervals includes: Constructing a digital twin model of the pipeline network: Using pipeline as-built drawings, GIS spatial data, pipe diameter parameters, laying date, soil geological parameters, and ground load data as input parameters, firstly, a fluid model inside the pipeline is established in computational fluid dynamics tools such as ANSYS Fluent or OpenFOAM to simulate the leakage flow and pressure distribution under different leak sizes and operating pressures; simultaneously, a pipeline-soil coupled solid vibration transmission simulation model is constructed in finite element simulation tools such as COMSOL Multiphysics or ABAQUS to simulate the propagation and attenuation law of leakage-induced pipe wall vibration in the pipe and soil, obtaining vibration amplitude data at different propagation distances; finally, the fluid model and the solid vibration transmission simulation model are jointly simulated to construct a digital twin model that maps one-to-one with the real pipeline network, i.e., the pipeline network digital twin model; The leak samples are generated by simulating the digital twin model of the pipeline network, and the leak detection model is trained by the leak samples. Then, the standardized dataset is input into the trained leak detection model for three-layer progressive fusion analysis to identify suspected leaking pipe sections. Specifically, by adjusting the leak location, leak size, operating pressure, and environmental noise parameters in the model, a massive dataset of leak samples covering different pipe materials, burial depths, and operating conditions is generated through simulation. This effectively supplements the deficiencies of real fault samples. The different operating conditions include normal water supply, peak water supply, and off-peak water supply. Subsequently, the leakage detection model is trained and calibrated offline using the leakage sample dataset to determine the normal nighttime minimum flow baseline value of the nighttime minimum flow anomaly identification algorithm, the normal gradient threshold of the pressure fluctuation gradient identification algorithm, and the baseline feature vector of the voiceprint feature matching algorithm; the leakage detection model includes a data-level fusion module, a feature-level fusion module, and a decision-level fusion module. After training is complete, input the standardized dataset (preprocessed by unifying and standardizing the spatiotemporal reference) into the trained leak detection model, and perform the following fusion analysis process: Data-level fusion: Mapping the spatiotemporal aligned zone-level flow, pressure condition data, and acoustic vibration data to a unified feature space and splicing them together to form a multi-source feature matrix; Feature-level fusion: Principal component analysis is used to reduce the dimensionality of the multi-source feature matrix, remove redundant and interfering features, and extract the core feature vector of leakage. Specifically, the dimensionality reduction process of the principal component analysis is as follows: calculate the covariance matrix of the standardized multi-source feature matrix, solve for the eigenvalues ​​and eigenvectors of the covariance matrix, and select eigenvalues ​​and eigenvectors with a cumulative contribution rate of not less than 1. The first k principal component feature vectors are used to remove interference features that are irrelevant to leakage, such as vehicle vibration and normal water flow impact, and the core feature vectors that characterize the leakage are extracted. Decision-level fusion: For the core feature vector of the leakage, the leakage probability results are output independently through the minimum flow rate algorithm at night, the pressure gradient algorithm, and the acoustic feature matching algorithm. The three results are weighted and fused using DS evidence theory to obtain the leakage confidence of each pipe segment. Pipe segments with leakage confidence exceeding the preset threshold are marked as suspected leakage pipe segments and the preliminary location interval is output. Specifically, the input to the nighttime minimum flow rate algorithm is a subset of the zone-level flow time-series features from the core feature vector of leakage, namely, flow time-series data extracted from zone-level flow and pressure condition data, supplemented by pressure baseline data for the same period. The core calculation logic of the nighttime minimum flow rate algorithm is as follows: select the low water consumption period from 2:00 AM to 4:00 AM each day as the calculation window, extract the minimum inflow rate within this window as the nighttime minimum flow rate (MNF), and compare it with the baseline value of the normal nighttime minimum flow rate for the same period in history for that zone. Compare the data and calculate the relative deviation rate of the flow rate: ; in, The mean of the minimum nighttime flow rate under normal operating conditions for the partition over the past 30 days is used; the relative deviation rate is linearly mapped to the interval [0,1] to obtain the first leakage probability. ; The input to the pressure gradient algorithm is a subset of pipe segment pressure features from the core feature vector of the leakage, specifically pressure time-series data extracted from the zone-level flow and pressure condition data, combined with the corresponding pipe segment length and topological location attributes. The core calculation logic of the pressure gradient algorithm is: calculating the pressure drop per unit pipe length between adjacent measuring points along the water flow direction, i.e., the pressure gradient. ; in, Let i be the pressure gradient of the i-th pipe segment. , These are the pressure values ​​at the upstream and downstream measuring points of the pipe section. The length of the pipe section is given; subsequently, similar to the minimum flow rate algorithm at night, the real-time pressure gradient is compared with the normal gradient threshold under the same operating conditions to calculate the degree of deviation and map it to obtain the second leakage probability. ; The input to the acoustic signature matching algorithm is a subset of the pipe wall vibration acoustic features in the core feature vector of the leakage, i.e., the raw waveform data extracted from the pipe segment-level pipe wall vibration acoustic data; the core calculation logic of the acoustic signature matching algorithm is: extracting 12-dimensional Mel frequency cepstral coefficients from the acoustic features to form a real-time feature vector. This is compared with the baseline feature vector in the pre-built standard leakage acoustic signature feature library. Calculate cosine similarity: ; in, , These are the magnitudes of the real-time feature vector and the baseline feature vector, respectively; the cosine similarity value is used as the voiceprint matching degree and mapped to the [0,1] interval to obtain the third-path leakage probability. The higher the similarity, the greater the probability of leakage; The process of weighted fusion of the three results using DS evidence theory is as follows: Assume the three algorithms correspond to three independent evidence bodies, and each evidence body outputs a basic probability assignment value for the proposition that a leak exists in the pipe section. , , Fusion is performed using Dempster's combinatorial rules: ; in, , , This represents the proposition supported by each of the three independent evidence bodies, which are, in order, the minimum flow rate algorithm at night, the pressure gradient algorithm, and the voiceprint feature matching algorithm; the proposition of each evidence body can take two states: "there is a leak in the pipe section" and "there is no leak in the pipe section"; The evidence conflict coefficient is calculated by summing the probability products of all combinations where the intersection of the three propositions is an empty set, which is the sum of the probabilities of all combinations where the judgment results of the three algorithms are completely contradictory. For the integrated target proposition The basic probability allocation value is the final output pipe segment leakage confidence level; in this embodiment, the preset threshold is set to 0.8, and pipe segments with a confidence level greater than or equal to 0.8 are marked as suspected leakage pipe segments.

[0024] It should be further explained that, in the specific implementation process, the process of using on-site detection equipment to perform step-by-step precise location verification of the suspected leaking pipe section includes: Acoustic correlators were installed at the valves at both ends of the suspected leaking pipe section to collect dual-channel leakage vibration signals; The time difference between the arrival of the leakage vibration signal at both ends of the sensor is calculated by cross-correlation algorithm. Combined with the propagation speed of the vibration signal in the pipeline, the location of the leak is calculated, and the range of the leak is narrowed down to a predetermined range, such as 2-5 meters. Specifically, the cross-correlation algorithm is as follows: Let the vibration signals collected by the two sensors at both ends be respectively... and Calculate the cross-correlation function of the two signals. The calculation formula is as follows: ; When the cross-correlation function reaches its maximum value, the corresponding time delay This refers to the time difference between the arrival of the leakage signal at the two sensors; combined with the known pipe wall vibration propagation velocity under the pipe material. and the distance between the two sensors The distance from the leak point to one of the sensors is calculated, thus narrowing down the leak point range to 2-5 meters; the calculation formula is as follows: ; Subsequently, within the narrowed interval, a ground-penetrating radar with a center frequency of 400MHz was used to perform a lateral grid scan along the pipeline route. High water-bearing areas and cavitation anomalies in the underground soil were identified through radar echo profiles, and signal interference caused by other pipeline intersections was eliminated by combining the pipeline network GIS data. Specifically, the spatial coordinates, burial depth, and orientation information of other pipelines within the narrowed interval were extracted from the pipeline network GIS data, and the spatial projection positions of the high water-bearing areas and cavitation anomalies identified in the ground-penetrating radar echo profiles were compared with the spatial positions of other pipelines. The spatial projection location refers to the location range on the horizontal plane of the ground surface corresponding to the underground high water-bearing area or cavity anomaly identified after ground penetrating radar scanning. That is, the underground anomaly is vertically projected onto the ground coordinate system to form a two-dimensional anomaly area. If the projection of an abnormal area on the horizontal plane of the ground coincides with the intersection or parallel segment of the projection of any other pipeline on the horizontal plane of the ground, the abnormality is determined to be an interference caused by the intersection or parallelism of other pipelines and is eliminated; if the spatial projection of the abnormal area does not coincide with the spatial position of any known other pipeline, the abnormality is retained as a soil disturbance feature suspected to be caused by leakage. Finally, electronic listening rods were used to listen to the ground at suspected locations point by point, and the background noise of the environment was compared with the characteristic sound of the leak to finally confirm the precise location of the leak point, controlling the positioning error to within 0.5 meters, and directly guiding the excavation and repair.

[0025] It should be further explained that, in the specific implementation process, the process of transmitting on-site verification data back to iteratively optimize the leakage assessment capability includes: After on-site detection confirms the leak, maintenance personnel use mobile terminals to transmit verification data such as the final leak location coordinates, actual leakage amount, pipe diameter, and cause of damage back to the cloud-based big data platform. The platform then associates and annotates the actual leakage data with the corresponding time period's pipe section-level pipe wall vibration acoustic data and zone-level flow and pressure operating condition data, adding them to the annotation sample library. Every quarter, the newly added annotation samples are used to perform incremental training on the multi-source data fusion analysis model, iteratively optimizing the model's feature recognition weights and continuously improving the accuracy of leak assessment.

[0026] like Figure 2 As shown, the present invention also provides an underground pipe network leakage monitoring system based on big data analysis, comprising: Data acquisition module: Acquires multi-source raw sensing data of pipeline network operation; Data processing module: Performs spatiotemporal standardization governance and standardization preprocessing on the multi-source raw sensing data to obtain a standardized dataset; Data analysis module: Constructs a digital twin model of the pipeline network, and uses the digital twin model of the pipeline network and a multi-source data fusion analysis algorithm to perform leakage assessment on the standardized dataset to obtain suspected leaking pipe sections; Detection and positioning module: performs step-by-step precise positioning and verification of the suspected leaking pipe section, and transmits the on-site verification data back to iteratively optimize the leak assessment capability.

[0027] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications or equivalent substitutions made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for monitoring leakage in underground pipe networks based on big data analysis, characterized in that, include: Collect multi-source raw sensing data on the operation of the target underground pipeline network; The multi-source raw sensing data is subjected to spatiotemporal standardization governance and standardization preprocessing to obtain a standardized dataset; A digital twin model of the pipeline network is constructed, and leakage is assessed on the standardized dataset based on the digital twin model of the pipeline network and a multi-source data fusion analysis algorithm to identify suspected leaking pipe sections. The suspected leaking pipe sections were precisely located and verified step by step, and the on-site verification data was transmitted back to iteratively optimize the leak assessment capability.

2. The method for monitoring leakage in underground pipe networks based on big data analysis according to claim 1, characterized in that, The process of collecting multi-source raw sensing data from the pipeline network includes: The target underground pipe network is divided into several independent DMA zones. Electromagnetic flow meters and pressure transmitters are installed at the main inlet pipes of each DMA zone to obtain zone-level flow and pressure data. Acoustic vibration sensors were deployed at the network nodes within each DMA partition to obtain acoustic data of pipe wall vibration at the pipe segment level. Portable ground-penetrating radar, infrared thermal imager, and tracer gas detection device are configured as precision positioning auxiliary sensing equipment, and on-site verification data can be collected and retrieved as needed.

3. The method for monitoring leakage in underground pipe networks based on big data analysis according to claim 1, characterized in that, The spatiotemporal standardization governance includes: Edge computing gateways are deployed in each pipeline area. The BeiDou high-precision timing module built into the edge computing gateway is used to stamp all multi-source raw sensing data with millisecond-level timestamps. At the same time, the spatial coordinates of all multi-source raw sensing data are uniformly mapped to the standard coordinate system to complete the spatiotemporal reference alignment.

4. The method for monitoring leakage in underground pipe networks based on big data analysis according to claim 3, characterized in that, The standardized preprocessing includes: The standardized dataset is obtained by performing preprocessing on the multi-source raw sensing data aligned with the spatiotemporal reference through an edge computing gateway. The preprocessing includes: using the 3σ criterion to remove outliers, using linear interpolation combined with historical data fitting to complete missing data caused by communication interruption, and converting data of different formats and protocols into a structured standard format containing timestamps, spatial coordinates, data values, and device ID fields.

5. The method for monitoring leakage in underground pipe networks based on big data analysis according to claim 2, characterized in that, It also includes hierarchical data transmission and storage: Standardized datasets are uploaded to the cloud big data platform according to priority; among them, the zone-level flow and pressure condition data are uploaded in real time at the minute level, and the pipe section-level pipe wall vibration acoustic data are uploaded in batches at the hour level. The cloud uses a time-series database to store runtime sequence data and a GIS spatial database to store pipeline attribute data, and establishes a spatiotemporal correlation index.

6. The method for monitoring leakage in underground pipe networks based on big data analysis according to claim 1, characterized in that, The process of constructing the digital twin model of the pipeline network includes: Using pipeline as-built drawings, GIS spatial data, pipe diameter parameters, laying year, soil geological parameters, and ground load data as inputs, and combining computational fluid dynamics tools and finite element vibration simulation, a high-fidelity digital twin model that maps one-to-one with the real pipeline network is constructed.

7. The method for monitoring leakage in underground pipe networks based on big data analysis according to claim 6, characterized in that, The process of assessing leakage in the standardized dataset based on the pipeline network digital twin model and multi-source data fusion analysis algorithm includes: The leak samples are generated by simulating the digital twin model of the pipeline network, and combined with the standardized dataset, suspected leaking pipe sections are identified through a three-level progressive fusion analysis at the data level, feature level, and decision level.

8. The method for monitoring leakage in underground pipe networks based on big data analysis according to claim 7, characterized in that, The process of identifying suspected leaky pipe sections through a three-level progressive fusion analysis at the data level, feature level, and decision level includes: After aligning the spatiotemporal reference, the partition-level flow rate, pressure condition data, and acoustic vibration data are mapped to a unified feature space and spliced ​​together to form a multi-source feature matrix. Principal component analysis was used to reduce the dimensionality of the multi-source feature matrix, remove redundant and interfering features, and obtain the core feature vector of leakage. For the core feature vector of the leakage, the leakage probability results are independently output by the minimum flow rate algorithm at night, the pressure gradient algorithm, and the acoustic feature matching algorithm. The three leakage probability results are weighted and fused using DS evidence theory to obtain the leakage confidence of each pipe segment. Pipe segments with leakage confidence exceeding a preset threshold are marked as suspected leakage pipe segments.

9. The method for monitoring leakage in underground pipe networks based on big data analysis according to claim 1, characterized in that, The process of performing step-by-step precise location verification of the suspected leaking pipe section includes: Acoustic correlators were installed at the valves at both ends of the suspected leaking pipe section to collect dual-channel leakage vibration signals; The arrival time difference of the dual-channel leakage vibration signal is calculated by cross-correlation algorithm, and the location of the leak is calculated by combining the propagation speed of the vibration signal in the pipeline, thus narrowing the range of the leak to a preset interval. In the narrowed area, ground-penetrating radar is used to perform lateral scanning to identify areas with high water content in the underground soil and abnormal cavities, and to eliminate signal interference caused by geological heterogeneity and intersecting pipelines. An electronic listening rod was used to verify each suspected leak point to confirm the precise location of the leak.

10. An underground pipe network leakage monitoring system applied to the underground pipe network leakage monitoring method based on big data analysis as described in any one of claims 1 to 9, characterized in that, include: Data acquisition module: Acquires multi-source raw sensing data of pipeline network operation; Data processing module: Performs spatiotemporal standardization governance and standardization preprocessing on the multi-source raw sensing data to obtain a standardized dataset; Data analysis module: Constructs a digital twin model of the pipeline network, and uses the digital twin model of the pipeline network and a multi-source data fusion analysis algorithm to perform leakage assessment on the standardized dataset to obtain suspected leaking pipe sections; Detection and positioning module: performs step-by-step precise positioning and verification of the suspected leaking pipe section, and transmits the on-site verification data back to iteratively optimize the leak assessment capability.

Images