A method and system for monitoring leaks in underground pipes

CN123006967APending Publication Date: 2026-10-30HENAN PROVINCIAL WATER CONSERVANCY SECOND ENG BUREAU GRP CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611235872.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-10-30

AI Technical Summary

Technical Problem

首先,地下管网覆盖范围广、管段数量众多,不同管段的管材、埋深、周边条件差异显著,单一或经验性的传感器设置方案难以在全管网范围内达到理想的监测效果

Benefits of technology

1、本发明通过结合地下管网静态信息与初始监测动态信息确定候选渗漏监测管段,并构建代表核验管段与延展监测管段的分层双路监测网络,该技术特征打破了传统方法对全网管段逐一盲目布设传感器的模式,通过对少数量化代表管段的精细化核验来推演同类延展管段的布设参数,极大降低了传感器的铺设数量与铺设成本,实现了小范围核验指导全网精准覆盖的高效运维机制。

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123006967A_ABST
    Figure CN123006967A_ABST
Patent Text Reader

Abstract

The application provides an underground pipeline leakage monitoring method and system, and relates to the technical field of underground pipelines.The method comprises the following steps: obtaining underground pipeline network information and initial leakage monitoring information; determining a plurality of representative verification pipeline segment information and a plurality of extended monitoring pipeline segment information based on the information of each candidate leakage monitoring pipeline segment; determining a plurality of candidate setting position information based on each representative verification pipeline segment information; generating a plurality of sets of front-probing sensor setting schemes based on the plurality of candidate setting position information; determining a sensor setting scheme of the representative verification pipeline segment and a sensor setting scheme of the extended monitoring pipeline segment based on the plurality of sets of front-probing sensor setting schemes; and performing underground pipeline leakage monitoring based on the sensor setting scheme of the representative verification pipeline segment and the sensor setting scheme of the extended monitoring pipeline segment. The method can efficiently and accurately determine the sensor setting scheme of the leakage monitoring pipeline segment of the underground pipeline network.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of underground pipeline technology, and specifically to a method and system for monitoring leakage in underground pipelines. Background Technology

[0002] As a fundamental infrastructure ensuring water supply, drainage, and energy transmission, the operational status of urban underground pipe networks directly impacts urban safety and resource utilization efficiency. Pipelines buried underground for extended periods are prone to leakage. Failure to detect and address leakage promptly not only wastes resources but can also trigger secondary disasters such as road collapses. To improve the timeliness and accuracy of leakage detection, deploying sensors for online monitoring in key pipe sections has become a crucial industry practice. However, traditional underground pipeline leakage monitoring methods rely heavily on manual experience and standardized deployment methods in sensor placement, lacking differentiated optimization tailored to the specific characteristics of each pipe section. This traditional approach has significant limitations. First, underground pipe networks cover a wide area with numerous pipe sections, exhibiting significant differences in pipe materials, burial depth, and surrounding conditions. A single or experience-based sensor placement scheme is unlikely to achieve ideal monitoring results across the entire network. Second, traditional methods often require the deployment and verification of formal sensors for each pipe section individually. This fails to effectively utilize the similarities between pipe sections or leverage the results of small-scale verification to guide the placement of sensors in other sections, resulting in a time-consuming, labor-intensive, and uneconomical process for determining the placement scheme. Furthermore, existing sensor deployments often rely on fixed locations or simple rule-based methods, lacking in-depth analysis of the correlation characteristics between detection points. This makes it difficult to extract transferable and scalable deployment patterns from the detection data obtained from verified pipe sections. This not only limits further improvements in leakage monitoring coverage and sensitivity but also leads to a waste of sensor resources and low monitoring deployment efficiency, failing to meet the needs of refined and intelligent operation and maintenance management of modern urban underground pipe networks.

[0003] Therefore, how to efficiently and accurately determine the sensor installation scheme for underground pipeline leakage monitoring sections is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem this invention addresses is how to efficiently and accurately determine the sensor installation scheme for underground pipeline leakage monitoring sections.

[0005] According to a first aspect, the present invention provides a method for monitoring leakage in underground pipelines, comprising: acquiring underground pipeline network information and initial leakage monitoring information; determining multiple candidate leakage monitoring pipe segments and information for each candidate leakage monitoring pipe segment based on the underground pipeline network information and the initial leakage monitoring information; determining multiple representative verification pipe segment information and multiple extended monitoring pipe segment information based on the information for each candidate leakage monitoring pipe segment; determining multiple candidate installation location information based on the information for each representative verification pipe segment; generating multiple sets of forward sensor installation schemes based on the multiple candidate installation location information; determining sensor installation schemes for the representative verification pipe segments and sensor installation schemes for the extended monitoring pipe segments based on the multiple sets of forward sensor installation schemes; and performing underground pipeline leakage monitoring based on the sensor installation schemes for the representative verification pipe segments and the sensor installation schemes for the extended monitoring pipe segments.

[0006] In one possible implementation, determining multiple representative verification pipe segment information and multiple extended monitoring pipe segment information based on the information of each candidate leakage monitoring pipe segment includes: clustering the information of each candidate leakage monitoring pipe segment using a clustering algorithm to obtain multiple candidate leakage monitoring pipe segment groups; determining initial representative verification pipe segment information for each candidate leakage monitoring pipe segment group and multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment based on the multiple candidate leakage monitoring pipe segment groups; and determining multiple representative verification pipe segment information and multiple extended monitoring pipe segment information based on the multiple candidate leakage monitoring pipe segment groups, the initial representative verification pipe segment information for each candidate leakage monitoring pipe segment group, and the multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment.

[0007] In one possible implementation, determining the sensor setup scheme for the representative verification pipe section and the sensor setup scheme for the extended monitoring pipe section based on the multiple sets of forward sensor setup schemes includes: controlling the robot to deploy each set of forward sensor setup schemes on multiple representative verification pipe sections and acquiring sensor data for each set of forward sensor setup schemes; determining multiple key detection point information and multiple auxiliary detection point information based on the sensor data of each set of forward sensor setup schemes; determining the sensor setup scheme for the representative verification pipe section based on the multiple key detection point information and the multiple auxiliary detection point information; determining multiple candidate formal sensor deployment information for each extended monitoring pipe section based on the multiple key detection point information, the multiple auxiliary detection point information, and the sensor setup scheme for the representative verification pipe section; and determining the sensor setup scheme for the extended monitoring pipe section based on the multiple candidate formal sensor deployment information for each extended monitoring pipe section.

[0008] In one possible implementation, determining the sensor setup scheme representing the verification pipe section based on the multiple key detection point information and the multiple auxiliary detection point information includes: constructing a sensor data map, which includes multiple nodes and edges between the nodes, the multiple nodes including multiple key detection nodes and multiple auxiliary detection nodes, the node features of the key detection nodes being key detection point information, and the node features of the auxiliary detection nodes being auxiliary detection point information; and processing the sensor data map based on a sensor setup analysis model to obtain the sensor setup scheme representing the verification pipe section.

[0009] According to a second aspect, the present invention provides an underground pipeline leakage monitoring system, comprising: an acquisition module for acquiring underground pipeline network information and initial leakage monitoring information; a candidate leakage monitoring pipe segment determination module for determining multiple candidate leakage monitoring pipe segments and information for each candidate leakage monitoring pipe segment based on the underground pipeline network information and the initial leakage monitoring information; a representative verification pipe segment determination module for determining multiple representative verification pipe segment information and multiple extended monitoring pipe segment information based on the information for each candidate leakage monitoring pipe segment; a candidate setting location determination module for determining multiple candidate setting location information based on the information for each representative verification pipe segment; a forward sensor setting scheme generation module for generating multiple sets of forward sensor setting schemes based on the multiple candidate setting location information; a sensor setting scheme determination module for determining sensor setting schemes for the representative verification pipe segments and sensor setting schemes for the extended monitoring pipe segments based on the multiple sets of forward sensor setting schemes; and a leakage monitoring module for performing underground pipeline leakage monitoring based on the sensor setting schemes for the representative verification pipe segments and the sensor setting schemes for the extended monitoring pipe segments.

[0010] In one possible implementation, the representative verification pipe segment determination module is further configured to: cluster multiple candidate leakage monitoring pipe segment groups using a clustering algorithm based on the information of each candidate leakage monitoring pipe segment; determine initial representative verification pipe segment information and multiple extended monitoring pipe segment information corresponding to each candidate leakage monitoring pipe segment group based on the multiple candidate leakage monitoring pipe segment groups; and determine multiple representative verification pipe segment information and multiple extended monitoring pipe segment information based on the multiple candidate leakage monitoring pipe segment groups, the initial representative verification pipe segment information of each candidate leakage monitoring pipe segment group, and the multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment.

[0011] In one possible implementation, the sensor setup scheme determination module is further configured to: control the robot to deploy each set of forward sensor setup schemes to multiple representative verification pipe sections, and acquire sensor data for each set of forward sensor setup schemes; determine multiple key detection point information and multiple auxiliary detection point information based on the sensor data of each set of forward sensor setup schemes; determine the sensor setup scheme for the representative verification pipe section based on the multiple key detection point information and the multiple auxiliary detection point information; determine multiple candidate formal sensor deployment information for each extended monitoring pipe section based on the multiple key detection point information, the multiple auxiliary detection point information, and the sensor setup scheme for the representative verification pipe section; and determine the sensor setup scheme for the extended monitoring pipe section based on the multiple candidate formal sensor deployment information for each extended monitoring pipe section.

[0012] In one possible implementation, determining the sensor setup scheme representing the verification pipe section based on the multiple key detection point information and the multiple auxiliary detection point information includes: constructing a sensor data map, which includes multiple nodes and edges between the nodes, the multiple nodes including multiple key detection nodes and multiple auxiliary detection nodes, the node features of the key detection nodes being key detection point information, and the node features of the auxiliary detection nodes being auxiliary detection point information; and processing the sensor data map based on a sensor setup analysis model to obtain the sensor setup scheme representing the verification pipe section.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method including: acquiring underground pipe network information and initial leakage monitoring information; determining multiple candidate leakage monitoring pipe segments and information for each candidate leakage monitoring pipe segment based on the underground pipe network information and the initial leakage monitoring information; determining multiple representative verification pipe segment information and multiple extended monitoring pipe segment information based on the information for each candidate leakage monitoring pipe segment; determining multiple candidate installation location information based on the information for each representative verification pipe segment; generating multiple sets of forward sensor installation schemes based on the multiple candidate installation location information; determining sensor installation schemes for the representative verification pipe segments and sensor installation schemes for the extended monitoring pipe segments based on the multiple sets of forward sensor installation schemes; and performing underground pipeline leakage monitoring based on the sensor installation schemes for the representative verification pipe segments and the sensor installation schemes for the extended monitoring pipe segments.

[0014] According to a fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the aforementioned method for monitoring underground pipeline leakage. The method includes: acquiring underground pipeline network information and initial leakage monitoring information; determining multiple candidate leakage monitoring pipe segments and information for each candidate leakage monitoring pipe segment based on the underground pipeline network information and the initial leakage monitoring information; determining multiple representative verification pipe segment information and multiple extended monitoring pipe segment information based on the information for each candidate leakage monitoring pipe segment; determining multiple candidate installation location information based on the information for each representative verification pipe segment; generating multiple sets of forward sensor installation schemes based on the multiple candidate installation location information; determining sensor installation schemes for the representative verification pipe segments and sensor installation schemes for the extended monitoring pipe segments based on the multiple sets of forward sensor installation schemes; and performing underground pipeline leakage monitoring based on the sensor installation schemes for the representative verification pipe segments and the sensor installation schemes for the extended monitoring pipe segments.

[0015] The present invention has at least the following beneficial effects: 1. This invention determines candidate leakage monitoring pipe sections by combining static information of underground pipe networks with initial monitoring dynamic information, and constructs a hierarchical dual-path monitoring network representing verification pipe sections and extended monitoring pipe sections. This technical feature breaks the traditional method of blindly deploying sensors one by one for each pipe section of the entire network. By refining the verification of a few quantitative representative pipe sections, the deployment parameters of similar extended pipe sections are deduced, which greatly reduces the number of sensors to be laid and the cost of laying them. It realizes an efficient operation and maintenance mechanism that guides the precise coverage of the entire network through small-scale verification.

[0016] 2. This invention generates multiple forward-exploration schemes based on candidate location information and utilizes robots to implement forward-exploration deployment and multi-source data acquisition. It can dynamically capture high-response leakage signals under different installation parameters. At the same time, it constructs a sensor data map that integrates spatial geometry and signal response characteristics. Through multi-round message passing and feature aggregation, it optimizes the sensor deployment scheme of representative pipe sections, avoiding signal blind spots and equipment redundancy caused by experience-based deployment. It significantly improves the leakage detection sensitivity and equipment utilization rate of complex pipe network nodes and high-risk areas.

[0017] 3. This invention extracts the coordination response patterns of key detection points and auxiliary detection points in representative verification pipe sections, and migrates them across pipe sections to each extended monitoring pipe section to generate candidate formal deployment information. Furthermore, it performs cross-constraint and peak-shifting parameter calibration in the global solution space. This global collaborative temporal and spatial optimization mechanism successfully eliminates the redundancy of detection points between adjacent pipe sections, avoids data link overload and transmission conflicts caused by high-frequency concurrency of monitoring data, and ensures high reliability and automated deployment capability of the entire underground pipe network monitoring. Attached Figure Description

[0018] Figure 1A schematic flowchart of an underground pipeline leakage monitoring method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the overall architecture of an underground pipeline leakage monitoring method provided in an embodiment of the present invention; Figure 3 A schematic diagram of an underground pipe network provided for an embodiment of the present invention; Figure 4 A schematic diagram of underground pipeline network operating pressure sequence data provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of a medium delivery sequence data provided in an embodiment of the present invention; Figure 6 A schematic diagram of an underground pipeline leakage risk distribution map provided for an embodiment of the present invention; Figure 7 This is a schematic diagram of a process for determining multiple representative verification pipe segment information and multiple extended monitoring pipe segment information, provided by an embodiment of the present invention. Figure 8 This is a flowchart illustrating a sensor setup scheme for determining a representative verification pipe section and a sensor setup scheme for extending a monitoring pipe section, provided as an embodiment of the present invention. Figure 9 A flowchart illustrating a sensor setup scheme for determining a representative verification pipe section, provided by an embodiment of the present invention; Figure 10 This is a schematic diagram of an underground pipeline leakage monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0020] In this embodiment of the invention, the following are provided: Figure 1 The method for monitoring leakage in underground pipelines, as shown in some embodiments, Figure 2 This is a schematic diagram of the overall architecture of an underground pipeline leakage monitoring method provided by an embodiment of the present invention. In some embodiments, Figure 1The underground pipeline leakage monitoring method shown includes steps S1 to S7: Step S1: Obtain underground pipe network information and initial leakage monitoring information.

[0021] Underground pipe networks are networks of pipes buried underground and connected in a specific manner to transport media such as water and gas. An underground pipe network consists of multiple underground pipe sections and their connections, with different underground pipe sections interconnected to form a complete transport network. Figure 3 This is a schematic diagram of an underground pipe network provided for an embodiment of the present invention.

[0022] Underground pipeline network information describes the spatial layout and structure of various pipe segments within an underground pipeline network. This information can be obtained through underground pipeline network design drawings, as-built records, pipeline network management platform records, and on-site measurement results. Underground pipeline network information may include the coverage area of ​​the underground pipeline network, the location and number of each underground pipe segment, the connection relationships between different underground pipe segments, segment length, pipe diameter, material type, burial depth, and the type of medium being transported.

[0023] Initial leakage monitoring information reflects the current operating status of underground pipeline networks, preliminary signs of leakage, and historical maintenance information. This information can be obtained by retrieving underground pipeline network operation and scheduling records, inlet and outlet metering records, and underground pipeline acoustic inspection records. Initial leakage monitoring information may include underground pipeline network operating pressure sequences, media transport volume sequences, abnormal sound records, corresponding monitoring time points, and historical maintenance records.

[0024] The operating pressure sequence of underground pipelines is a set of numerical values ​​recording the forces exerted by the transported medium on the pipe wall in chronological order of monitoring time. Figure 4 This is a schematic diagram of underground pipeline network operating pressure sequence data provided in an embodiment of the present invention.

[0025] See attached document Figure 4 , attached Figure 4 This diagram illustrates the operating pressure sequence of an underground pipeline network. The horizontal axis represents monitoring time in hours (h), and the vertical axis represents pressure values ​​in MPa. The diagram includes a pressure sequence curve of the underground pipeline network, used to show how the operating pressure of the underground pipeline network changes over monitoring time.

[0026] The media delivery sequence is a set of numerical values ​​of the flow rate of the media delivered in the pipeline per unit time, recorded in chronological order of monitoring time. Figure 5 This is a schematic diagram of a medium delivery sequence data provided in an embodiment of the present invention.

[0027] See attached document Figure 5 , attached Figure 5This diagram illustrates the sequence of media transport volume in an underground pipeline network. The horizontal axis represents monitoring time in hours (h), and the vertical axis represents media transport volume in cubic meters (m³). 3 / h. The figure includes a sequence curve of underground pipeline medium transport volume, used to show how the underground pipeline medium transport volume changes over monitoring time.

[0028] Step S2: Based on the underground pipeline network information and the initial leakage monitoring information, determine multiple candidate leakage monitoring pipe sections and information for each candidate leakage monitoring pipe section.

[0029] In some embodiments, a candidate pipe segment determination model can be used to determine multiple candidate leakage monitoring pipe segments and information for each candidate leakage monitoring pipe segment. The candidate pipe segment determination model is a Transformer model. The inputs to the candidate pipe segment determination model are the underground pipe network information and the initial leakage monitoring information, and the outputs of the candidate pipe segment determination model are multiple candidate leakage monitoring pipe segments and information for each candidate leakage monitoring segment.

[0030] The Transformer model is a learning model based on self-attention computation. It consists of an encoder and a decoder. The encoder computes the relationships between different locations in the input through a self-attention layer and performs nonlinear mapping through a feedforward network layer. The decoder generates the target output based on the encoding results. The Transformer model can simultaneously handle the pipe segment connection relationships in underground pipe networks and monitoring values ​​arranged according to acquisition time, and identify the propagation relationships of leakage anomalies between different pipe segments.

[0031] Multiple candidate leakage monitoring sections are underground pipe sections identified through a candidate section determination model that initially indicate potential underground pipeline leakage and require further verification. Information for each candidate leakage monitoring section may include its location and number, pipeline structure information, operational anomaly information, and leakage risk index.

[0032] Pipeline structure information may include the pipe length, diameter, material type, and burial depth of each candidate leakage monitoring pipe segment.

[0033] Anomaly information can include abnormal pressure fluctuations, flow imbalances, and frequency of abnormal sounds for each candidate leak monitoring pipe section.

[0034] Underground pipe network information reflects the spatial layout, connection relationships, and structural conditions of various underground pipe segments. It can limit the possible propagation of leakage anomalies along which connected underground pipe segments, and also explain the differences in signal transmission after leakage occurs in different underground pipe segments. Initial leakage monitoring information reflects pressure changes, media transport imbalances, abnormal sounds, and maintenance status that have already occurred during the current operation of the underground pipe network. It can indicate whether leakage anomalies are forming and in which underground pipe segments they may be concentrated. Underground pipe network information is used to determine the propagation range of anomalies and the impact on pipe segment structures, while initial leakage monitoring information provides current anomaly manifestations. Based on this, the Transformer model can identify underground pipe segments in the underground pipe network that require further verification.

[0035] The Transformer model can establish associative data units for various underground pipe segments in an underground pipeline network based on their connections, and map initial leakage monitoring information arranged according to acquisition time to relevant underground pipe segments. The self-attention layer in the Transformer model encoder can calculate the correlation between any underground pipe segment and its connected segments, allowing for comparison of abnormal pressure fluctuations, flow imbalances, and abnormal sounds occurring in a particular underground pipe segment with changes occurring in upstream and downstream segments within a similar timeframe. If the pressure in an underground pipe segment continues to decrease, while connected underground pipe segments show an increase in the difference in medium transport volume and an increase in the frequency of abnormal sounds, the self-attention layer will enhance the correlation between related abnormal information. The pipe diameter, material type, burial depth, and historical maintenance history of the underground pipe segments participate in the nonlinear calculations of the feedforward network layer to determine the likelihood of leakage under different pipeline structure conditions for the same monitored anomaly. Based on the degree of anomaly concentration, the duration of anomalies, the anomaly transmission relationship between adjacent underground pipe sections, and the impact of pipeline structure, the decoder identifies multiple candidate leakage monitoring pipe sections that may initially have underground pipeline leakage and require further verification. At the same time, it compiles the location and number of each candidate leakage monitoring pipe section, pipeline structure information, operational anomaly information, and leakage risk index, and finally determines the information of each candidate leakage monitoring pipe section.

[0036] In some embodiments, determining multiple candidate leakage monitoring segments based on the underground pipeline network information and the initial leakage monitoring information, and the information of each candidate leakage monitoring segment includes steps S21 to S23: Step S21: Determine the structural pressure bearing capacity of each underground pipe segment and the topological leakage susceptibility coefficient of each underground pipe segment based on the underground pipe network information.

[0037] In some embodiments, deep neural networks can be used to determine the structural pressure bearing capacity of each underground pipe segment and the topological leakage susceptibility coefficient of each underground pipe segment.

[0038] A deep neural network (DNN) is a neural network model consisting of an input layer, multiple hidden layers, and an output layer. Through neurons in the hidden layers, a deep neural network performs continuous linear calculations and nonlinear transformations on the input values, learning complex correspondences from the input content, and then generates the target result through the output layer.

[0039] The structural pressure bearing capacity of each underground pipe segment is determined by a deep neural network, which represents the level of external pressure intensity that each underground pipe segment can withstand.

[0040] The topological leakage susceptibility coefficient for each underground pipe segment is determined by a deep neural network. It is a coefficient representing the degree of leakage susceptibility of each underground pipe segment under the pipe network topology and geometric connection relationship.

[0041] The deep neural network transforms physical attributes such as pipe segment length, diameter, material type, and burial depth from underground pipe network information into continuous feature maps through the input layer. Hidden layer neurons utilize multi-layer nonlinear transformations to extract the response characteristics of different pipe materials and dimensional parameters under internal media pressure and external soil loads. For underground pipe segments with deeper burial depths, higher soil loads, and larger diameters, the model activates high-pressure-bearing features in the hidden layers; conversely, for shallower burial segments with lower loads, low-pressure-bearing features are assigned. The final output layer maps these external pressure conditions to the structural bearing capacity of each underground pipe segment. The model also processes topological geometric attributes such as the connection relationships between different underground pipe segments, the coverage area of ​​the underground pipe network, and pipe segment lengths. The model can perform sliding calculations on topological connections, analyzing the intersection density, number of connected branches, and turning angles of each underground pipe segment in the network. When an underground pipe segment is located at a complex node where multiple pipes intersect or at a steep bend at the end of a pipe network, the model determines that its spatial constraints and connection complexity in the geometric topology of the pipe network are significant. Therefore, it is inferred that the underground pipe segment is more prone to leakage in the static topology layout. Finally, the topological leakage susceptibility coefficient of each underground pipe segment is calculated and derived by the output layer.

[0042] Step S22: Based on the structural pressure bearing capacity of each underground pipe segment, the topological leakage susceptibility coefficient of each underground pipe segment, and the initial leakage monitoring information, determine the leakage risk distribution map of the underground pipe network and the cumulative score of media transport anomalies for each underground pipe segment.

[0043] In some embodiments, deep neural networks can be used to determine the leakage risk distribution map of underground pipeline networks and the cumulative score of media transport anomalies for each underground pipeline segment.

[0044] The underground pipeline leakage risk distribution map is a thermal image generated by a deep neural network that can show the distribution of the probability of leakage at various geographical locations in the entire underground pipeline network. Figure 6This is a schematic diagram of an underground pipeline network leakage risk distribution map provided in an embodiment of the present invention. The underground pipeline network leakage risk distribution map uses different colored areas and numerical labels to indicate the degree of leakage risk at different locations in the underground pipeline network.

[0045] The cumulative score for abnormal media transport in each underground pipeline segment is determined by a deep neural network. It is a value used to quantify the degree to which the operating pressure fluctuation and media transport imbalance of each underground pipeline segment deviate from the normal state over a period of time.

[0046] The deep neural network uses hidden layers to map the structural bearing capacity level and topological leakage susceptibility coefficient of each underground pipe segment to the three-dimensional spatial coordinate system of the pipeline network, constructing a static physical vulnerability background field for the entire pipeline network. The model then receives the operating pressure sequence and media delivery sequence of the underground pipeline network from the initial leakage monitoring information through a time-series processing unit, calculating the pressure drop magnitude and the inlet / outlet flow imbalance difference for each underground pipe segment within a sliding time window. When an underground pipe segment with a low structural bearing capacity level and a high topological leakage susceptibility coefficient experiences a continuous decrease in operating pressure and a reduction in media delivery at the corresponding monitoring time point in the initial leakage monitoring information, the hidden layer neurons perform multi-layer nonlinear fusion of this static topological susceptibility and dynamic imbalance characteristics, calculating the total operational deviation of the underground pipe segment in the spatiotemporal dimension. This is then decoded by a fully connected layer into a cumulative score for media delivery anomalies for each underground pipe segment, reflecting the severity of the deviation in the pipe segment's operating conditions. Deep neural networks can simultaneously incorporate abnormal sound records from initial leakage monitoring information, integrate the location of abnormal sound occurrences with the cumulative score of abnormal medium transport in each underground pipe section and the static physical vulnerability background field, and extrapolate the risk field strength distribution on the spatial plane. Finally, through the back projection operator, a leakage risk distribution map of the underground pipe network is generated, which uses different shades of color to indicate the probability of actual leakage at various points in the entire network.

[0047] Step S23: Based on the underground pipeline network leakage risk distribution map and the cumulative score of medium transport anomalies for each underground pipeline segment, determine multiple candidate leakage monitoring segments and information for each candidate leakage monitoring segment.

[0048] In some embodiments, a convolutional neural network can be used to determine multiple candidate leak monitoring segments and information for each candidate leak monitoring segment.

[0049] Convolutional Neural Networks (CNNs) consist of convolutional layers, pooling layers, and fully connected layers. Convolutional layers perform sliding calculations on two-dimensional or multi-dimensional spatial feature matrices through convolutional kernels, enabling the automatic extraction of local spatial features and high-dimensional image patterns. Fully connected layers can fuse and decode the extracted spatial features and scalar features.

[0050] The convolutional neural network (CNN) uses convolutional layers to perform a sliding scan of the underground pipeline network leakage risk distribution map, extracting the spatial center coordinates, risk boundary contours, and risk peak density of high-risk color areas. Simultaneously, fully connected layers receive the cumulative score of media transport anomalies for each underground pipe segment, aligning the cumulative score of a single segment with its risk intensity in the underground pipeline network leakage risk distribution map using dual-channel features. Hidden layer neurons establish a cross-matching mechanism between global spatial risk and individual pipe segment operational imbalances. When an underground pipe segment is located in the core area of ​​a high-risk patch in the underground pipeline network leakage risk distribution map, and its cumulative score of media transport anomalies exceeds a set high deviation threshold, the model's decision neurons assign that underground pipe segment a very high candidate retention weight. The CNN performs adaptive filtering and decision convergence in the latent space, automatically eliminating interfering pipe segments with only isolated, weak risks and no media transport anomalies, accurately identifying underground pipe segments with significant spatial risks and severe operational deviations, and then adjudicating and outputting the selected underground pipe segments as candidate leakage monitoring segments.

[0051] Step S3: Based on the information of each candidate leakage monitoring pipe segment, determine multiple representative verification pipe segment information and multiple extended monitoring pipe segment information.

[0052] In some embodiments, Figure 7 This is a flowchart illustrating a process for determining multiple representative verification pipe segment information and multiple extended monitoring pipe segment information, provided by an embodiment of the present invention. The determination of multiple representative verification pipe segment information and multiple extended monitoring pipe segment information includes steps S31 to S33: Step S31: Based on the information of each candidate leakage monitoring pipe segment, a clustering algorithm is used to cluster multiple candidate leakage monitoring pipe segment groups.

[0053] In some embodiments, the clustering algorithm is the K-means clustering algorithm.

[0054] K-means clustering is a clustering algorithm that divides multiple samples into a predetermined number of groups based on the numerical distance between samples. By repeatedly calculating the distance between each sample and its cluster center and updating the cluster center, K-means clustering ensures that samples within the same group have similar values, while samples in different groups exhibit significant numerical differences. K-means clustering can group multiple candidate leak monitoring pipe sections according to their pipe structure and leak signs. The number of groups can be preset manually.

[0055] Multiple candidate leakage monitoring pipe segment groups are formed by dividing the information of multiple candidate leakage monitoring pipe segments according to the similarity of leakage monitoring features using the K-means clustering algorithm.

[0056] As an example, specifically, the material type in the information of each candidate leak monitoring pipe segment is first numerically encoded. The K-means clustering algorithm can set a predetermined number of cluster centers in all candidate leak monitoring pipe segments and calculate the numerical distance between each candidate leak monitoring pipe segment and each cluster center. The distance calculation considers both differences in pipe structure and operational anomalies; candidate leak monitoring pipe segments with similar pipe structures and similar leakage signs are assigned to the same cluster center group. After one round of assignment, the algorithm re-determines the cluster centers based on the average value of all candidate leak monitoring pipe segment information within each group and examines the degree of numerical dispersion within each group. Then, it continues to calculate the distance between each candidate leak monitoring pipe segment and the updated cluster centers. Group assignment and cluster center updates are continuously performed alternately until the positions of each cluster center no longer change significantly. After stabilization, each group has similar pipe structure conditions and leakage anomaly manifestations, while significant differences remain between different groups, thus obtaining multiple candidate leak monitoring pipe segment groups.

[0057] Clustering the information of each candidate leak monitoring segment allows for grouping all candidate leak monitoring segments based on their similarity in pipeline structural condition and abnormal leak behavior. After clustering, segments within the same candidate leak monitoring segment group exhibit relatively consistent information characteristics, while different candidate leak monitoring segment groups retain more significant differences. By forming multiple candidate leak monitoring segment groups, segments with similar pipeline structural conditions and abnormal leak behavior can be grouped together, making the characteristic differences between different candidate leak monitoring segment groups clearer and thus reflecting the distribution of candidate leak monitoring segments in terms of pipeline structural condition and abnormal leak behavior.

[0058] Step S32: Based on the multiple candidate leakage monitoring pipe segment groups, determine the initial representative verification pipe segment information for each candidate leakage monitoring pipe segment group and the multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment.

[0059] In some embodiments, a verification segment determination model can be used to determine the initial representative verification segment information for each candidate leakage monitoring segment group and the information of multiple extended monitoring segments corresponding to each initial representative verification segment. The verification segment determination model is a deep neural network. The input to the verification segment determination model is the multiple candidate leakage monitoring segment groups, and the output of the verification segment determination model is the initial representative verification segment information for each candidate leakage monitoring segment group and the information of multiple extended monitoring segments corresponding to each initial representative verification segment.

[0060] The initial representative verification segment information for each candidate leak monitoring segment group is determined by the verification segment determination model and is used to characterize the most typical segment features within each candidate leak monitoring segment group. The initial representative verification segment information for each candidate leak monitoring segment group may include the location and number of the initial representative verification segment, pipeline structure information, operational anomaly information, and leak risk index.

[0061] Each candidate leak monitoring segment group includes an initial representative verification segment information.

[0062] The information for multiple extended monitoring segments corresponding to each initial representative verification segment is determined through a verification segment determination model. This information is used to monitor multiple segments in conjunction with each initial representative verification segment. The information for each extended monitoring segment may include the location and number of each extended monitoring segment, pipeline structure information, operational anomaly information, leakage risk index, etc.

[0063] Multiple candidate leakage monitoring pipe segment groups are presented in the form of groups, which are sets of candidate pipe segments with similar pipe structures and similar leakage signs. This clustering grouping form directly defines the similarity search boundary of leakage characteristics, so that the model can find the typical representative pipe segment that best reflects the characteristics within a specific similarity range, and identify the network of related target pipe segments that need to be monitored in the same similarity range.

[0064] After receiving data from multiple candidate leak monitoring pipe segment groups, the hidden layer of the deep neural network analyzes the common characteristics of different candidate leak monitoring pipe segments within each candidate leak monitoring pipe segment group in terms of pipeline structural status and leak anomaly performance through continuous linear calculations and nonlinear transformations. The model can deeply evaluate the deviation and dispersion of the leak risk index and operational anomaly information of each pipe segment within the group in the hidden feature space. The deep neural network can compare the similarity of each candidate leak monitoring pipe segment with the common characteristics within the corresponding candidate leak monitoring pipe segment group, and determine the verification necessity of each candidate leak monitoring pipe segment in combination with the leak risk index. Then, the candidate leak monitoring pipe segment that can reflect the main pipeline structural status and leak anomaly performance of the corresponding candidate leak monitoring pipe segment group is determined as the initial representative verification pipe segment, thus obtaining the initial representative verification pipe segment information for each candidate leak monitoring pipe segment group. The model then uses each initial representative verification pipe segment as a benchmark to analyze the similarity of other candidate leak monitoring pipe segments in the corresponding candidate leak monitoring pipe segment group with each initial representative verification pipe segment in terms of pipeline structural status and leak anomaly performance. Deep neural networks, relying on the weight updates of the inter-layer parameter matrix, can accurately calculate the subordinate and cooperative relationship and attenuation ratio between these peripheral pipe segments and the core pipe segment in terms of feature amplitude. Deep neural networks can keenly identify peripheral pipe segments that also have a certain abnormal sound frequency but have not yet reached extreme values, and which, in terms of physical topology, can form a monitoring encirclement of the core pipe segment, as supplementary observation segments. The output layer ultimately performs coordinate and structural attribute matching and decoding on these repeatedly screened and weighted pipe segments, accurately outputting initial representative verification pipe segment information to characterize the most typical pipe segment features within each candidate leakage monitoring pipe segment group, as well as information on multiple extended monitoring pipe segments used to cooperate with each initial representative verification pipe segment in monitoring.

[0065] Step S33: Based on the multiple candidate leakage monitoring pipe segment groups, the initial representative verification pipe segment information of each candidate leakage monitoring pipe segment group, and the multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment, determine multiple representative verification pipe segment information and multiple extended monitoring pipe segment information.

[0066] In some embodiments, a second verification pipe segment determination model can be used to determine multiple representative verification pipe segment information and multiple extended monitoring pipe segment information. The second verification pipe segment determination model is a deep neural network. The inputs to the second verification pipe segment determination model are the multiple candidate leakage monitoring pipe segment groups, the initial representative verification pipe segment information of each candidate leakage monitoring pipe segment group, and the multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment. The outputs of the second verification pipe segment determination model are the multiple representative verification pipe segment information and the multiple extended monitoring pipe segment information.

[0067] Multiple representative verification pipe segment information refers to specific information on multiple underground pipe segments within multiple candidate leakage monitoring pipe segment groups, representing the characteristics of multiple typical pipe segments. Each candidate leakage monitoring pipe segment group corresponds to multiple representative verification pipe segment information. The information for each representative verification pipe segment may include the location and number of each representative verification pipe segment, pipeline structure information, operational anomaly information, leakage risk index, etc.

[0068] Multiple extended monitoring segment information refers to specific information on multiple underground pipe segments used to coordinate with multiple representative verification segments for monitoring and to expand the coverage of leakage monitoring. The information for each extended monitoring segment may include the location and number of each extended monitoring segment, pipeline structure information, operational anomaly information, leakage risk index, etc.

[0069] Multiple representative verification pipe segment information and multiple extended monitoring pipe segment information are used to delineate the key detection range and extended detection range for subsequent underground pipeline leakage monitoring. Multiple representative verification pipe segments can be used to identify key pipe segments for priority sensor deployment and leakage signal detection, enabling subsequent sensor deployment schemes to select representative verification locations for different candidate leakage monitoring pipe segment groups. Multiple extended monitoring pipe segments can be used to determine the range of pipe segments for supplementary monitoring around the representative verification pipe segments, allowing subsequent sensor deployment schemes to further cover relevant underground pipe segments beyond the key detection range, improving the scope and accuracy of underground pipeline leakage monitoring.

[0070] Multiple candidate leakage monitoring pipe segment groups define the classification range of different leakage characteristics. The initial representative verification pipe segment information of each candidate leakage monitoring pipe segment group and the information of multiple extended monitoring pipe segments corresponding to each initial representative verification pipe segment provide the initially determined typical detection targets and supporting observation network. These preliminary evaluation results reveal the primary and secondary distribution and preliminary monitoring layout among pipe segments, which can support the model to perform secondary fine-tuning and layout optimization verification from a global perspective.

[0071] The deep neural network utilizes hidden layers to perform global dependency mapping on the features of each initial representative verification pipe segment and the corresponding distribution of extended monitoring pipe segments. The model's internal multi-layer nonlinear computation module performs further cross-validation on the boundary overlap areas between different pipe segment groups. If the model calculates that the initial representative verification pipe segments of two adjacent candidate leakage monitoring pipe segment groups are too close in physical space and their reflected abnormal sound frequencies are highly coupled, the network will suppress redundant typical feature expressions through a penalty weight mechanism and reselect a pipe segment with the second highest leakage risk index but wider spatial coverage within the group to replace it. This adaptive weight adjustment allows the model to overcome the local field-of-view limitations of the initial extraction. The deep neural network can simultaneously verify the spatial coverage continuity of multiple extended monitoring pipe segments corresponding to each initial representative verification pipe segment. When the model finds that some extended pipe segments cannot effectively cover the end of the pipe network connection or that there are detection blind spots, the model supplements or removes them based on the overall network topology. After dimensionality reduction and back projection of the output layer, the deep neural network can output multiple representative verification pipe segment information that eliminates redundant interference, has a more reasonable spatial distribution, and more accurate feature representativeness, as well as multiple extended monitoring pipe segment information for expanding the monitoring range.

[0072] Step S4: Determine multiple candidate setting location information based on the verification pipe section information of each representative.

[0073] In some embodiments, a candidate location determination model can be used to determine multiple candidate installation location information. The candidate location determination model is a deep neural network. The input to the candidate location determination model is information representing each verification pipe segment, and the output of the candidate location determination model is multiple candidate installation location information.

[0074] Multiple candidate placement locations are determined by a candidate location determination model and are used to identify alternative locations on each representative verification pipe segment where a forward sensor can be placed. Each candidate placement location may include the pipe segment number corresponding to the representative verification pipe segment, the spatial coordinates of the candidate placement location, the axial distance between the candidate placement location and the starting point of the pipe segment, the installation direction of the pipe surface, the width of the installable surface, the detection coverage length, and the compatible sensor type, etc.

[0075] Each representative verification pipe segment reflects its actual location, pipe structure conditions, operational anomalies, and leakage risk level. The segment's length, diameter, material type, and burial depth constrain the pipe surface area from which forward sensors can be installed and the types of sensors that can be used. The representative verification pipe segment's pressure anomaly fluctuation rate, flow imbalance, abnormal sound frequency, and leakage risk index indicate which pipe locations require more direct leakage signals. Each representative verification pipe segment's information allows the model to simultaneously assess installation feasibility and testing necessity, identifying multiple locations from the representative verification pipe segment suitable for forward sensor deployment.

[0076] The deep neural network first performs deep encoding and feature vectorization of key physical attribute parameters such as pipe length, diameter, material type, and burial depth in each representative verification pipe segment through the first feature extraction layer. Multiple nonlinear mapping mechanisms within the hidden layers perform high-frequency matching calculations and constraint checks between these pipe segment physical structure data and the physical compatibility conditions of various sensor installations. Based on the pipe structure information in each representative verification pipe segment, the deep neural network determines the degree of compatibility between different pipe locations and sensor installation conditions, and identifies the candidate location range for deploying forward sensors. Combining the location and pipe structure information of each representative verification pipe segment, the deep neural network eliminates pipe locations unsuitable for deploying forward sensors. Furthermore, the deep neural network can further mine the abnormal sound source range and the axial wave peak segment with the most intense pressure fluctuations in the operational anomaly information, and through the superposition calculation of feature maps, pinpoint the axial distance range where high-frequency abnormal sound transmission is most concentrated and where weak leakage signals are most easily captured. The deep neural network can continuously eliminate inferior candidate locations in the layer-by-layer feedforward propagation process, such as those buried too deep, making excavation difficult, or whose installable surface width does not meet the size requirements of specific compatible sensor types. The output layer transforms the effective installation points that have undergone multiple rounds of screening, fully meet the pipeline structure adaptation requirements, and can cover key high-risk leakage areas into specific spatial three-dimensional coordinates and axial distance values, and finally accurately outputs multiple candidate setting location information.

[0077] Step S5: Generate multiple forward sensor setting schemes based on the multiple candidate setting location information.

[0078] In some embodiments, a configuration scheme generation model can be used to generate multiple forward sensor configuration schemes. The configuration scheme generation model is a generative adversarial network (GAN). The input to the configuration scheme generation model is the multiple candidate configuration location information, and the output of the configuration scheme generation model is the multiple forward sensor configuration schemes.

[0079] Generative Adversarial Networks (GANs) are generative models consisting of a generator and a discriminator. The generator produces new candidate results based on the input, while the discriminator judges how closely the generated results resemble the training samples. During adversarial training, the generator and discriminator continuously adjust their parameters so that the generator output gradually approximates the numerical distribution of the training samples.

[0080] Multiple forward sensor deployment schemes are a set of alternative deployment options for conducting sensor testing on multiple representative verification pipe sections. Each forward sensor deployment scheme may include sensor location coordinates, sensor type, number of sensors, sensor installation direction, sensor sampling frequency, sensor start-up sequence, and sensor monitoring duration.

[0081] Multiple candidate location information limits the specific location, installation direction, installable surface width, detection coverage length, and compatible sensor type of each representative verification pipe section where a forward probe sensor can be installed. These location and parameter conditions define the adjustable value space of the setting scheme in terms of the number of sensors, installation method, and monitoring range, enabling the generative adversarial network to generate different setting schemes and perform differentiated combinations within this space.

[0082] In a generative adversarial network (GAN), the generator generates a combination of sensor placement coordinates, sensor type, number of sensors, and sensor installation direction based on the spatial coordinates, installation direction, available surface width, and detection coverage length of each candidate placement location from multiple candidate placement location information. The discriminator compares the generated scheme with existing reasonable placement samples to evaluate the similarity in sensor coverage and installation feasibility. If the generator's scheme has an excessive number of sensors at certain candidate placement locations, leading to overlapping detection coverage, the discriminator will output a lower similarity score. The generator will then adjust the number of sensors at the corresponding locations and reallocate the installation direction to reduce coverage overlap. If the generated scheme has insufficient detection coverage on some representative verification pipe sections, the discriminator will also output a lower similarity score. The generator will then adjust its generation parameters based on the discriminator's judgment, increasing the number of sensors at the corresponding locations or adjusting the sensor placement and installation direction to supplement uncovered pipe locations in subsequent generated schemes. The generator and discriminator continuously adjust their parameters during repeated adversarial training. The generator can gradually learn the correspondence between candidate setting location information and sensor coverage effect. Different setting schemes show obvious differences in the distribution of the number of sensors and the combination of installation directions. Finally, the generator can output multiple sets of forward sensor setting schemes for multiple candidate setting locations.

[0083] Step S6: Based on the multiple sets of forward sensor setting schemes, determine the sensor setting scheme for the representative verification pipe section and the sensor setting scheme for the extended monitoring pipe section.

[0084] In some embodiments, Figure 8 This is a flowchart illustrating the process of determining a sensor setup scheme for a representative verification pipe section and a sensor setup scheme for an extended monitoring pipe section, as provided in an embodiment of the present invention. The determination of the sensor setup scheme for the representative verification pipe section and the sensor setup scheme for the extended monitoring pipe section includes steps S61 to S65: Step S61: Control the robot to deploy each set of forward sensor setup schemes in multiple representative verification pipe sections and acquire sensor data for each set of forward sensor setup schemes.

[0085] The sensor data for each set of forward sensor configuration schemes is obtained by controlling the robot to install the corresponding sensors on multiple representative verification pipe sections according to each set of forward sensor configuration schemes, and the data collected by the corresponding sensors during the testing and monitoring process.

[0086] The sensor data for each forward sensor setup scheme may include the sensor setup location coordinates, sensor data acquisition time points, pressure value sequences, flow rate sequences, sound wave amplitude sequences, vibration acceleration value sequences, soil moisture content value sequences, etc.

[0087] Step S62: Based on the sensor data of each set of forward sensor settings, determine multiple key detection point information and multiple auxiliary detection point information.

[0088] In some embodiments, a detection point determination model can be used to determine multiple key detection point information and multiple auxiliary detection point information. The detection point determination model is a Transformer model. The input to the detection point determination model is the sensor data of each forward sensor setup scheme, and the output of the detection point determination model is the multiple key detection point information and multiple auxiliary detection point information.

[0089] Multiple key detection point information is used to identify multiple representative verification pipe sections that have a high response amplitude to underground pipeline leakage signals and can cover the main suspected leakage locations. The information for each key detection point may include the location coordinates of the key detection point, the pipe section number corresponding to the representative verification pipe section, the sensor type, the leakage signal amplitude, the leakage signal response time, the detection coverage length, and the signal amplitude difference under different forward detection sensor settings.

[0090] Multiple auxiliary detection point information is used to identify multiple representative verification pipe sections that can supplement the key detection points in terms of pipe locations not covered by the key detection points and provide leakage signal comparison results. The information of each auxiliary detection point can include the location coordinates of the auxiliary detection point, the pipe section number corresponding to the verification pipe section, the sensor type, the leakage signal amplitude, the leakage signal response time, the detection coverage length, and the signal amplitude difference under different forward detection sensor setting schemes, etc.

[0091] The sensor data from each forward sensor setup comprehensively reflects the monitoring response of each sensor location under that setup. Response differences between different sensor locations can be compared, and monitoring results from different forward sensor setups can be cross-referenced, thus fully demonstrating the contribution of each sensor location to the leakage monitoring process. Based on the correlation between different sensor locations and the overall variation patterns among different forward sensor setups, the Transformer model can distinguish between detection points that primarily identify leaks and those that provide coverage, supplementation, and signal comparison.

[0092] The Transformer model arranges the sensor data of each forward sensor setup according to the sensor data acquisition time point, and incorporates the sensor location coordinates and the corresponding pipe segment numbers representing the verification pipe sections into the positional relationship calculation. The self-attention layer simultaneously calculates the relationship between changes at the same detection point at different time points, the response relationship between adjacent detection points within a short time frame, and the relationship between results formed by different forward sensor setups in the same pipeline area. When pressure drop, flow imbalance, increased acoustic amplitude, abnormal vibration acceleration, and increased soil moisture content occur consecutively in time and remain stable across multiple forward sensor setups, the relevant detection points are identified as having a high response amplitude to underground pipeline leakage signals. Detection points whose detection coverage length can cover the main suspected leakage locations and whose leakage signal response time can clearly point to the abnormal process are organized into multiple key detection point information. For some detection points whose leakage signal amplitude is not at the main response level, but whose detection coverage length can supplement the pipeline locations not covered by key detection points, or can provide comparison results of normal and abnormal signals at adjacent locations, the Transformer model will organize the relevant detection points into multiple auxiliary detection point information. The decoder outputs the location coordinates of the detection point, the corresponding pipe segment number representing the verification pipe segment, the sensor type, the leakage signal amplitude, the leakage signal response time, the detection coverage length, and the signal amplitude difference under different forward sensor setting schemes.

[0093] Step S63: Determine the sensor setup scheme for the representative verification pipe section based on the information of the multiple key detection points and the information of the multiple auxiliary detection points.

[0094] In some embodiments, Figure 9 This is a flowchart illustrating a sensor setup scheme for determining a representative verification pipe section, provided by an embodiment of the present invention. The process of determining the sensor setup scheme for the representative verification pipe section includes steps S631-S632: Step S631: Construct a sensor data map. The sensor data map includes multiple nodes and edges between multiple nodes. The multiple nodes include multiple key detection nodes and multiple auxiliary detection nodes. The node features of the key detection nodes are key detection point information, and the node features of the auxiliary detection nodes are auxiliary detection point information.

[0095] A sensor data map is a graph data structure composed of multiple key detection nodes, multiple auxiliary detection nodes, and edges between the nodes. The sensor data map is used to represent the sensor deployment at each detection point and the positional connections between different detection points.

[0096] Multiple key detection nodes are nodes in the sensor data map used to represent multiple key detection points. Each key detection node corresponds to one key detection point, and the node features of the key detection node are the key detection point information.

[0097] Multiple auxiliary detection nodes are nodes in the sensor data map used to represent multiple auxiliary detection points. Each auxiliary detection node corresponds to one auxiliary detection point, and the node characteristics of the auxiliary detection node are the auxiliary detection point information.

[0098] Edges between multiple nodes are used to connect adjacent detection nodes whose detection coverage overlaps. Each edge includes the spatial distance between adjacent detection nodes, their relative azimuth angle, and the axial distance of the corresponding pipe segment.

[0099] Step S632: Based on the sensor setting analysis model, process the sensor data map to obtain a sensor setting scheme representing the verification pipe section.

[0100] In some embodiments, the sensor setup analysis model is a graph neural network. The input to the sensor setup analysis model is the sensor data map, and the output of the sensor setup analysis model is a sensor setup scheme representing the verification pipe section.

[0101] Graph Neural Networks (GNNs) are deep learning models specifically designed for processing graph-structured data. GNNs update the feature representations of each node progressively by performing message passing and feature aggregation between neighboring nodes, enabling each node to incorporate information from its neighbors. GNNs can fully utilize the structural relationships reflected by node features and edges in a graph to effectively analyze and predict graph data.

[0102] The sensor deployment scheme for the representative verification section is determined by processing sensor data maps using a sensor deployment analysis model. It serves to guide the formal sensor deployment within the representative verification section. The sensor deployment scheme for the representative verification section may include sensor location coordinates, sensor type, number of sensors, sensor installation direction, sensor sampling frequency, sensor startup sequence, and sensor monitoring duration.

[0103] The sensor data map visually represents the physical distribution and signal response characteristics of various detection points on the verification pipe section using a topological structure. Multiple key detection nodes and multiple auxiliary detection nodes correspond to the key and auxiliary detection points, respectively. Node features encompass information on key and auxiliary detection points, including leakage signal amplitude, response time, and detection coverage length under different forward sensor setups. Edges between nodes represent the spatial distance, relative azimuth, and axial distance between detection points and the pipe section. This structured organization unifies and integrates scattered detection point information with spatial geometric relationships, facilitating the understanding and processing of complex network dependencies by the graph neural network. Through node features and edge information, the graph neural network can effectively capture the signal coverage overlap and monitoring synergy between adjacent detection nodes. The various dimensions of the node features provide signal response intensity and blind zone supplementation indicators under different schemes, while the positional relationships on the edges reflect the transmission and influence patterns of these response indicators in the pipe network space. The construction of the sensor data map allows for the orderly organization of the actual test results and spatial topological relationships of all detection points, providing standardized input for the global optimization decision-making of the graph neural network, while reducing redundant computation and improving the model's inference efficiency and decision stability. Because sensor data maps provide structured node and edge information, graph neural networks can fully utilize spatial topological relationships for feature aggregation, avoiding the data disorder and local optima problems that traditional methods encounter when dealing with scattered deployment points. This structural advantage makes graph neural networks more reliable in complex decision-making processes that balance individual point signal acquisition sensitivity with overall deployment redundancy. By comprehensively analyzing the test characteristics and spatial relationships of each node in the sensor data map, graph neural networks can select the sensor setup scheme that represents the verification pipe section with the strongest signal response and the most streamlined equipment configuration.

[0104] When processing sensor data maps, the graph neural network uses the key detection point information of each key detection node and the auxiliary detection point information of each auxiliary detection node as the initial features of the node. It aggregates the features of surrounding nodes along the edges connecting adjacent detection nodes to the current node. During the aggregation process, the spatial distance, relative azimuth angle, and axial distance of the corresponding pipe segment recorded by the edges determine the degree of influence of surrounding nodes on the current node. Neighboring nodes that are spatially close and have a high degree of overlap in their detection coverage have a greater impact on the features of the current node. After multiple rounds of message passing, the features of each detection node incorporate the leakage signal amplitude and response time information of surrounding detection points. Based on this, the graph neural network judges whether the similarity and number of responses of auxiliary detection nodes around a key detection node are sufficient. Surrounding areas with similar responses and sufficient numbers are considered to have achieved sufficient coverage, and the number of corresponding sensors can be appropriately reduced. Surrounding areas with significant response differences or sparse auxiliary detection nodes are considered to have incomplete detection coverage, requiring the retention of the corresponding key detection node and the addition of auxiliary detection nodes or adjustment of the installation direction of the corresponding sensors. The graph neural network further determines the sensor installation direction based on the sensor type and detection coverage length of each detection node, so that the detection coverage of adjacent sensors can be connected and duplicate coverage can be reduced. It also determines the sensor start-up sequence according to the monitoring coordination relationship between different detection nodes, and finally sorts out and outputs the sensor setting scheme representing the verification pipe section.

[0105] Step S64: Based on the information of multiple key detection points, multiple auxiliary detection points, and the sensor setting scheme representing the verification pipe section, determine the deployment information of multiple candidate formal sensors for each extended monitoring pipe section.

[0106] In some embodiments, a candidate deployment information determination model can be used to determine the deployment information of multiple candidate formal sensors for each extended monitoring pipe segment. The candidate deployment information determination model is a deep neural network. The inputs to the candidate deployment information determination model are the multiple key detection point information, the multiple auxiliary detection point information, and the sensor setup scheme representing the verification pipe segment. The output of the candidate deployment information determination model is the deployment information of multiple candidate formal sensors for each extended monitoring pipe segment.

[0107] The candidate formal sensor deployment information for each extended monitoring pipe segment indicates multiple alternative locations on each segment where sensors can be formally deployed, along with their corresponding sensor settings. Each candidate formal sensor deployment information may include the segment number, sensor location coordinates, sensor type, sensor installation direction, sensor sampling frequency, sensor startup sequence, and sensor monitoring duration.

[0108] Multiple key detection point information reflects the distribution of detection points with primary leakage identification capabilities within the representative verification pipe section, while multiple auxiliary detection point information reflects the distribution of detection points serving to supplement coverage and provide signal comparison. The information from multiple key detection points, multiple auxiliary detection points, and the sensor placement scheme of the representative verification pipe section collectively reflects the division of functions and corresponding deployment patterns of different detection points within that section. The model can leverage the established correspondence between the representative verification pipe section and extended monitoring pipe sections to transfer the deployment patterns within the representative verification pipe section, enabling the effective placement methods in the representative verification pipe section to be adapted to the corresponding extended monitoring pipe sections, thereby determining the candidate formal deployment information for each extended monitoring pipe section.

[0109] The deep neural network uses the sensor setup scheme representing the verification pipe section as a deployment reference. It learns the correspondence between sensor placement locations and high-response leakage signals from multiple key detection point information, and the setting patterns of supplementary detection coverage locations and signal comparison locations from multiple auxiliary detection point information. The deep neural network can convert the sensor placement locations on the representative verification pipe section from the original axial distance to the corresponding axial position on the extended monitoring pipe section, based on the established association between each extended monitoring pipe section and its corresponding representative verification pipe section. Sensor type and installation direction are matched based on verified adaptation results from multiple key and auxiliary detection point information. The sensor sampling frequency forms alternative values ​​based on the leakage signal change rate. The sensor start-up sequence and monitoring duration follow an effective arrangement for anomaly detection and signal comparison. The hidden layer of the deep neural network retains several position offset results around the converted corresponding axial position to adapt to potential positional limitations that may occur during the actual installation of the extended monitoring pipe section. Each position offset result must maintain the correspondence between key and auxiliary detection coverage and avoid changing the adaptation relationship between sensor type and detection task. The output layer organizes the segment number, sensor location coordinates, sensor type, sensor installation direction, sensor sampling frequency, sensor start-up sequence, and sensor monitoring duration for each extended monitoring segment, forming multiple candidate formal sensor deployment information for each extended monitoring segment.

[0110] Step S65. Determine the sensor configuration scheme for each extended monitoring pipe segment based on the multiple candidate formal sensor deployment information for each extended monitoring pipe segment.

[0111] In some embodiments, a configuration scheme determination model can be used to determine the sensor configuration scheme for extended monitoring pipe segments. The configuration scheme determination model is a deep neural network. The input to the configuration scheme determination model is multiple candidate formal sensor deployment information for each extended monitoring pipe segment, and the output of the configuration scheme determination model is the sensor configuration scheme for the extended monitoring pipe segment.

[0112] The sensor deployment plan for extended monitoring pipeline sections is a comprehensive scheme to guide the formal sensor deployment of all extended monitoring pipeline sections. The sensor deployment plan for extended monitoring pipeline sections may include sensor location coordinates, sensor type, number of sensors, sensor installation direction, sensor sampling frequency, sensor start-up sequence, and sensor monitoring duration.

[0113] The multiple candidate formal sensor deployment information for each extended monitoring pipe segment corresponds to various alternative formal deployment methods for the same extended monitoring pipe segment. There are differences in the overall deployment effect among different candidate formal sensor deployment information. The model compares the adaptability of each candidate formal sensor deployment information with the corresponding extended monitoring pipe segment and comprehensively analyzes the overall coordination relationship among different candidate formal sensor deployment information. In this way, it can determine the sensor deployment information suitable for formal deployment from the multiple candidate formal sensor deployment information for each extended monitoring pipe segment and form a coordinated and consistent formal sensor deployment scheme among all extended monitoring pipe segments.

[0114] After reading the deployment information of multiple candidate sensors for each extended monitoring pipe segment, the deep neural network can map all its internal parameter combinations into a massive number of discrete feature points in a high-dimensional solution space. The hidden layer analyzes the positional connectivity, detection coverage, and sensor parameter coordination among the candidate sensor deployment information of different extended monitoring pipe segments through nonlinear transformations. The model selects candidate sensor deployment information combinations with more complete detection coverage and fewer redundant deployments by comparing the completeness of detection coverage and the degree of cross-segment redundancy under different combinations. In the rigorous gradient descent optimization process, the deep neural network continuously compares the spatial distance and axial span of adjacent candidate positions on different extended pipe segments. When two physically close candidate sensor positions are found at the boundary of adjacent extended monitoring pipe segments, and the detection coverage of the two candidate sensor positions overlaps significantly, the model eliminates redundant candidate sensor positions with low signal response sensitivity and weak detection coverage through a penalty term. The model further verifies the coordination of sensor activation sequences among different extended pipe segments in the entire underground pipeline network, and fine-tunes the activation time nodes through time-series peak-shifting simulation to avoid instantaneous congestion in the data transmission channel. Finally, the output layer performs a final summary and serialization operation on the coordinates of all sensor nodes, sensor types, sensor quantities, and various monitoring parameters after global filtering and parameter coordination calibration, accurately integrating and outputting a sensor setting scheme suitable for the extended monitoring pipe segments with seamless overall perimeter coverage.

[0115] Step S7: Based on the sensor setting scheme of the representative verification pipe section and the sensor setting scheme of the extended monitoring pipe section, underground pipeline leakage monitoring is carried out.

[0116] Once the sensor setup scheme for the representative verification pipe section and the sensor setup scheme for the extended monitoring pipe section are determined, leakage monitoring of the underground pipeline is carried out based on the sensor setup scheme for the representative verification pipe section and the sensor setup scheme for the extended monitoring pipe section.

[0117] This step is existing technology and will not be described in detail here.

[0118] Based on the same inventive concept Figure 10 This is a schematic diagram of an underground pipeline leakage monitoring system provided in an embodiment of the present invention. The underground pipeline leakage monitoring system includes: Module 81 is used to acquire underground pipe network information and initial leakage monitoring information; The candidate leakage monitoring pipe segment determination module 82 is used to determine multiple candidate leakage monitoring pipe segments and information of each candidate leakage monitoring segment based on the underground pipe network information and the initial leakage monitoring information. The representative verification pipe section determination module 83 is used to determine multiple representative verification pipe section information and multiple extended monitoring pipe section information based on the information of each candidate leakage monitoring pipe section; The candidate setting location determination module 84 is used to determine multiple candidate setting location information based on the verification pipe segment information of each representative; The forward sensor setting scheme generation module 85 is used to generate multiple forward sensor setting schemes based on the multiple candidate setting position information. The sensor setting scheme determination module 86 is used to determine the sensor setting scheme representing the verification pipe section and the sensor setting scheme for the extended monitoring pipe section based on the multiple sets of forward sensor setting schemes. The leakage monitoring module 87 is used to monitor underground pipeline leakage based on the sensor setting scheme of the representative verification pipe section and the sensor setting scheme of the extended monitoring pipe section.

[0119] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0120] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and are considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for monitoring leakage in underground pipelines, characterized in that, include: Obtain information on underground pipe networks and initial leakage monitoring information; Based on the underground pipeline network information and the initial leakage monitoring information, multiple candidate leakage monitoring pipe sections are determined, along with information for each candidate leakage monitoring pipe section; Based on the information of each candidate leakage monitoring pipe segment, multiple representative verification pipe segment information and multiple extended monitoring pipe segment information are determined; Multiple candidate installation location information are determined based on the verification pipe section information of each representative; Multiple forward sensor setting schemes are generated based on the multiple candidate setting location information; Based on the aforementioned multiple sets of forward sensor setting schemes, determine the sensor setting scheme for the representative verification pipe section and the sensor setting scheme for the extended monitoring pipe section; Underground pipeline leakage monitoring is carried out based on the sensor setting scheme of the representative verification pipe section and the sensor setting scheme of the extended monitoring pipe section.

2. The method for monitoring leakage in underground pipelines as described in claim 1, characterized in that, The determination of multiple representative verification pipe segment information and multiple extended monitoring pipe segment information based on the information of each candidate leakage monitoring pipe segment includes: Based on the information of each candidate leakage monitoring pipe segment, a clustering algorithm is used to cluster them to obtain multiple candidate leakage monitoring pipe segment groups; Based on the multiple candidate leakage monitoring pipe segment groups, the initial representative verification pipe segment information of each candidate leakage monitoring pipe segment group and the multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment are determined. Based on the multiple candidate leakage monitoring pipe segment groups, the initial representative verification pipe segment information of each candidate leakage monitoring pipe segment group, and the multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment, multiple representative verification pipe segment information and multiple extended monitoring pipe segment information are determined.

3. The method for monitoring leakage in underground pipelines as described in claim 1, characterized in that, The process of determining the sensor setup scheme for the representative verification section and the sensor setup scheme for the extended monitoring section based on the multiple sets of forward sensor setup schemes includes: The robot is controlled to deploy each set of forward sensor configuration schemes in multiple representative verification pipe sections and acquire sensor data for each set of forward sensor configuration schemes. Based on the sensor data of each set of forward sensor settings, multiple key detection point information and multiple auxiliary detection point information are determined; Based on the information of the multiple key detection points and the information of the multiple auxiliary detection points, a sensor setting scheme representing the verification pipe section is determined; Based on the information of multiple key detection points, multiple auxiliary detection points, and the sensor setting scheme representing the verification pipe section, the deployment information of multiple candidate formal sensors for each extended monitoring pipe section is determined. The sensor setup scheme for each extended monitoring pipe section is determined based on the deployment information of multiple candidate formal sensors for each extended monitoring pipe section.

4. The method for monitoring leakage in underground pipelines as described in claim 3, characterized in that, The sensor setup scheme for determining the representative verification pipe section based on the information of the multiple key detection points and the information of the multiple auxiliary detection points includes: Construct a sensor data map, which includes multiple nodes and edges between them. The multiple nodes include multiple key detection nodes and multiple auxiliary detection nodes. The node features of key detection nodes are key detection point information, and the node features of auxiliary detection nodes are auxiliary detection point information. The sensor data map is processed based on the sensor setup analysis model to obtain a sensor setup scheme representing the verification pipe section.

5. A leakage monitoring system for underground pipelines, characterized in that, include: The acquisition module is used to acquire underground pipe network information and initial leakage monitoring information; The candidate leakage monitoring pipe segment determination module is used to determine multiple candidate leakage monitoring pipe segments and information for each candidate leakage monitoring segment based on the underground pipe network information and the initial leakage monitoring information. The representative verification pipe section determination module is used to determine multiple representative verification pipe section information and multiple extended monitoring pipe section information based on each candidate leakage monitoring pipe section information; The candidate setting location determination module is used to determine multiple candidate setting location information based on the verification pipe segment information of each representative; The forward sensor setting scheme generation module is used to generate multiple forward sensor setting schemes based on the multiple candidate setting location information. The sensor setting scheme determination module is used to determine the sensor setting scheme for the representative verification pipe section and the sensor setting scheme for the extended monitoring pipe section based on the multiple sets of forward sensor setting schemes. The leakage monitoring module is used to monitor underground pipeline leakage based on the sensor setting scheme of the representative verification pipe section and the sensor setting scheme of the extended monitoring pipe section.

6. The underground pipeline leakage monitoring system as described in claim 5, characterized in that, The module for determining the representative verification pipe section is also used for: Based on the information of each candidate leakage monitoring pipe segment, a clustering algorithm is used to cluster them to obtain multiple candidate leakage monitoring pipe segment groups; Based on the multiple candidate leakage monitoring pipe segment groups, the initial representative verification pipe segment information of each candidate leakage monitoring pipe segment group and the multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment are determined. Based on the multiple candidate leakage monitoring pipe segment groups, the initial representative verification pipe segment information of each candidate leakage monitoring pipe segment group, and the multiple extended monitoring pipe segment information corresponding to each initial representative verification pipe segment, multiple representative verification pipe segment information and multiple extended monitoring pipe segment information are determined.

7. The underground pipeline leakage monitoring system as described in claim 5, characterized in that, The sensor setting scheme determination module is also used for: The robot is controlled to deploy each set of forward sensor configuration schemes in multiple representative verification pipe sections and acquire sensor data for each set of forward sensor configuration schemes. Based on the sensor data of each set of forward sensor settings, multiple key detection point information and multiple auxiliary detection point information are determined; Based on the information of the multiple key detection points and the information of the multiple auxiliary detection points, a sensor setting scheme representing the verification pipe section is determined; Based on the information of multiple key detection points, multiple auxiliary detection points, and the sensor setting scheme representing the verification pipe section, the deployment information of multiple candidate formal sensors for each extended monitoring pipe section is determined. The sensor setup scheme for each extended monitoring pipe section is determined based on the deployment information of multiple candidate formal sensors for each extended monitoring pipe section.

8. The underground pipeline leakage monitoring system as described in claim 7, characterized in that, The sensor setup scheme for determining the representative verification pipe section based on the information of the multiple key detection points and the information of the multiple auxiliary detection points includes: Construct a sensor data map, which includes multiple nodes and edges between them. The multiple nodes include multiple key detection nodes and multiple auxiliary detection nodes. The node features of key detection nodes are key detection point information, and the node features of auxiliary detection nodes are auxiliary detection point information. The sensor data map is processed based on the sensor setup analysis model to obtain a sensor setup scheme representing the verification pipe section.

9. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the underground pipeline leakage monitoring method as claimed in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the underground pipeline leakage monitoring method as described in any one of claims 1 to 4.