Underground pipe duct safety operation and maintenance method and system based on Internet of Things

Abstract

The invention relates to an underground pipe duct safety operation and maintenance method based on the Internet of Things, and the method comprises the following steps: S1, obtaining pipe network GIS data and historical risk point data, carrying out the node layout optimization based on risk assessment, and obtaining the optimized sensor layout and network topology; s2, acquiring original data of various sensors based on the arranged sensors, and preprocessing the original data to obtain a standardized data packet; s3, performing multi-modal data fusion according to the standardized data packet, the historical operation and maintenance record and the GIS data to obtain fused data; s4, according to the fused data, on the basis of a multi-modal model, underground pipe duct anomaly detection including environment anomaly, pipeline structure anomaly and operation state anomaly detection is carried out, and anomaly detection results are obtained; and S5, according to an anomaly detection result, combining official website GIS data with the optimized sensor layout and network topology, and accurately locking a risk point. The safety management level and the emergency response capability of the urban underground pipe network are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of underground pipeline operation and maintenance, and in particular to an Internet of Things-based method and system for safe operation and maintenance of underground pipelines. Background Technology

[0002] With the acceleration of urbanization, underground pipeline networks (such as drainage, water supply, and gas infrastructure) are becoming increasingly complex, operating in variable environments, and posing growing management challenges. Traditional pipeline network operation and maintenance methods rely heavily on manual inspections and periodic maintenance, which are not only inefficient and costly but also fail to provide timely detection and rapid response to sudden risks. In recent years, emerging technologies such as the Internet of Things (IoT), big data, and artificial intelligence have developed rapidly and demonstrated enormous potential in the intelligent management of urban infrastructure. Various new sensors can monitor and collect multi-dimensional information such as pipeline network operating status, structural health, and environmental changes in real time. Combined with spatial information technologies such as GIS (Geographic Information System), this provides rich data support for the entire lifecycle management of pipeline networks. However, the large volume and complex types of multi-source heterogeneous data during pipeline network operation and maintenance, along with issues such as information silos and decentralized decision-making, remain serious problems, making it difficult to achieve efficient integration, identification, and handling of abnormal risks. Summary of the Invention

[0003] To address the aforementioned issues, the present invention aims to provide an IoT-based method and system for the safe operation and maintenance of underground pipelines, achieving a fully intelligent closed loop from data collection and fusion analysis to risk identification and efficient operation and maintenance, thereby improving the safety management level and emergency response capabilities of urban underground pipeline networks.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: A method for safe operation and maintenance of underground pipelines based on the Internet of Things includes the following steps: S1: Obtain pipeline GIS data and historical risk point data, optimize node layout based on risk assessment, and obtain optimized sensor layout and network topology; S2: Based on the deployed sensors, acquire raw data from various sensors and preprocess it to obtain standardized data packets; S3: Based on standardized data packets, historical operation and maintenance records, and GIS data, multimodal data fusion is performed to obtain fused data; S4: Based on the fused data and a multimodal model, perform anomaly detection for underground pipelines, including environmental anomalies, pipeline structure anomalies, and operational status anomalies, and obtain anomaly detection results. S5: Based on the anomaly detection results, combined with official website GIS data and optimized sensor layout and network topology, the risk points are accurately identified.

[0005] Furthermore, by acquiring pipeline network GIS data and historical risk point data, and based on risk assessment and node layout optimization, an optimized sensor layout and network topology are obtained, specifically: Establish a risk assessment indicator system and calculate the comprehensive risk value R for each node / segment k. k : R k =αF k +βI k +γP k ; Where α, β, γ are weighting coefficients; F k For fault frequency, I k For the degree of influence and P k For risk probability; Based on the comprehensive risk value R k The pipeline nodes are sorted, and the node set K that can effectively monitor the area of ​​greatest risk impact is selected based on the pipeline network topology. Based on risk points and weights {(k,R k )}, deployable location {j}, total cost constraint C total Based on the maximum coverage location selection model, the spatial distribution of sensors after optimized layout is obtained: ; Where M is the set of possible deployment locations; a kj For the discriminant variable: if placing a sensor at deployment point j can cover risk k, then a kj= 1, otherwise a kj =0;C j The cost of installing the sensor at deployment point j; C total For budget; x j For decision variables: if a sensor is deployed at location j, x j =1, otherwise 0; y k For the coverage variable: if risk point k is covered by at least one sensor, y k =1, otherwise 0; The optimal communication route and data aggregation node location are determined by a topology optimization algorithm, network connectivity constraints are applied, and network energy consumption or latency is minimized.

[0006] Furthermore, the optimal communication route and data aggregation node location are determined through topology optimization algorithms, network connectivity constraints are applied, and network energy consumption or latency is minimized, as detailed below: Based on the spatial distribution of the sensors after optimization, coordinate information is obtained, and the communication radius r0 of all nodes is defined. For any two nodes i and j, if their distance d ijIf r ≤ r0, then a communication edge is established between the two, and all nodes are constructed into an undirected graph to reflect the wireless communication connection relationship; For the constructed undirected graph, the depth-first search algorithm is used to detect the connectivity of the entire sensor network. If the network is not strongly connected, relay nodes are added to ensure connectivity, thereby ensuring the smooth collection and transmission of subsequent data. Based on network connectivity, to reduce overall transmission energy consumption, a minimum spanning tree algorithm (such as Prim or Kruskal) is applied, using the physical distance between nodes as edge weights, to extract the backbone network structure with the minimum total weight. The optimization objective is to connect all nodes with the minimum total cost, resulting in an acyclic backbone structure.

[0007] Based on network size, data traffic distribution, and transmission latency requirements, the number of sink nodes is determined, and the K-means clustering algorithm is used to cluster all sensor nodes x. i The coordinates are used to cluster the network, dividing it into M subnetworks C. m For each subnetwork C m , will converge node S m Assume that all nodes in this cluster are connected to S. m The location with the minimum sum of distances is chosen to minimize the total data transmission distance or power consumption of all sensor nodes within the area to their convergence node. ; Within each sub-network, for all ordinary nodes to the sink node, the Dijkstra shortest path algorithm is used to calculate the main path with the minimum energy consumption, with the edge weight as the link energy consumption. The final output includes the spatial distribution of each sensor and aggregation node, the optimal routing structure for energy consumption within the backbone network and each sub-network, and the overall network energy consumption performance indicators.

[0008] Further preprocessing is performed as follows: To address noise, spikes, and outliers in the raw data, preprocessing algorithms such as moving average, median filtering, or wavelet denoising are used to remove distorted measurements and ensure data quality. After cleaning, the data is organized into a standardized data format according to network protocols and requirements of the upper-level system. Each data packet contains: sensor ID, type label, timestamp, raw or processed data value, and quality identification metadata; all fields are aligned according to the agreed standard fields. The standardized data is repackaged effectively using TLV and structure to optimize data length and parsing efficiency, preparing it for network transmission.

[0009] Furthermore, based on standardized data packets, historical operation and maintenance records, and GIS data, multimodal data fusion is performed to obtain the fused data, as follows: For standardized data packets, historical operation and maintenance records, and GIS data, feature extraction is performed, and feature alignment is completed based on common features of space, time, and entity ID: sensor collection points are mapped and overlaid on the GIS map according to spatial location to achieve a one-to-one correspondence in physical location; real-time data is synchronized with historical operation and maintenance events according to a unified standard time axis to compare the operating status and maintenance operations within a specific time period; sensor IDs are used as indexes to establish the correlation of features from multiple data sources. The aligned features of the multi-source data are spliced ​​and combined to form a comprehensive feature set that includes environmental, geographical, and operational characteristics, resulting in fused data.

[0010] Furthermore, based on the fused data and a multimodal model, anomaly detection is performed on underground pipelines, and the anomaly detection results are obtained as follows: Based on the fused data, the environmental, structural, and operational status data are respectively formed into vectors X(env), X(struct), and X(ops); GIS spatial adjacency information is organized into graph-structured data, and numerical features are normalized, time series data are synchronously completed, and the basis for multimodal joint input is established. First, input the respective parameterized multi-head self-attention Transformer encoders to extract the deep latent representation within the modality: Z (m) =MultiHeadAttn(Q (m) ,K (m) V (m) ),m∈{env,struct,ops}; Pipeline topology data is used to encode the spatial neighborhood characteristics of each monitoring point using a graph neural network: ; in, This represents the feature representation of node i in the network of the l-th layer graph neural network; Let i be a neighboring node of node i; The self-loop weight matrix of the l-th layer The transformation matrix represents the features of the neighbors, which projects the features of the neighboring nodes onto the current node. It is a non-linear activation function; A cross-attention mechanism is used to establish information flow among environmental, structural, and operational state modalities, enabling intermodal pairing and signal complementarity. ; in, and Let i and j be the representation vectors of modes i′ and j′, respectively; i′, j′ ∈ {env, struct, ops}; Information from mode j′ is passed to mode i′ through cross attention; To query the weight matrix, used to assign Z (i′) Linear transformation into a query vector; For the key weight matrix, Z (j′) A linear transformation is used to obtain a key vector; For the value weight matrix, put Z (j′) Transform into a value vector; d k is the dimension of the key vector; the superscript T indicates transpose; Based on a gating fusion mechanism, the contributions of the original modality and the interaction modality are adaptively balanced to improve the fusion effectiveness. ; in, Indicates that Z (i′) With C i′,j′ Concatenate along the feature dimension; Here is the weight matrix of the gated network; is the bias vector of the gated network; Indicates the gating weight; ⊙ represents the product of corresponding elements; The features of mode i′ after gated fusion; Multimodal features are stitched together in the fusion layer: ; Based on the fusion features, the output branch predicts the scores S for the three types of anomalies respectively. env, S struct ,S ops : ; Among them, W m b m Let m be the first-level weights and biases. w is the activation function. m b m′ For output layer weights and biases; It is the sigmoid function; Various anomalies are weighted and combined to form an overall risk score, which is then multiplied by the confidence level to produce the final anomaly strength:

[0011] Determine if AnomalyScore > θ t θ t The threshold is used for abnormal warnings.

[0012] Furthermore, based on the anomaly detection results, combined with official website GIS data and optimized sensor layout and network topology, risk points were precisely identified, as follows: The anomaly detection results output by the anomaly detection model, the corresponding high-risk sensors, pipe sections or nodes, are associated with the GIS database through the sensor ID to obtain their accurate location in the real geographic space and pipeline network. By utilizing the current optimized sensor layout and the latest network topology, we analyze the location attributes of anomalies in the pipeline system to determine whether they are located at critical nodes in the pipeline network. Combining pipeline flow direction, upstream and downstream relationships of nodes, and hydraulic conditions, we analyze the possible propagation paths and impact range of abnormal signals. On the GIS platform, with anomalies as the center, a safety buffer zone or risk impact radius is set. Combined with the distribution of historical risk points, spatial overlay analysis is performed on the surrounding pipe sections, important facilities and residential areas that may be affected to form a preliminary candidate set of high-risk areas. Taking into account anomaly scores, frequency of historical risk occurrences, topological importance of nodes, and spatial clustering, a weighted model is used to calculate the final risk score for each candidate point: ; Among them, S anom,i D represents the anomaly detection score. cluster,i For spatial clustering density, C topology,i H represents the topological centrality of the pipeline network. history,i Historical risks are weighted, with α, β, γ, and δ as weighting parameters; nodes with scores higher than the threshold are selected as the risk points locked in this instance.

[0013] Furthermore, based on the anomaly detection results and risk points, visualization is performed, and the detection report is distributed to the operations and maintenance personnel, as detailed below: Using a GIS platform, the precisely identified risk points are visually marked on the map, with different colors, icons, or heat maps representing different risk levels and anomaly types; and detailed attributes of the risk points are displayed, including: anomaly type, detection time, anomaly score, pipeline number, geographical location, surrounding infrastructure, and historical risk occurrence data; For high-risk points or the diffusion and linkage relationships between multiple risk points, dynamic layers are used to display the potential propagation links of risk events and the historical abnormal evolution process, helping operation and maintenance personnel to conduct trend tracking and causal analysis.

[0014] An Internet of Things (IoT)-based underground pipeline safety operation and maintenance system includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the IoT-based underground pipeline safety operation and maintenance method described above.

[0015] A computer storage medium storing a plurality of instructions adapted for loading by a processor and executing the method steps described above.

[0016] The present invention has the following beneficial effects: 1. This invention realizes fully automated closed-loop management of the entire process, from pipeline network GIS and historical data collection, node optimization and layout, sensor data standardization, to multimodal fusion and intelligent anomaly detection. Through the deep fusion of multi-source heterogeneous data and the introduction of multimodal detection models, it significantly improves the accuracy of risk point identification and the timeliness of anomaly response, effectively reduces the workload of manual inspection, and realizes early intelligent warning of underground pipeline safety risks. 2. This invention closely integrates anomaly detection results with GIS spatial data, optimized sensor layout, and network topology, enabling precise mapping and in-depth analysis of risk points at multiple levels of "point-line-network-surface". Through spatial buffering and topology diffusion methods, it can lock and infer the risk impact area, providing a basis for quantitative assessment and intervention of abnormal events. 3. This invention integrates the Internet of Things, multimodal information fusion, and intelligent analysis into a method for safe operation and maintenance of underground pipelines, realizing a closed-loop intelligent system from data collection and fusion analysis to risk identification and efficient operation and maintenance, thereby improving the safety management level and emergency response capability of urban underground pipeline networks. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: refer to Figure 1 In this embodiment, an IoT-based method for the safe operation and maintenance of underground pipelines is provided, including the following steps: S1: Obtain pipeline GIS data and historical risk point data, optimize node layout based on risk assessment, and obtain optimized sensor layout and network topology; S2: Based on the deployed sensors, acquire raw data from various sensors and preprocess it to obtain standardized data packets; S3: Based on standardized data packets, historical operation and maintenance records, and GIS data, multimodal data fusion is performed to obtain fused data; S4: Based on the fused data and a multimodal model, perform anomaly detection for underground pipelines, including environmental anomalies, pipeline structure anomalies, and operational status anomalies, and obtain anomaly detection results. S5: Based on the anomaly detection results, combined with official website GIS data and optimized sensor layout and network topology, the risk points are accurately identified.

[0019] In this embodiment, pipeline GIS data and historical risk point data are acquired, and node layout optimization based on risk assessment is performed to obtain an optimized sensor layout and network topology, specifically as follows: Establish a risk assessment indicator system and calculate the comprehensive risk value R for each node / segment k. k : R k =αF k +βI k +γP k ; Where α, β, and γ are weighting coefficients, which can be set using the AHP (Analytic Hierarchy Process) method; F k For fault frequency, I k For the degree of influence and P k For risk probability; Based on the comprehensive risk value R k The pipeline nodes are sorted, and the node set K that can effectively monitor the area of ​​greatest risk impact is selected based on the pipeline network topology. Based on risk points and weights {(k,R k )}, deployable location {j}, total cost constraint C total Based on the maximum coverage location selection model, the spatial distribution of sensors after optimized layout is obtained: ; Where M is the set of possible deployment locations (i.e., the locations of all deployable sensors); a kj For the discriminant variable: if placing a sensor at deployment point j can cover risk k, then a kj= 1, otherwise a kj =0 (Coverage relationship is determined by the sensing radius); C j The cost of installing sensors at deployment point j (which can be a weighted total cost including quantity, cost, construction, etc.); C total For budget; x j For decision variables: if a sensor is deployed at location j, x j =1, otherwise 0; y k For the coverage variable: if risk point k is covered by at least one sensor, y k =1, otherwise 0; The optimal communication route and data aggregation node location are determined by a topology optimization algorithm, network connectivity constraints are applied, and network energy consumption or latency is minimized.

[0020] In this embodiment, the optimal communication route and data aggregation node location are determined through a topology optimization algorithm, network connectivity constraints are applied, and network energy consumption or latency is minimized, as detailed below: Based on the spatial distribution of the sensors after optimization, coordinate information is obtained, and the communication radius r0 of all nodes is defined. For any two nodes i and j, if their distance d ij If r ≤ r0, then a communication edge is established between the two, and all nodes are constructed into an undirected graph to reflect the wireless communication connection relationship; For the constructed undirected graph, the depth-first search algorithm is used to detect the connectivity of the entire sensor network. If the network is not strongly connected, relay nodes are added to ensure connectivity, thereby ensuring the smooth collection and transmission of subsequent data. Building upon network connectivity, to reduce overall transmission energy consumption, a minimum spanning tree algorithm (such as Prim's or Kruskal's algorithm) is applied, using the physical distance between nodes as edge weights, to extract the backbone network structure with the minimum total weight. This structure ensures that all nodes are interconnected with the lowest total connection cost, effectively reducing network deployment and maintenance costs. The optimization objective is to connect all nodes with the minimum total cost, achieving a loop-free backbone structure.

[0021] Based on network size, data traffic distribution, and transmission latency requirements, the number of sink nodes is determined, and the K-means clustering algorithm is used to cluster all sensor nodes x. i The coordinates are used to cluster the network, dividing it into M subnetworks C. m For each subnetwork C m , will converge node S m Assume that all nodes in this cluster are connected to S. m The location with the minimum sum of distances is chosen to minimize the total data transmission distance or power consumption of all sensor nodes within the area to their convergence node. ; Within each sub-network, for all ordinary nodes to the aggregation node, the link energy consumption is determined by the edge weight, and the Dijkstra shortest path algorithm is used to calculate the main path with the minimum energy consumption. This enables high-speed and low-energy aggregation of node data, further improving the energy efficiency of data transmission and reducing the overall network energy consumption. The final output includes the spatial distribution of each sensor and aggregation node, the energy-optimized routing structure within the backbone network and each subnetwork, and the overall network energy performance indicators. This not only ensures coverage and communication connectivity but also achieves global optimization of energy consumption (or latency, etc.), meeting the efficiency and cost-effectiveness requirements of actual deployment.

[0022] In this embodiment, the preprocessing is as follows: To address noise, spikes, and outliers in the raw data, preprocessing algorithms such as moving average, median filtering, or wavelet denoising are used to remove distorted measurements and ensure data quality. After cleaning, the data is organized into a standardized data format according to network protocols and requirements of the upper-level system. Each data packet contains: sensor ID, type label, timestamp, raw or processed data value, and quality identification metadata; all fields are aligned according to the agreed standard fields. The standardized data is repackaged effectively using a TLV (Type-Length-Value) structure to optimize data length and parsing efficiency, preparing it for network transmission.

[0023] In this embodiment, multimodal data fusion is performed based on standardized data packets, historical operation and maintenance records, and GIS data to obtain fused data, as follows: Feature extraction is performed on standardized data packets, historical operation and maintenance records, and GIS data. Feature alignment is then completed based on common spatial, temporal, and entity ID characteristics: sensor acquisition points are mapped and overlaid onto the GIS map according to spatial location (latitude and longitude / geocoding) to achieve a one-to-one correspondence in physical location; real-time data is synchronized with historical operation and maintenance events according to a unified standard timeline to compare the operating status and maintenance operations within a specific time period; and sensor IDs are used as indexes to establish the correlation of features among multi-source data. The aligned features of the multi-source data are spliced ​​and combined to form a comprehensive feature set that includes environmental, geographical, and operational characteristics, resulting in fused data.

[0024] In this embodiment, based on the fused data and a multimodal model, anomaly detection of underground pipelines is performed, and the anomaly detection results are obtained, as follows: Based on the fused data, the environmental, structural, and operational status data are respectively formed into vectors X(env), X(struct), and X(ops); GIS spatial adjacency information is organized into graph-structured data, and numerical features are normalized, time series data are synchronously completed, and the basis for multimodal joint input is established. First, input the respective parameterized multi-head self-attention Transformer encoders to extract the deep latent representation within the modality: Z (m) =MultiHeadAttn(Q (m) ,K (m) V (m) ),m∈{env,struct,ops}; Pipeline topology data is used to encode the spatial neighborhood characteristics of each monitoring point using a graph neural network: ; in, This represents the feature representation of node i in the network of the l-th layer graph neural network; Let i be a neighboring node of node i; The self-loop weight matrix of the l-th layer The transformation matrix represents the features of the neighbors, which projects the features of the neighboring nodes onto the current node. It is a non-linear activation function; A cross-attention mechanism is used to establish information flow among environmental, structural, and operational state modalities, enabling intermodal pairing and signal complementarity. ; in, and Let i and j be the representation vectors of modes i′ and j′, respectively; i′, j′ ∈ {env, struct, ops}; Information from mode j′ is passed to mode i′ through cross attention; To query the weight matrix, used to assign Z (i′) Linear transformation into a query vector; For the key weight matrix, Z (j′) A linear transformation is used to obtain a key vector; For the value weight matrix, put Z (j′) Transform into a value vector; d k is the dimension of the key vector; the superscript T indicates transpose; Based on a gating fusion mechanism, the contributions of the original modality and the interaction modality are adaptively balanced to improve the fusion effectiveness. ; in, Indicates that Z (i′) With C i′,j′ Concatenate along the feature dimension; Here is the weight matrix of the gated network; is the bias vector of the gated network; ⊙ represents the gating weight; ⊙ represents the product of corresponding elements (Hadamard). The features of mode i′ after gated fusion; Multimodal features are stitched together in the fusion layer: ; Based on the fusion features, the output branch predicts the scores S for the three types of anomalies respectively. env, S struct ,S ops : ; Among them, W m b m Let m be the first-level weights and biases. w is the activation function. m b m′ For output layer weights and biases; It is the sigmoid function; Various anomalies are weighted and combined to form an overall risk score, which is then multiplied by the confidence level to produce the final anomaly strength:

[0025] Determine if AnomalyScore > θ t θ t The threshold is used for abnormal warnings.

[0026] In this embodiment, based on the anomaly detection results, combined with official website GIS data and optimized sensor layout and network topology, risk points are accurately identified, as follows: The anomaly detection results output by the anomaly detection model, the corresponding high-risk sensors, pipe sections or nodes (including type, time and anomaly score), are associated with the GIS database through sensor ID to obtain their accurate location in real geographic space and pipeline network (such as latitude and longitude coordinates, pipeline network zoning, and associated pipe sections). By utilizing the current optimized sensor layout and the latest network topology, we analyze the location attributes of anomalies in the pipeline system to determine whether they are located at critical nodes in the pipeline network (such as main pipes, junctions, branches, bottlenecks, etc.); and by combining pipeline flow direction, upstream and downstream relationships of nodes, and hydraulic conditions, we analyze the possible propagation paths and impact range of abnormal signals. On the GIS platform, with anomalies as the center, a safety buffer zone or risk impact radius is set. Combined with the distribution of historical risk points, spatial overlay analysis is performed on the surrounding pipe sections, important facilities and residential areas that may be affected to form a preliminary candidate set of high-risk areas. By combining anomaly scores, historical risk frequency, topological importance of nodes (such as degree centrality and betweenness), and spatial clustering, a weighted model is used to calculate the final risk score for each candidate point. ; Among them, S anom,i D represents the anomaly detection score. cluster,i For spatial clustering density, C topology,i H represents the topological centrality of the pipeline network. history,i Historical risks are weighted, with α, β, γ, and δ as weighting parameters; nodes with scores higher than the threshold are selected as the risk points locked in this instance.

[0027] In this embodiment, the anomaly detection results and risk points are visualized, and the detection report is distributed to the operations and maintenance personnel, as detailed below: Using a GIS platform, the precisely identified risk points (including abnormal sensor nodes, abnormal pipe sections and their affected areas) are visually marked on the map, and different risk levels and anomaly types are represented by different colors, icons or heat maps; and the detailed attributes of the risk points are displayed, including: anomaly type (such as structural, environmental or operational anomaly), detection time, anomaly score, pipeline number, geographical location, surrounding infrastructure and historical risk occurrence data. For high-risk points or the diffusion and linkage relationships between multiple risk points, dynamic layers are used to display the potential propagation links of risk events and the historical abnormal evolution process, helping operation and maintenance personnel to conduct trend tracking and causal analysis.

[0028] An Internet of Things (IoT)-based underground pipeline safety operation and maintenance system includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the IoT-based underground pipeline safety operation and maintenance method described above.

[0029] A computer storage medium storing a plurality of instructions adapted for loading by a processor and executing the method steps described above.

[0030] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0031] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0032] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0033] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0034] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for safe operation and maintenance of underground pipelines based on the Internet of Things, characterized in that, Includes the following steps: S1: Obtain pipeline GIS data and historical risk point data, optimize node layout based on risk assessment, and obtain optimized sensor layout and network topology; S2: Based on the deployed sensors, acquire raw data from various sensors and preprocess it to obtain standardized data packets; S3: Based on standardized data packets, historical operation and maintenance records, and GIS data, multimodal data fusion is performed to obtain fused data; S4: Based on the fused data and a multimodal model, perform anomaly detection for underground pipelines, including environmental anomalies, pipeline structure anomalies, and operational status anomalies, and obtain anomaly detection results. S5: Based on the anomaly detection results, combined with official website GIS data and optimized sensor layout and network topology, the risk points are accurately identified.

2. The method for safe operation and maintenance of underground pipelines based on the Internet of Things according to claim 1, characterized in that, The acquisition of pipeline network GIS data and historical risk point data, and the optimization of node layout based on risk assessment, to obtain optimized sensor layout and network topology, specifically involves: Establish a risk assessment indicator system and calculate the comprehensive risk value R for each node / segment k. k : R k =αF k +βI k +γP k ; Where α, β, γ are weighting coefficients; F k For fault frequency, I k For the degree of influence and P k Risk probability; Based on the comprehensive risk value R k The pipeline nodes are sorted, and the node set K that can effectively monitor the area of ​​greatest risk impact is selected based on the pipeline network topology. Based on risk points and weights {(k,R k )}, deployable location {j}, total cost constraint C total Based on the maximum coverage location selection model, the spatial distribution of sensors after optimized layout is obtained: ； Where M is the set of possible deployment locations; a kj For the discriminant variable: if placing a sensor at deployment point j can cover risk k, then a kj= 1, otherwise a kj =0;C j The cost of installing the sensor at deployment point j; C total For budget; x j For decision variables: if a sensor is deployed at location j, x j =1, otherwise 0; y k For the coverage variable: if risk point k is covered by at least one sensor, y k =1, otherwise 0; The optimal communication route and data aggregation node location are determined by a topology optimization algorithm, network connectivity constraints are applied, and network energy consumption or latency is minimized.

3. A method for safe operation and maintenance of underground pipelines based on the Internet of Things as described in claim 2, characterized in that, The optimal communication route and data aggregation node location are determined through a topology optimization algorithm, network connectivity constraints are applied, and network energy consumption or latency is minimized, as detailed below: Based on the spatial distribution of the sensors after optimization, coordinate information is obtained, and the communication radius r0 of all nodes is defined. For any two nodes i and j, if their distance d ij If r ≤ r0, then a communication edge is established between the two, and all nodes are constructed into an undirected graph to reflect the wireless communication connection relationship; For the constructed undirected graph, the depth-first search algorithm is used to detect the connectivity of the entire sensor network. If the network is not strongly connected, relay nodes are added to ensure connectivity, thereby ensuring the smooth collection and transmission of subsequent data. Based on network connectivity, to reduce overall transmission energy consumption, a minimum spanning tree algorithm (such as Prim or Kruskal) is applied, using the physical distance between nodes as edge weights, to extract the backbone network structure with the minimum total weight. The optimization objective is to connect all nodes with the minimum total cost, resulting in an acyclic backbone structure. ； Based on network size, data traffic distribution, and transmission latency requirements, the number of sink nodes is determined, and the K-means clustering algorithm is used to cluster all sensor nodes x. i The coordinates are used to cluster the network, dividing it into M subnetworks C. m For each subnetwork C m , will converge node S m Assume that all nodes in this cluster are connected to S. m The location with the minimum sum of distances is chosen to minimize the total data transmission distance or power consumption of all sensor nodes within the area to their convergence node. ； Within each sub-network, for all ordinary nodes to the sink node, the Dijkstra shortest path algorithm is used to calculate the main path with the minimum energy consumption, with the edge weight as the link energy consumption. The final output includes the spatial distribution of each sensor and aggregation node, the optimal routing structure for energy consumption within the backbone network and each sub-network, and the overall network energy consumption performance indicators.

4. A method for safe operation and maintenance of underground pipelines based on the Internet of Things as described in claim 1, characterized in that, The preprocessing is as follows: To address noise, spikes, and outliers in the raw data, preprocessing algorithms such as moving average, median filtering, or wavelet denoising are used to remove distorted measurements and ensure data quality. After cleaning, the data is organized into a standardized data format according to network protocols and requirements of the upper-level system. Each data packet contains: sensor ID, type label, timestamp, raw or processed data value, and quality identification metadata; all fields are aligned according to the agreed standard fields. The standardized data is repackaged effectively using TLV and structure to optimize data length and parsing efficiency, preparing it for network transmission.

5. A method for safe operation and maintenance of underground pipelines based on the Internet of Things as described in claim 1, characterized in that, The process involves multimodal data fusion based on standardized data packets, historical operation and maintenance records, and GIS data to obtain the fused data, as detailed below: Feature extraction is performed on standardized data packets, historical operation and maintenance records, and GIS data, and feature alignment is completed based on common spatial, temporal, and entity ID features. Sensor data collection points are mapped and overlaid onto a GIS map according to their spatial location to achieve a one-to-one correspondence in physical location; real-time data is synchronized with historical operation and maintenance events according to a unified standard timeline to compare the operating status and maintenance operations within a specific time period; sensor IDs are used as indexes to establish the correlation of features from multiple data sources. The aligned features of the multi-source data are spliced ​​and combined to form a comprehensive feature set that includes environmental, geographical, and operational characteristics, resulting in fused data.

6. A method for safe operation and maintenance of underground pipelines based on the Internet of Things according to claim 5, characterized in that, Based on the fused data and a multimodal model, anomaly detection of underground pipelines is performed, and the anomaly detection results are obtained, as detailed below: Based on the fused data, the environmental, structural, and operational status data are respectively formed into vectors X(env), X(struct), and X(ops); GIS spatial adjacency information is organized into graph-structured data, and numerical features are normalized, time series data are synchronously completed, and the basis for multimodal joint input is established. First, input the parameterized multi-head self-attention Transformer encoders to extract the deep latent representation within the modality: Z (m) =MultiHeadAttn(Q (m) ,K (m) V (m) ),m∈{env,struct,ops}; Pipeline topology data is used to encode the spatial neighborhood characteristics of each monitoring point using a graph neural network: ； in, This represents the feature representation of node i in the network of the l-th layer graph neural network; Let i be a neighboring node of node i; The self-loop weight matrix of the l-th layer The transformation matrix represents the features of the neighbors, which projects the features of the neighboring nodes onto the current node. It is a non-linear activation function; A cross-attention mechanism is used to establish information flow among environmental, structural, and operational state modalities, enabling intermodal pairing and signal complementarity. ； in, and Let i and j be the representation vectors of modes i′ and j′, respectively; i′, j′ ∈ {env, struct, ops}; Information from mode j′ is passed to mode i′ through cross attention; To query the weight matrix, used to assign Z (i′) Linear transformation into a query vector; For the key weight matrix, Z (j′) A linear transformation is used to obtain a key vector; For the value weight matrix, put Z (j′) Transform into a value vector; d k is the dimension of the key vector; the superscript T indicates transpose; Based on a gating fusion mechanism, the contributions of the original modalities and the interaction modalities are adaptively balanced to improve the effectiveness of fusion: ； in, Indicates that Z (i′) With C i′,j′ Concatenate along the feature dimension; Here is the weight matrix of the gated network; is the bias vector of the gated network; Indicates the gating weight; ⊙ represents the element-wise product; The features of mode i′ after gated fusion; Multimodal features are stitched together in the fusion layer: ； Based on the fusion features, the output branch predicts the scores S for the three types of anomalies respectively. env, S struct ,S ops : ； Among them, W m b m Let m be the first-level weights and biases. w is the activation function. m b m′ For output layer weights and biases; It is the sigmoid function; Various anomalies are weighted and combined to form an overall risk score, which is then multiplied by the confidence level to produce the final anomaly strength: ； Determine if AnomalyScore > θ t θ t The threshold is used for abnormal warnings.

7. A method for safe operation and maintenance of underground pipelines based on the Internet of Things as described in claim 6, characterized in that, Based on the anomaly detection results, combined with official website GIS data and optimized sensor layout and network topology, risk points are accurately identified, as follows: The anomaly detection results output by the anomaly detection model, the corresponding high-risk sensors, pipe sections or nodes, are associated with the GIS database through the sensor ID to obtain their accurate location in the real geographic space and pipeline network. By utilizing the current optimized sensor layout and the latest network topology, we analyze the location attributes of anomalies in the pipeline system to determine whether they are located at critical nodes in the pipeline network. Combining pipeline flow direction, upstream and downstream relationships of nodes, and hydraulic conditions, we analyze the possible propagation paths and impact range of abnormal signals. On the GIS platform, with anomalies as the center, a safety buffer zone or risk impact radius is set. Combined with the distribution of historical risk points, spatial overlay analysis is performed on the surrounding pipe sections, important facilities and residential areas that may be affected to form a preliminary candidate set of high-risk areas. Taking into account anomaly scores, frequency of historical risk occurrences, topological importance, and spatial clustering of nodes, a weighted model is used to calculate the final risk score for each candidate point: ； Among them, S anom,i D represents the anomaly detection score. cluster,i For spatial clustering density, C topology,i H represents the topological centrality of the pipeline network. history,i Historical risks are weighted, with α, β, γ, and δ as weighting parameters; nodes with scores higher than the threshold are selected as the risk points locked in this instance.

8. A method for safe operation and maintenance of underground pipelines based on the Internet of Things according to claim 1, characterized in that, Based on the anomaly detection results and risk points, the data is visualized, and the detection report is distributed to the operations and maintenance personnel, as detailed below: Using a GIS platform, the precisely identified risk points are visually marked on the map, and different risk levels and anomaly types are represented by different colors, icons or heat maps; It also displays detailed attributes of the risk points, including: anomaly type, detection time, anomaly score, pipeline number, geographical location, surrounding infrastructure, and historical risk occurrence data; For high-risk points or the diffusion and linkage relationships between multiple risk points, dynamic layers are used to display the potential propagation links of risk events and the historical abnormal evolution process, helping operation and maintenance personnel to conduct trend tracking and causal analysis.

9. An Internet of Things-based underground pipeline safety operation and maintenance system, characterized in that, It includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the Internet of Things-based underground pipeline safety operation and maintenance method as described in any one of claims 1-8.

10. A computer storage medium, characterized in that, The computer storage medium stores a plurality of instructions, which are adapted to be loaded by a processor and executed as method steps as claimed in any one of claims 1 to 8.

Images