Underground pipeline monitoring method based on multi-source heterogeneous sensor data fusion and storage medium

Abstract

The application discloses a kind of multi-source heterogeneous sensor data fusion underground pipeline monitoring method and storage medium, comprising the following steps: obtaining the static surveying and mapping data of target underground pipeline area, and according to static surveying and mapping data, construct three-dimensional geometric model with unique asset coding;Sensor is deployed in the selected physical position of target underground pipeline area, and the identifier of each sensor is bound with the asset coding of deployment position;Receive dynamic monitoring data collected by each sensor;Dynamic monitoring data is automatically associated and mapped to model entity with the same asset coding in three-dimensional geometric model;Based on the dynamic monitoring data mapped to model entity, drive physical simulation model to calculate, generate state prediction result corresponding to model entity;The method and storage medium can deeply fuse static spatial information and dynamic operation data, and can simulate, predict physical entity behavior in virtual space.

Description

Technical Field

[0001] This invention relates to the field of computer surveying technology, and in particular to a method and storage medium for monitoring underground pipelines using multi-source heterogeneous sensor data fusion. Background Technology

[0002] Underground pipelines (including power cables, oil and gas pipelines, water supply and drainage pipelines, etc.) are the "lifelines" of urban and industrial infrastructure, and their safe and stable operation is of paramount importance. With the development of information and intelligent technologies, three-dimensional visualization modeling and online status monitoring of underground pipelines have become an industry trend.

[0003] Currently, the industry mainly adopts the following two types of technical solutions: 1. 3D modeling and management based on static surveying data: According to standards such as the "Technical Specification for Urban Underground Pipeline Detection" (CJJ 61-2003), static attribute data such as spatial coordinates, burial depth, and pipe diameter of pipelines are obtained through surveying equipment such as GPS and total stations, and detailed 3D geometric models are constructed using 3D modeling software. This type of model can accurately reflect the spatial layout and static archives of pipelines, providing an intuitive tool for planning, design, and asset management. However, this type of model is essentially a "static snapshot," which is fixed once built and cannot reflect the real-time status changes of pipelines during operation caused by load changes, environmental erosion, geological activities, etc., such as cable joint temperature rise, small pipe displacement, and soil settlement. 2. Independent Internet of Things (IoT) monitoring system: Sensors such as temperature, displacement, stress, and gas concentration are deployed in some important pipeline sections. The operating status data is collected in real time through an IoT platform and displayed graphically and with over-threshold alarms on independent monitoring software or large screens. However, this type of system is usually isolated from the 3D spatial model. Maintenance personnel need to frequently switch between the 3D model system and the monitoring system and manually compare data to associate a specific alarm message with a specific spatial location. This disconnect between "data" and "spatial context" leads to unintuitive status perception, difficulty in information fusion, and low efficiency in anomaly localization, which is particularly detrimental to rapid decision-making in emergency situations.

[0004] In addition, existing technologies also have the following deeper problems: Low level of automation in data association: Existing methods rely heavily on complex manual configuration or fuzzy matching based on spatial coordinates when associating sensor data with 3D models. The process is cumbersome, error-prone, and difficult to adapt to large-scale, dynamically changing sensor networks.

[0005] The monitoring deployment lacks scientific rigor: sensor placement often relies on experience or simple, uniform coverage, failing to link it to the importance and risk level of assets, resulting in wasted monitoring resources or blind spots in key areas.

[0006] Lack of mechanistic models and predictive capabilities: Existing monitoring is mostly limited to displaying the current state and threshold alarms, which is a reactive response. There is a lack of mechanistic model analysis based on pipeline physical characteristics (such as heat conduction, stress and strain), making it impossible to simulate and predict operating trends, and difficult to achieve "predictive maintenance".

[0007] Insufficient system closed-loop capability: The chain from status perception and early warning to operation and maintenance is broken, and early warning information is difficult to automatically convert into executable work orders or control instructions, making it impossible to form an intelligent operation and maintenance closed loop of "perception-analysis-decision-execution-optimization".

[0008] Therefore, existing technologies urgently need a new method that can deeply integrate static spatial information with dynamic operational data, and simulate and predict the behavior of physical entities in virtual space, thereby achieving forward-looking and intelligent management of underground pipelines throughout their entire life cycle. Summary of the Invention

[0009] To address the shortcomings of the existing technologies, the present invention aims to provide a method and storage medium for monitoring underground pipelines by fusing multi-source heterogeneous sensor data, which can deeply integrate static spatial information and dynamic operational data and simulate and predict the behavior of physical entities in virtual space.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: A method for monitoring underground pipelines using multi-source heterogeneous sensor data fusion, comprising the following steps: Acquire static mapping data of the target underground pipeline area, and construct a three-dimensional geometric model with a unique asset code based on the static mapping data; Based on a preset deployment strategy, sensors are deployed at selected physical locations in the target underground pipeline area, and the identifier of each sensor is bound to the asset code of the deployment location. Receive dynamic monitoring data collected by each of the sensors, wherein the dynamic monitoring data encapsulates a corresponding sensor identifier; Based on the binding relationship between the sensor identifier and the asset code, the dynamic monitoring data is automatically associated and mapped to model entities with the same asset code in the three-dimensional geometric model; Based on the dynamic monitoring data mapped onto the model entity, the physical simulation model is driven to perform calculations and generate state prediction results corresponding to the model entity. Based on the dynamic monitoring data and the state prediction results, the three-dimensional geometric model is visualized for state updates and early warning displays.

[0011] Preferably, the step of automatically associating and mapping the dynamic monitoring data based on the binding relationship between the sensor identifier and the asset code specifically includes: Parse the received dynamic monitoring data packets and extract the sensor identifiers from them; Query the pre-set binding relationship list to obtain one or more asset codes corresponding to the sensor identifier; In the three-dimensional geometric model, locate the model entity whose asset code matches the acquired asset code; The monitoring values ​​from the dynamic monitoring data are assigned to the located model entity as its real-time state attributes.

[0012] Preferably, for monitoring parameters that require the generation of a spatially continuous distribution field, the method further includes: For a model entity associated with the same asset code, obtain the monitoring values ​​and preset logical locations of all the sensors bound to it; Based on the monitored values ​​and logical locations, a spatial interpolation algorithm is used to calculate the parameter distribution field of the entire model entity. The calculated parameter distribution field is mapped and rendered as a state cloud map covering the surface of the model entity.

[0013] Preferably, the step of deploying sensors at selected physical locations based on a preset deployment strategy specifically includes: In the aforementioned three-dimensional geometric model, key asset codes are selected based on risk analysis models or historical operation and maintenance data. Based on the selected key asset codes, generate sensor deployment work orders for the corresponding physical entities; According to the deployment work order, on-site personnel install sensors on the physical pipelines or ancillary facilities corresponding to the key asset codes and complete the binding information entry of sensor identifiers and asset codes.

[0014] Preferably, the sensor acquires dynamic monitoring data using a dynamic interval strategy, specifically including: When the monitored parameters are within the normal range, the sensor collects and reports data at the first preset cycle. When the monitored parameters exceed the preset threshold, the sensor automatically switches to a second preset period shorter than the first preset period for high-frequency acquisition and real-time reporting.

[0015] Preferably, the method further includes coordinated scheduling of large-scale sensor clusters: Assign different dedicated wake-up time slots to each sensor node in the network; Each sensor node is only woken up and performs data acquisition or reporting within its assigned wake-up time slot, and enters a low-power sleep state in other time slots.

[0016] Preferably, the physical simulation model is a thermo-mechanical coupling simulation model based on finite element analysis, used to calculate the stress distribution of the pipeline structure based on real-time temperature data and predict potential stress concentration areas.

[0017] Preferably, the step of visualizing the status update and displaying early warnings for the three-dimensional geometric model specifically includes: The state prediction result is compared with a preset safety threshold. When the state prediction result exceeds the safety threshold, the corresponding model entity is highlighted in the three-dimensional geometric model, and a warning message containing the asset code, warning type and prediction value is generated and pushed.

[0018] Preferably, the method further includes the step of forming an operation and maintenance closed loop: Record early warning information and corresponding response decisions; Based on the decision-making process, control instructions or maintenance plans are generated, and the execution results are tracked. The execution results are fed back to the system to optimize the parameters of the risk analysis model or the physical simulation model.

[0019] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described above.

[0020] Compared with the prior art, the present invention has the following beneficial effects: 1. By establishing and utilizing a unique asset coding system, this invention innovatively achieves the logical binding of sensor identifiers with 3D model entities. Dynamic monitoring data can be automatically and losslessly associated with and mapped to the corresponding 3D model based on this binding relationship, completely changing the traditional methods of manual configuration or complex coordinate matching. This greatly improves the automation, accuracy, and efficiency of data fusion, laying the foundation for building high-fidelity digital twins.

[0021] 2. In a unified 3D visualization environment, each pipeline or facility not only displays its static attributes but also presents its dynamic status in real time, including temperature cloud maps, displacement vectors, stress distribution, and simulation prediction results. Maintenance personnel can see the results promptly, gaining a clear understanding of the overall situation and local details, greatly reducing the information literacy threshold and making anomaly location, root cause analysis, and emergency decision-making more intuitive, rapid, and accurate.

[0022] 3. Through a "digital deployment strategy based on risk analysis models or historical data," operations and maintenance personnel can intelligently select key assets for monitoring and deployment in a 3D model, and the system automatically generates deployment work orders. This approach achieves a shift from "experience-driven" or "blind full coverage" to "data-driven, precise investment," significantly reducing the overall cost of sensor deployment and maintenance while ensuring that key risk points are effectively monitored.

[0023] 4. Through a "dynamic interval acquisition and collaborative scheduling" mechanism, the system saves energy and network bandwidth by performing low-frequency inspections when parameters are normal, and automatically switches to high-frequency real-time reporting when parameters are abnormal. Combined with "dedicated wake-up time slots" allocated for large-scale sensor clusters, network conflicts are effectively avoided, ensuring system stability. This innovation achieves high real-time anomaly response under low power consumption and low network load, optimizing the overall system performance.

[0024] 5. By introducing "simulation models based on physical laws (such as thermo-mechanical coupled finite element models)," digital twins can simulate and calculate the stress distribution and heat conduction within pipelines based on real-time data, and proactively predict potential hot spots, stress concentration zones, and other risks. This upgrades the management model from passive "post-event alarms" to proactive "pre-event warnings," significantly improving the predictability and proactivity of safety management.

[0025] 6. Through a closed-loop process of "recording-decision-execution-feedback," maintenance plans or control instructions can be automatically generated based on early warning information, and the results of handling can be tracked to optimize the model. This forms an intelligent operation and maintenance ecosystem of "continuous evaluation and self-optimization," which significantly improves the automation and intelligence level of pipeline system management and provides a complete technical path for achieving predictive maintenance. Attached Figure Description

[0026] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the underground pipeline monitoring method based on multi-source heterogeneous sensor data fusion according to the present invention. Figure 2 This is a flowchart of the time series technology of the present invention. Detailed Implementation

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0028] It should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or device.

[0029] refer to Figure 1 This embodiment uses the construction and application of a digital twin monitoring system in the underground pipeline group of a ±800kV converter station as an example for illustration, but the application of the present invention is not limited to this scenario.

[0030] Implementation Process of a Digital Twin Monitoring Method for Underground Pipelines in High-Voltage Converter Stations The implementation of this embodiment relies on a complete hardware and software system, whose logical architecture follows the process of data acquisition, fusion, simulation and application.

[0031] S101: Constructing a 3D geometric model with asset codes First, a total station and GPS combined survey was conducted on approximately 20 kilometers of underground pipeline network within the target area (a ±800kV converter station). The survey strictly followed the "Technical Specification for Urban Underground Pipeline Detection" (CJJ 61-2003) to obtain static survey data, including the three-dimensional coordinates (accuracy: plane ≤ ±5cm, elevation ≤ ±3cm), pipe diameter, material, burial depth, and laying year of all pipeline feature points.

[0032] Core operation: Asset coding and binding. When constructing detailed 3D geometric models using a 3D engine (such as a custom platform based on Unity3D), each model entity is assigned a globally unique "asset code." The coding rule adopts a "facility type-region-serial number" structure, for example: Power cable: DLQ-03A-015 (indicating cable number 15 in area 03A) Manhole: GJ-203 (indicating manhole number 203) Connector: JT-DLQ-03A-015-01 (indicates connector number 1 of DLQ-03A-015 cable) This code is completely consistent with the nameplates and asset ledgers of pipelines in the physical world, forming the "identity card" of the digital twin. The completed model not only has a geometric shape, but is also a semantic model that carries complete asset attributes and coded information.

[0033] S102: Deploy and bind sensors based on digital strategies The deployment of sensors is not about blindly covering everything, but rather about making intelligent decisions based on digital models.

[0034] Step 1 (Digital Deployment Decision): Maintenance personnel utilize the integrated risk analysis model on the 3D digital twin platform. This model integrates factors such as pipeline load history, fault records, soil corrosivity data, and spatial location (e.g., proximity to roads) to assess the risk of all pipeline assets at the station. The system automatically selects the top 20% of asset codes with the highest risk scores (e.g., DLQ-03A-015, GJ-101) and marks them as critical assets. Subsequently, based on these critical asset codes, the platform automatically generates detailed sensor deployment work orders, specifying the installation location (asset code), sensor type (e.g., cable surface temperature, well water level), and technical requirements.

[0035] Step Two (On-site Installation and Information Binding): On-site personnel, according to the work order, install a temperature sensor (identifier Temp_015_Mid) at the intermediate joint of cable with asset code DLQ-03A-015, and a water level sensor (identifier WL_101) inside manhole GJ-101. After installation, using a mobile terminal APP, scan the QR code on the sensor entity (identifier) ​​and the QR code on the on-site asset nameplate (asset code) to complete the binding. The binding information (Temp_015_Mid -> DLQ-03A-015, WL_101 -> GJ-101) is uploaded in real-time to the binding relationship list in the central database. This list serves as a "routing table" for the automatic association of all subsequent data.

[0036] S103: Receive dynamic monitoring data The deployed sensors have begun operating, employing an intelligent dynamic interval strategy for data acquisition: Normal low-frequency mode: When the temperature sensor monitoring value is within a safe range (e.g., <65℃), data is collected once every first preset period (30 seconds). However, in order to save communication power consumption and network bandwidth, the average value is packaged and reported to the IoT platform every 10 readings (i.e., 5 minutes).

[0037] Abnormal High-Frequency Mode: On a certain day, the sensor detected a temperature rise to 68℃ (exceeding the preset threshold of 65℃), immediately triggering a mode switch. High-frequency data acquisition is performed at a second preset period (2 seconds), and each data packet is reported in real time (network latency <16 seconds). The data packet format is JSON: {“sensor_id”: “Temp_015_Mid”, “value”: 68.2, “timestamp”: “2023-10-27T14:05:30Z”}.

[0038] For large-scale sensor clusters, IoT gateways employ a time-division multiple access (TDMA) cooperative scheduling algorithm. The gateway divides a day into numerous time slots (e.g., each slot is 200ms) and allocates a dedicated wake-up time slot to each node based on its ID hash value. The sensor Temp_015_Mid only wakes up, collects, and transmits data within its designated time slot, then immediately enters a low-power sleep state until its next designated time slot arrives. This fundamentally avoids network congestion and data collisions that can occur when a large number of nodes report simultaneously.

[0039] S104: Automatic Association Mapping of Dynamic Data and 3D Models The IoT platform forwards the received data packets to the data fusion engine of the digital twin system.

[0040] Step 1 (Parsing and Querying): The fusion engine parses the data packet and extracts the key field: sensor_id = "Temp_015_Mid". Then, it queries the "binding relationship list" stored in the database to quickly find the asset code corresponding to this sensor ID: DLQ-03A-015.

[0041] Step Two (Location and Attribute Binding): The fusion engine locates the cable model entity with asset code DLQ-03A-015 in the asset database of the 3D scene. Then, it assigns the value=68.2 and timestamp from the data packet as real-time attributes to this model entity. In the 3D interface, when the user clicks on this cable, the attribute panel displays the current temperature: 68.2℃ (update time: 14:05:30).

[0042] Step 3 (Spatial Interpolation and Cloud Map Generation): To display the temperature distribution of the entire cable rather than a single point, the system initiates a spatial interpolation algorithm and a time series alignment technique. Assuming that three temperature sensors (including Temp_015_Mid) are deployed on cable DLQ-03A-015, the system acquires the real-time temperature values ​​of these three sensors (whose preset logical positions on the cable model are known, such as percentages from the starting point).

[0043] Algorithm Application: Kriging interpolation is employed. This algorithm considers not only the distance between known and unknown points but also analyzes the spatial autocorrelation between known points using a variogram. The calculation process is as follows: 1. Calculate the experimental variability function based on the temperature values ​​at three known points.

[0044] 2. Fit a theoretical variogram model (such as a spherical model).

[0045] 3. For any point on the cable to be interpolated, use the fitted variogram model to calculate the spatial correlation weight between that point and all known points.

[0046] 4. Using these weights, perform a weighted average of the temperature values ​​at the known points to obtain the optimal linear unbiased estimate of the temperature at that point.

[0047] refer to Figure 2 Time series alignment techniques In the "digital twin system for underground pipelines based on multi-source heterogeneous sensor data fusion", there exists a fundamental contradiction: In the real world, at a precise instant, the temperature at point A, the displacement at point B, and the gas concentration at point C on a cable are simultaneously present, collectively forming a complete "health snapshot" of the pipeline at that moment. However, the data collected by your system is collected at different frequencies: temperature sensors may report every 20 seconds, displacement sensors every 5 minutes, and gas sensors every 10 minutes. Even within the same second, due to minute differences in chip processing and network transmission, the timestamps of data packets arriving at the server can deviate by milliseconds. There is also a variable network latency in data transmission from the sensors to the central server.

[0048] Without time series alignment, directly "pasting" the latest received sensor data with different timestamps onto a 3D model for simulation or cloud map generation will lead to an absurd result: the data you use to calculate the "current" thermal stress is actually the temperature from 20 seconds ago, the displacement from 2 minutes ago, and the gas data from 8 minutes ago. This "spatiotemporal misalignment" in data fusion will distort the state of the digital twin, rendering simulation predictions meaningless, and early warnings may produce false alarms or missed alarms.

[0049] The purpose of time series alignment is to reconstruct a snapshot of the physical world's state at every "moment" in the digital world. Its core process, as shown in the diagram below, acts as a crucial data preprocessing module, transforming asynchronous "data streams" into synchronous "state snapshots."

[0050] The implementation of this technology follows the specific steps below: Step 1: Establish the "Target Timeline" The system defines a virtual, uniform target time series, such as [T0, T1, T2, ...], where each T_n interval may be 1 second or the step size of the simulation engine. All subsequent fusion and analysis will be based on these "target times".

[0051] Step 2: Data interpolation estimation For each "target time" T, the system needs to know the "estimated value" of each sensor at time T. Since the sensor may not report data exactly at time T, interpolation is required.

[0052] Specific method: The system will find the two real data points (t1, v1) and (t2, v2) with timestamps closest to T in the data cache of the sensor.

[0053] Algorithm (taking linear interpolation as an example): Estimated value V_T = v1 + ( (T - t1) / (t2 - t1) ) * (v2 -v1). This assumes that the data changes linearly between two measured points.

[0054] Result: Through this operation, the original state in which only temperature had measured values ​​at time T, while displacement and gas data were missing, was transformed into a complete dataset in which all three had reasonable estimated values ​​at time T.

[0055] Step 3: Output a snapshot of the synchronized data The alignment engine outputs a series of synchronized data packets arranged along the "target time axis." Each data packet contains aligned data from all sensors at time T. This data is strictly consistent in the time dimension and can be safely used for subsequent fusion and simulation.

[0056] This time series alignment technique works in conjunction with "spatial interpolation": To generate a temperature cloud map of a cable at time T, data from multiple temperature sensors on the cable at time T are required.

[0057] Time alignment first ensures that all sensor data used for spatial interpolation are values ​​at time T, not values ​​from different times. Only in this way can the generated contour map represent a true, simultaneous temperature distribution cross-section.

[0058] Time series alignment techniques are also closely coupled with other techniques, which together determine the accuracy of the system: 1. Collaborate with "Collecting Dynamic Monitoring Data": When the sensor is in a normal low-frequency mode (such as reporting every 5 minutes), the alignment engine interpolates based on sparse data points to generate a state snapshot every minute, which is sufficient to maintain the basic situation.

[0059] When a sensor triggers an abnormally high-frequency mode (such as reporting per second), the alignment engine automatically increases the density of the target timeline (e.g., also becoming per second) and uses the denser data for high-precision interpolation. This ensures that, in the early stages of an incident, the digital twin can synchronously track and simulate the evolution of the anomaly with "slow-motion" timing accuracy.

[0060] 2. Collaboration with "Physical Simulation Models": Physical models such as finite element simulation require that the input parameters (temperature, pressure, boundary conditions) must be in the same state at the same time step.

[0061] The time series alignment technique provides the simulation kernel with "time step consistent input data packets", which enables the simulation to perform calculations based on correct causal relationships, thus making the predictions reliable.

[0062] Time series alignment technology is not only a background tool, but also a key technology for achieving "dynamic perception" and "predictive early warning".

[0063] Its core values ​​include: It ensures the "spatiotemporal consistency" of the digital twin: This is the lifeline of high-fidelity digital twins, ensuring a strict correspondence between the virtual model and the physical entity in time and state.

[0064] It unlocks precise simulation and prediction capabilities: providing clean and synchronous input data for physical simulation models, enabling trend prediction and stress analysis based on mechanistic models, and achieving a leap from "monitoring the current situation" to "predicting the future".

[0065] The system's response accuracy to emergencies has been improved: by linking with dynamic data acquisition strategies, it can automatically switch to a high-precision alignment mode when an anomaly occurs, providing a highly timely data base for accurate early warning and decision-making.

[0066] Visualization and Rendering: Through the above calculations, the estimated temperature of each point on the cable is obtained, forming a continuous temperature distribution field. The graphics engine renders it as a color cloud map with a gradient from blue (low temperature) to red (high temperature) based on the temperature, overlaying it on the surface of the 3D cable model to intuitively present the location of the "hot spots".

[0067] S105: State Prediction Based on Physical Simulation Model Once real-time temperature data (especially abnormally high temperature data) is mapped onto the model, the system automatically triggers the physical simulation kernel to perform calculations. In this example, the simulation kernel is a thermo-mechanical coupling model based on finite element analysis.

[0068] Finite element modeling: The simulation model of the cable has been established in advance, and the cable structure (conductor, insulation layer, sheath) has been discretized into thousands of finite element meshes.

[0069] Boundary condition loading: The cable surface temperature distribution field generated in step S104 is used as the boundary condition and loaded onto the corresponding nodes of the finite element model.

[0070] The simulation kernel solves the heat conduction differential equation to calculate the precise temperature of the cable conductor core (typically 10-15°C higher than surface measurements). Then, based on this temperature field and the material's coefficient of thermal expansion, the mechanical equilibrium equation is solved to calculate the distribution of thermal stress in the cable insulation and metal sheath caused by thermal expansion and contraction.

[0071] Predicted Output: Simulation results identify regions where thermal stress exceeds 80% of the material's allowable stress and mark them as potential stress concentration risk areas. The predicted results (stress values, risk area coordinates) are linked back to asset code DLQ-03A-015.

[0072] S106: Visual Status Updates and Early Warnings The digital twin platform updates the 3D scene synchronously.

[0073] 1. Status Update: The color of cable DLQ-03A-015 changes in real time according to the temperature cloud map, and the attribute panel is updated synchronously with the core temperature and maximum stress value calculated by simulation.

[0074] 2. Warning Trigger: The warning module compares the maximum thermal stress value calculated by simulation (125 MPa) with the safety threshold (100 MPa) for this type of cable in the material library. Since 125 MPa > 100 MPa, the warning condition is triggered.

[0075] 3. Visual Early Warning: In the 3D model, the section of cable DLQ-03A-015 with excessive stress begins to flash red. Simultaneously, the system automatically generates a structured early warning message: Asset under warning: DLQ-03A-015; Type: Excessive thermal stress; Predicted value: 125MPa; Threshold: 100MPa; Recommendation: Check the load and schedule infrared retesting. This information is pushed to relevant maintenance personnel via pop-up windows, SMS, and mobile app.

[0076] S107: Form a closed loop for operation and maintenance and optimize the model Early warning is not the end, but the starting point of intelligent operation and maintenance.

[0077] 1. Recording and Decision-Making: The system records the entire context of this alert. Based on the alert information, the operations manager makes a decision in the system: "Issue an inspection work order".

[0078] 2. Work Order Generation and Execution: Based on the decision, the system automatically generates a digital work order, associates it with asset DLQ-03A-015, and sends it to the mobile terminal of the inspection personnel. After on-site confirmation, the inspection personnel provide feedback on the terminal: "The on-site retest temperature is too high; an application has been made to adjust the load on this line."

[0079] 3. Feedback and Optimization: After load adjustment, new temperature data is transmitted back to the system. The system records the complete causal chain of "load reduction X% → temperature drop Y℃ → stress drop to Z MPa" as a closed-loop case. This case data is fed into the risk analysis model and the physical simulation model. Based on this, the risk analysis model learns and assigns a higher risk weight to cables under similar operating conditions in the future; the physical simulation model may use this data to calibrate its thermal conductivity coefficient, making the next prediction more accurate. Thus, a complete intelligent closed loop of "perception-early warning-decision-execution-learning" is completed.

[0080] S108: System Carrier All steps S101 to S107 described above are implemented by digital twin monitoring platform software deployed on a server. This software is written as a computer program and stored in a computer-readable storage medium (such as a solid-state drive) in the data center. When the server processor loads and executes this program, all the methods described above are fully implemented.

[0081] This embodiment details a complete technical process from static modeling, intelligent site selection, data acquisition, automatic fusion, simulation prediction to visualization, early warning, and closed-loop optimization. By expanding and specifying each step (especially the asset coding and binding process, Kriging interpolation algorithm, finite element simulation process, dynamic acquisition and scheduling logic, and closed-loop optimization mechanism), the feasibility, advancement, and completeness of the invention's technical solution are fully demonstrated. This method successfully upgrades underground pipeline management from a static, passive, and experience-driven model to a dynamic, predictive, and data-driven intelligent digital twin model.

[0082] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for monitoring underground pipelines using multi-source heterogeneous sensor data fusion, characterized in that, Includes the following steps: Acquire static mapping data of the target underground pipeline area, and construct a three-dimensional geometric model with a unique asset code based on the static mapping data; Based on a preset deployment strategy, sensors are deployed at selected physical locations in the target underground pipeline area, and the identifier of each sensor is bound to the asset code of the deployment location. Receive dynamic monitoring data collected by each of the sensors, wherein the dynamic monitoring data encapsulates a corresponding sensor identifier; Based on the binding relationship between the sensor identifier and the asset code, the dynamic monitoring data is automatically associated and mapped to model entities with the same asset code in the three-dimensional geometric model; Based on the dynamic monitoring data mapped onto the model entity, the physical simulation model is driven to perform calculations and generate state prediction results corresponding to the model entity. Based on the dynamic monitoring data and the state prediction results, the three-dimensional geometric model is visualized for state updates and early warning displays.

2. The method according to claim 1, characterized in that, The step of automatically associating and mapping the dynamic monitoring data based on the binding relationship between the sensor identifier and the asset code specifically includes: Parse the received dynamic monitoring data packets and extract the sensor identifiers from them; Query the pre-set binding relationship list to obtain one or more asset codes corresponding to the sensor identifier; In the three-dimensional geometric model, locate the model entity whose asset code matches the acquired asset code; The monitoring values ​​from the dynamic monitoring data are assigned to the located model entity as the real-time status attribute of the model entity.

3. The method according to claim 2, characterized in that, For monitoring parameters that require the generation of a spatially continuous distribution field, the method further includes: For a model entity associated with the same asset code, obtain the monitoring values ​​and preset logical locations of all bound sensors on that model entity; Based on the monitored values ​​and logical locations, spatial interpolation algorithms and time series alignment techniques are used to calculate the parameter distribution field of the entire model entity. The calculated parameter distribution field is mapped and rendered as a state cloud map covering the surface of the model entity.

4. The method according to claim 1, characterized in that, The step of deploying sensors at selected physical locations based on a preset deployment strategy specifically includes: In the aforementioned three-dimensional geometric model, key asset codes are selected based on risk analysis models or historical operation and maintenance data. Based on the selected key asset codes, generate sensor deployment work orders for the corresponding physical entities; According to the deployment work order, on-site personnel install sensors on the physical pipelines or ancillary facilities corresponding to the key asset codes and complete the binding information entry of sensor identifiers and asset codes.

5. The method according to claim 1, characterized in that, The sensor employs a dynamic interval strategy to collect dynamic monitoring data, specifically including: When the monitored parameters are within the normal range, the sensor collects and reports data at the first preset cycle. When the monitored parameters exceed the preset threshold, the sensor automatically switches to a second preset period shorter than the first preset period for high-frequency acquisition and real-time reporting.

6. The method according to claim 5, characterized in that, The method also includes coordinated scheduling of the sensor cluster: Different dedicated wake-up time slots are allocated to each sensor node within the network; Each sensor node is only woken up and performs data acquisition or reporting within its assigned wake-up time slot, and enters a low-power sleep state in other time slots.

7. The method according to claim 1, characterized in that, The physical simulation model is a thermo-mechanical coupling simulation model based on finite element analysis, used to calculate the stress distribution of the pipeline structure based on real-time temperature data and predict potential stress concentration areas.

8. The method according to claim 1, characterized in that, The steps for visualizing the status update and providing early warnings for the three-dimensional geometric model specifically include: The state prediction result is compared with a preset safety threshold. When the state prediction result exceeds the safety threshold, the corresponding model entity is marked in the three-dimensional geometric model, and a warning message containing asset code, warning type and prediction value is generated and pushed.

9. The method according to claim 1, characterized in that, The method also includes the step of forming a closed loop of operation and maintenance: Record early warning information and corresponding response decisions; Based on the decision-making process, control instructions or maintenance plans are generated, and the execution results are tracked. The execution results are fed back to the model to optimize the parameters of the risk analysis model or the physical simulation model.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 9.

Images