Pipe gallery disease monitoring and diagnosing method based on multi-source heterogeneous data

Abstract

The invention provides a pipe gallery disease monitoring and diagnosis method based on multi-source heterogeneous data, and relates to the technical field of disease monitoring and diagnosis, and the method comprises the steps: carrying out the data collection through a multi-mode sensor network disposed in an underground pipe gallery; fusing the multi-source heterogeneous data set based on a graph neural network to generate a dynamic sensing characteristic spectrum of the whole domain of the pipe gallery; carrying out disease evolution mode identification, and determining a pipe gallery disease risk level; and triggering the self-adaptive early warning strategy, executing the self-adaptive early warning strategy to generate a decision instruction set, and pushing the decision instruction set to the visual monitoring platform. The technical problem that potential problems are difficult to find in time and the operation efficiency of the pipe gallery is affected due to the fact that detection and risk assessment of the pipe gallery diseases depend on periodicity is solved, real-time monitoring and early recognition of the underground pipe gallery diseases are achieved through effective integration and processing of the multi-source heterogeneous data, and the method and the device have the advantages of being high in practicability and the like. And the efficiency and the safety of pipe gallery management are improved.

Description

Technical Field

[0001] The present application relates to the technical field of disease monitoring and diagnosis, and in particular to a pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data. Background Art

[0002] Defect monitoring and diagnosis in underground utility corridors involves using various monitoring methods and technical means to conduct real-time or periodic health assessments of the structure and internal facilities (such as pipes, cables, and drainage systems) of underground utility corridors, promptly identifying and diagnosing potential defects or faults. Currently, defect detection and risk assessment in underground utility corridors rely on periodic inspections. Manual inspections of equipment and facilities within the corridor are performed at set intervals to determine the corridor's operational status and potential hazards. Because inspection intervals are typically fixed, potential defects that occur between inspection cycles may go undetected. For example, further deterioration of defects between the end of an inspection cycle and the next inspection could lead to functional failure of the corridor equipment or a serious accident.

[0003] In summary, the existing technology has a technical problem that it relies on periodic detection and risk assessment of pipeline corridor defects, which makes it difficult to detect potential problems in a timely manner, thereby affecting the efficiency and safety of pipeline corridor operation. Summary of the Invention

[0004] The purpose of this application is to provide a pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data to solve the technical problem in the existing technology that it is difficult to detect potential problems in a timely manner due to reliance on periodic pipeline corridor disease detection and risk assessment, thereby affecting the efficiency and safety of pipeline corridor operation.

[0005] In view of the above problems, the present application provides a pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data, which includes: collecting data through a multimodal sensor network deployed in the underground pipeline corridor to construct a multi-source heterogeneous data set; fusing the multi-source heterogeneous data set based on a graph neural network to generate a dynamic perception feature map of the entire pipeline corridor; performing disease evolution pattern recognition on the dynamic perception feature map to determine the pipeline corridor disease risk level; triggering an adaptive early warning strategy according to the pipeline corridor disease risk level, executing the adaptive early warning strategy to generate a decision instruction set, and pushing the decision instruction set to a visualization monitoring platform.

[0006] Optionally, the multimodal sensor network includes distributed fiber optic sensors, UWB positioning base stations and patrol robots; data is collected through the distributed fiber optic sensors to generate a first time series data stream; data positioning of the pipeline corridor is performed based on the UWB positioning base station to generate second spatiotemporal trajectory data; the patrol robot is controlled by a lidar to scan along a preset path to obtain three-dimensional point cloud data of the pipeline corridor; the first time series data stream, the second spatiotemporal trajectory data and the three-dimensional point cloud data of the pipeline corridor are standardized and preprocessed using edge computing nodes to obtain the multi-source heterogeneous data set.

[0007] Optionally, an attention mechanism is used to dynamically allocate weights of multi-source data, and spatial correlation features are extracted through convolution calculations; sensor nodes are mapped to graph vertices according to the spatial correlation features, and multi-source data weights are mapped to dynamic edge weights; the corridor topology structure is obtained, and a graph neural network is constructed according to the corridor topology structure combined with the dynamic edge weights and the graph vertices; the first time series data stream, the second spatiotemporal trajectory data and the corridor three-dimensional point cloud data are fused with spatiotemporal features through the graph neural network to generate the dynamic perception feature map.

[0008] Optionally, the dynamic perception feature map is traversed to perform deep residual analysis to obtain multiple residual data; based on the multiple residual data, the pipeline corridor is simulated for damage to obtain a simulated evolution path; degradation judgment is performed according to the simulated evolution path, and the disease development stage is determined according to the judgment result; the life of the pipeline corridor is calculated according to the disease development stage to obtain the remaining service life data of the pipeline corridor; the disease development stage and the remaining service life of the pipeline corridor are correlated and integrated to determine the disease evolution pattern.

[0009] Optionally, a risk diffusion analysis is performed on the pipeline corridor according to the disease development stage and the remaining service life of the pipeline corridor to obtain a risk diffusion coefficient; the risk diffusion coefficient is used as an index to retrieve the disease evolution pattern and set three risk levels; an analysis of the functional impact of the disease on the pipeline corridor is performed, and the three risk levels are weighted and adjusted according to the functional impact coefficient, and the pipeline corridor disease risk level is dynamically divided.

[0010] Optionally, a multi-level risk threshold matrix is ​​set according to the pipeline corridor disease risk level, and the multi-level risk threshold matrix includes a first-level threshold and a second-level threshold; the dynamic perception feature map is traversed to extract the pipeline corridor disease index, and the pipeline corridor disease index is matched and determined with the multi-level risk threshold matrix; it is judged whether the pipeline corridor disease index exceeds the first-level threshold, when the pipeline corridor disease index exceeds the first-level threshold, the inspection robot is activated to perform local re-testing and verification on the pipeline corridor, and a re-testing and verification result is generated; based on the re-testing and verification result, the pipeline corridor disease index is updated, the pipeline corridor disease update index is determined, and it is judged whether the pipeline corridor disease update index exceeds the second-level threshold, when the pipeline corridor disease index exceeds the second-level threshold, the adaptive early warning strategy is triggered.

[0011] Optionally, the adaptive early warning strategy is executed, the disease development stage of the disease evolution model is called, the disease development stage is matched with the three-dimensional point cloud data of the pipeline corridor, and a stress concentration area vector diagram is generated; a historical repair case library is introduced, and the historical repair case library is searched for similar diseases to obtain disease construction record data; the disease construction record data is mapped to the stress concentration area vector diagram for planning, multiple groups of emergency arrival paths are determined, and the decision instruction set is formulated according to the multiple groups of emergency arrival paths.

[0012] Optionally, the decision instruction set is pushed to the visual monitoring platform, and the three-dimensional point cloud data of the pipeline corridor is dynamically annotated based on the decision instruction set to obtain multiple disease location labels; damage repair analysis is performed according to the multiple disease location labels, and a repair priority heat map is constructed; and damage superimposed navigation of the pipeline corridor is performed according to the repair priority heat map through the visual monitoring platform.

[0013] The technical solution provided in this application has at least the following beneficial effects: Data is collected through a multimodal sensor network deployed in the underground tunnel to construct a multi-source heterogeneous data set. This data set is then fused based on a graph neural network to generate a dynamic perception feature map for the entire tunnel. Disease evolution patterns are identified on this dynamic perception feature map to determine the tunnel disease risk level. An adaptive early warning strategy is triggered based on the tunnel disease risk level, and the adaptive early warning strategy is executed to generate a decision instruction set, which is then pushed to a visualization monitoring platform. In other words, data is collected through a multimodal sensor network, the collected multi-source heterogeneous data is fused to generate a dynamic perception feature map, identify potential disease evolution paths, trigger an adaptive early warning strategy based on the identified disease risk level, and generate a decision instruction set to promptly identify potential diseases and safety hazards, thereby improving the efficiency and safety of tunnel management.

[0014] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, which can be implemented in accordance with the contents of the description, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically listed below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in this application or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and a person of ordinary skill in the art can obtain other drawings based on the provided drawings without creative work.

[0016] Figure 1 This is a flowchart of the pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data in this application; Figure 2 This is a flow chart of determining the disease evolution pattern in the pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data in this application. DETAILED DESCRIPTION

[0017] This application solves the technical problem in the existing technology of relying on periodic detection and risk assessment of pipeline corridor diseases, which makes it difficult to detect potential problems in a timely manner, thereby affecting the efficiency and safety of pipeline corridor operations, by providing a pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data. Data is collected through a multimodal sensor network, and the collected multi-source heterogeneous data is fused to generate a dynamic perception feature map, identify potential disease evolution paths, trigger an adaptive early warning strategy based on the identified disease risk level, and generate a decision instruction set to timely identify potential diseases and safety hazards, thereby improving the efficiency and safety of pipeline corridor management.

[0018] Below, the technical solutions in this application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited to the example embodiments described herein. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. It should also be noted that, for the convenience of description, only the parts related to this application, rather than all of them, are shown in the accompanying drawings.

[0019] For examples, please see the attached Figure 1The present application provides a pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data, wherein the pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data specifically includes the following steps: S100: Data is collected through a multimodal sensor network deployed in underground pipeline corridors to construct a multi-source heterogeneous dataset.

[0020] Furthermore, the present application S100 includes: S110: The multimodal sensor network includes distributed fiber optic sensors, UWB positioning base stations and inspection robots; S120: Data is collected through the distributed fiber optic sensors to generate a first time series data stream; S130: Data positioning of the pipeline corridor is performed based on the UWB positioning base station to generate second spatiotemporal trajectory data; S140: The inspection robot controls the laser radar to scan along a preset path to obtain three-dimensional point cloud data of the pipeline corridor; S150: The first time series data stream, the second spatiotemporal trajectory data and the three-dimensional point cloud data of the pipeline corridor are standardized and preprocessed using edge computing nodes to obtain the multi-source heterogeneous data set.

[0021] Specifically, a multimodal sensor network is deployed within underground utility corridors to simultaneously collect different types of data and monitor the corridor's condition from multiple dimensions. Underground utility corridors are underground facilities specifically designed to house and protect various urban utility pipelines (such as electricity, communications, water, and gas), thereby improving the safety, maintainability, and protection of urban infrastructure. A multimodal sensor network, comprised of sensors of different types (modalities), collects data from the corridor across multiple dimensions using a variety of methods, including fiber optic sensors, positioning base stations, and robots, enabling simultaneous monitoring of different corridor parameters.

[0022] Distributed fiber optic sensors are sensors that can be distributed along optical fiber lines to monitor various environmental changes. They can sense physical quantities such as temperature, pressure, stress, and water leakage, and are widely used for long-distance monitoring. UWB positioning base stations are a broadband radio communication technology that can provide high-precision positioning services, used to obtain centimeter-level location information of personnel and equipment. Inspection robots are automated robots specifically designed for inspection and patrol tasks within pipeline corridors. They scan along preset paths and use sensors such as lidar to obtain three-dimensional point cloud data within the corridor. Robots can be equipped with various sensors, such as lidar, thermal imagers, and cameras, for data collection, fault detection, and environmental monitoring.

[0023] Data collected by distributed fiber optic sensors (such as temperature, stress, and water leakage signals) generates a first time-series data stream. This stream is continuously updated over time, is temporally ordered, and reflects changes in the tunnel environment at different points in time. UWB positioning base stations can acquire real-time location information of personnel and equipment within the tunnel, generating second-order spatiotemporal trajectory data. This data includes not only location coordinates but also timestamps, thus forming the movement trajectory of personnel or equipment within the tunnel. Inspection robots equipped with lidar and other sensors acquire 3D data within the tunnel during scanning, generating 3D point cloud data that accurately depicts the spatial position and shape of the pipelines, as well as any potential cracks, deformations, and other issues. The 3D point cloud data collected by the lidar can be used to construct a 3D model of the tunnel. This model details the spatial layout of the tunnel and the location of the equipment within it. To enhance the inspection robot's fault detection capabilities, in addition to the lidar, the robot can also be equipped with a thermal imager to detect temperature changes within the tunnel and obtain thermal anomaly image data. Thermal imagers can reveal abnormal heat sources on the surface of pipes or equipment areas. For example, pipes with abnormal temperatures may indicate overheating, blockage or leakage in the area.

[0024] All collected data (including time series data streams, spatiotemporal trajectory data, and three-dimensional point cloud data) will be standardized and pre-processed by edge computing nodes. Edge computing nodes are devices near the data generation source. They are responsible for preliminary processing of the data collected by sensors, such as denoising, data format unification, time alignment, etc., thereby converting data from different sensors into a format that can be processed uniformly.

[0025] Edge computing nodes convert time-series data streams, spatiotemporal trajectory data, and 3D point cloud data into a unified format and then perform spatiotemporal alignment. Because time-series data streams (such as temperature, stress, and water leakage data), spatiotemporal trajectory data (such as the location of equipment, personnel, or robots within the tunnel), and 3D point cloud data (such as spatial data acquired by LiDAR) come from different sources, their formats, units, and accuracy may differ. Edge computing nodes standardize this data and convert it into a unified time format.

[0026] Because time series data, trajectory data, and point cloud data all have temporal and spatial attributes, effective fusion and analysis cannot be performed if their temporal or spatial information is out of sync. Data from different sensors (such as temperature sensors, UWB positioning systems, and lidar) need to be converted to the same time scale. Assuming that time series data is recorded in seconds and trajectory data is recorded in milliseconds, spatiotemporal alignment ensures that all data can be compared and analyzed at the same point in time. Because different data sources may have different spatial coordinate systems, they need to be converted to a unified spatial coordinate system. For example, 3D point cloud data may use a local coordinate system, while trajectory data uses a global coordinate system. Through spatial alignment, all data can be integrated within the same spatial framework, ensuring that they accurately represent the structure and dynamic information of the pipeline corridor.

[0027] After standardization and spatiotemporal alignment, data from different sensors is successfully integrated into a multi-source, heterogeneous dataset consisting of temporally aligned and spatially unified time series data, trajectory data, and 3D point cloud data. This multimodal sensor network simultaneously collects different types of data (such as temperature, stress, and position) to comprehensively reflect the actual conditions within the tunnel, avoiding the blind spots of a single data source. Preprocessing through edge computing ensures data timeliness, enabling early detection of potential problems.

[0028] S200: The multi-source heterogeneous data sets are integrated based on a graph neural network to generate a dynamic perception feature map of the entire pipeline corridor.

[0029] Furthermore, the present application S200 includes: S210: Use the attention mechanism to dynamically assign weights to multi-source data, and extract spatial correlation features through convolution calculation; S220: Map sensor nodes to graph vertices according to the spatial correlation features, and map multi-source data weights to dynamic edge weights; S230: Obtain the topological structure of the pipeline corridor, and construct a graph neural network according to the pipeline corridor topology structure combined with the dynamic edge weights and the graph vertices; S240: Use the graph neural network to perform spatiotemporal feature fusion on the first time series data stream, the second spatiotemporal trajectory data and the pipeline corridor three-dimensional point cloud data to generate the dynamic perception feature map.

[0030] Specifically, the attention mechanism is used to process different types of sensor data (such as time series data, trajectory data, and 3D point cloud data), assigning different weights to different sensor nodes (such as temperature sensors, UWB positioning sensors, and lidar). The weights are dynamically adjusted based on the real-time importance of the sensor data, thereby giving more attention to critical data and neglecting less important data. The attention mechanism is a dynamic weighting method in deep learning that allows the model to focus on the importance of different parts of the input data when processing it. Through the attention mechanism, each input feature is automatically assigned a weight, ensuring that more attention is paid to the key parts.

[0031] Convolution is used to extract spatial correlation features from the sensor data in the tunnel. This operation identifies spatial relationships between sensor nodes (such as distance and interactions) and helps understand the spatial dependencies between different areas within the tunnel. The convolution operation involves sliding a small window (or convolution kernel) over the input data and performing a weighted summation during the sliding process. This effectively extracts local features and patterns in the input data, such as edges and textures in images. Convolution processes sensor data using a sliding window (typically a small matrix, called a convolution kernel). The convolution kernel slides over the sensor data matrix and calculates a weighted sum at each location.

[0032] Convolution operations can be used to extract spatial correlation features from the sensor data in the tunnel. These features, known as spatial dependencies between different sensors, indicate the degree of mutual influence between sensor data. For example, a temperature sensor may exhibit spatial correlation with surrounding humidity sensors, indicating that changes in humidity have a certain impact on temperature. Spatial correlation features refer to the spatial relationships or dependencies between different parts of the data, reflecting the location, data dependencies, and mutual influence between different sensors.

[0033] Sensor data in the tunnel is converted into a graph structure to facilitate processing using graph theory methods such as graph neural networks. A graph is constructed to represent the relationships between tunnel sensors, and the correlation between different sensor data is reflected by dynamically adjusting edge weights. Nodes containing various sensor types in the tunnel are mapped to graph vertices, with each sensor location being considered a vertex in the graph. Spatial correlation features (such as the relationship between temperature changes and stress changes) are used to assess the mutual influence between different sensors. The strength of the relationship between sensors is represented by edge weights, which reflect the dependency of the data between the two sensors. The dynamic edge weights mean that the strength of this relationship may change over time and with environmental changes.

[0034] In graph theory, an edge connects two vertices, and the edge weight represents the strength of the relationship between them. Dynamic edge weights mean that edge weights may change over time or under different environmental conditions. In the context of pipeline corridor monitoring, the size of edge weights may depend on the spatial correlation characteristics between different sensor nodes. By mapping sensor nodes as graph vertices and mapping dynamic data weights as edge weights, the graph structure required for a graph neural network is constructed. Edges represent the relationship between different sensor nodes, and edge weights indicate the strength of this relationship.

[0035] A graph neural network (GNN) is constructed by capturing the tunnel's topology (i.e., the spatial layout and connectivity between sensor nodes). The tunnel's topology encompasses the physical and data connectivity between sensors and devices within the tunnel. The topology determines how sensor nodes communicate and interact, representing the relative spatial positions of different sensors and the data exchange paths between them. The topology may include two primary pieces of information: the positional relationships of sensor nodes (e.g., sensor A is to the left of sensor B, sensor C is farther away from sensor A), and the direction of data flow (e.g., temperature data from node A is transmitted to node C via node B).

[0036] Graph neural networks perform training and inference based on topological structures, node characteristics, and dynamic edge weights, analyzing the relationships between different sensor data. In a graph neural network, each sensor node (vertex) is connected by edges, and the edge weights reflect the strength of the relationships between these nodes. In the context of utility corridor monitoring, graph neural networks can process sensor data through graph structures, identify relationships between different sensor nodes, and predict the status of utility corridors.

[0037] A graph neural network (GNN) processes and fuses multi-source spatiotemporal data collected from the tunnel to generate a dynamic perception feature map. The first time-series data stream, the second spatiotemporal trajectory data, and the tunnel's 3D point cloud data are fed into the GNN. The GNN integrates an LSTM network to model periodic (such as seasonal variations) and sudden (such as equipment failure) features in the first time-series data stream. The LSTM uses its gating mechanism to learn and memorize long-term dependent temporal information, thereby capturing dynamic changes in the tunnel environment. The GNN processes the spatiotemporal trajectory data and 3D point cloud data, capturing the spatial relationships and temporal changes between sensor nodes within a graph structure. After LSTM modeling and GNN processing, the resulting dynamic perception feature map reflects the status of various areas within the tunnel. For example, the map may indicate abnormal temperature, excessive humidity, or pressure fluctuations in a particular area, and provide a risk assessment for that area. The dynamic perception feature map contains real-time monitoring information for the tunnel, reflecting key indicators such as the health status and risk level of each area within the tunnel.

[0038] By leveraging the advantages of graph neural networks and attention mechanisms, we can effectively process multi-source heterogeneous data, thereby achieving comprehensive monitoring and accurate diagnosis of the health status of underground pipeline corridors, timely discovering potential disease risks, and taking corresponding early warning measures.

[0039] S300: Perform disease evolution pattern recognition on the dynamic perception feature map to determine the pipeline corridor disease risk level.

[0040] Further, as attached Figure 2 As shown, this application S300 includes: S310: traverse the dynamic perception feature map to perform deep residual analysis and obtain multiple residual data; S320: simulate the disease of the pipeline corridor based on the multiple residual data to obtain a simulated evolution path; S330: make degradation judgment according to the simulated evolution path, and determine the disease development stage according to the judgment result; S340: calculate the life of the pipeline corridor according to the disease development stage to obtain the remaining service life data of the pipeline corridor; S350: correlate and integrate the disease development stage and the remaining service life of the pipeline corridor to determine the disease evolution pattern.

[0041] Specifically, by traversing the dynamic perception feature map, the health status of each area of ​​the tunnel is analyzed and microscopic defect characteristics such as structural cracks, leakage, and corrosion are identified. The feature map generated by the graph neural network is compared with a standard or reference model, and residuals (i.e., the difference between the data) are calculated. Areas with large residuals may indicate problems such as cracks, leakage, or corrosion in the tunnel. These residuals are analyzed using deep learning to determine the microscopic morphological characteristics of the defects. Deep residual analysis is a technique that identifies anomalies or performance degradation by analyzing residuals in the data (i.e., the difference between the actual value and the predicted value). Residuals refer to the difference between the actual observed value and the model's predicted value, revealing potential faults or anomalies in various areas of the tunnel, such as the width of cracks or the extent of leakage.

[0042] Based on residual data, damage simulation is performed on the tunnel, simulating the progression of damage under different loads and environmental factors. Incorporating a load-environment coupling model, the evolution paths of various defects within the tunnel (such as cracks and corrosion) are simulated. Load factors may include equipment operation and pipeline pressure, while environmental factors may include temperature fluctuations and humidity. Load-environment coupling refers to the combined effects of loads (such as equipment operation and tunnel weight) and environmental factors (such as temperature and humidity) on the tunnel structure during actual use. Defect simulation must consider these factors to accurately predict the evolution path of defects. By incorporating a load-environment coupling model, defect simulation can predict the evolution of defects in the tunnel under different environmental conditions. For example, cracks in a tunnel can rapidly expand under high humidity and frequent vibration. The simulation results can provide predictions of defect progression and help formulate appropriate maintenance strategies.

[0043] Degradation assessment is performed based on the evolutionary path derived from the simulation to determine the tunnel's disease progression stage. Degradation assessment assesses the tunnel's degree of degradation by analyzing its health and the rate of disease spread. Degradation assessment involves analyzing the tunnel's health to determine the extent of degradation (aging, damage, corrosion, etc.), helping to assess the tunnel's service life and maintenance cycle. A tunnel may be in different stages of degradation, such as mild, severe, or critical. Each stage represents a different degree of disease impact and requires different maintenance or replacement strategies. For example, if cracks in a tunnel have reached a critical state, immediate reinforcement or repair may be necessary.

[0044] Based on the stage of damage development, the lifespan prediction system combines historical data, current damage conditions, and prediction models to estimate how long the corridor can continue to be used in its current condition. For example, if a corridor is at a certain stage of degradation, a lifespan calculation can predict the likelihood of a major failure within the next few years, alerting relevant departments to proactively address or repair the problem. Based on the different stages of damage development, the lifespan prediction can determine the most likely retirement date for the corridor. The remaining service life of a corridor refers to the length of time it can continue to be used in its current healthy state.

[0045] Correlating the stages of disease progression with the remaining service life allows for the identification of a pipeline corridor's disease evolution pattern. By integrating the disease progression and remaining service life at each stage, a comprehensive disease evolution pattern is formed, demonstrating the entire process of a pipeline corridor from healthy to damaged. For example, one section of a pipeline corridor may experience a short-term decline in strength due to corrosion, while another section may gradually lose stability due to crack propagation.

[0046] By combining the digital twin model with safety rheology theory, dynamic perception feature maps are used to identify the evolutionary patterns of defects. Digital twins are a technology that uses a digital mirror of a physical entity to reflect its physical state in real time. By collecting real-time data from the pipeline corridor (such as temperature, stress, deformation, leakage, etc.) and environmental data (such as humidity and temperature fluctuations) and synchronizing this data with the virtual model, the corridor's operational status and health changes can be tracked in real time. The corridor health data obtained through the digital twin model is combined with safety rheology theory to identify the evolutionary patterns of defects. Safety rheology theory focuses on the flow, deformation, and failure processes of materials under external forces. In the process of identifying defects in pipeline corridors, it is used to analyze how factors such as loads, temperature changes, and pressure fluctuations affect the corridor's safety and structural changes during long-term use. Using rheology theory, the stress and strain states of the pipeline corridor under different conditions are simulated to identify risk points that may lead to defects, such as crack propagation and corrosion areas.

[0047] Combining the digital twin model and safety rheology theory to identify the disease evolution pattern of dynamic perception feature maps can not only improve the recognition accuracy of the evolution of pipeline corridor diseases, but also enable effective preventive measures to be taken before pipeline corridor diseases occur, ensuring the safety and operational efficiency of the pipeline corridor.

[0048] Furthermore, the present application further comprises the following steps: S360: Conduct risk diffusion analysis on the pipeline corridor according to the disease development stage and the remaining service life of the pipeline corridor to obtain the risk diffusion coefficient; S370: Use the risk diffusion coefficient as an index to search the disease evolution pattern and set three risk levels; S380: Conduct disease functional impact analysis on the pipeline corridor, perform weighted adjustment on the three risk levels according to the functional impact coefficient, and dynamically divide the pipeline corridor disease risk level.

[0049] Specifically, through risk diffusion analysis and functional impact analysis of pipeline corridor diseases, the disease risk level of the pipeline corridor is dynamically assessed, and corresponding preventive measures are taken in a timely manner at the early stage of the disease, thereby reducing risks and optimizing pipeline corridor management. Based on the pipeline corridor's disease development stage and remaining service life, a risk diffusion analysis is conducted on the pipeline corridor to obtain a risk diffusion coefficient, which is used to assess the potential for disease propagation in the pipeline corridor and the impact of different diseases on the structural stability and operational safety of the pipeline corridor. For example, assuming that cracks in a certain area of ​​the pipeline corridor have entered a more serious degradation stage and the ambient humidity in this area is high, the risk of crack expansion is relatively large, and the risk diffusion coefficient will be high. The risk diffusion coefficient is usually calculated by analyzing the speed of disease expansion, the correlation between diseases, and the overall bearing capacity of the pipeline corridor structure.

[0050] Using the risk diffusion coefficient as an index, the disease evolution pattern is searched and three risk levels are assigned based on different assessment criteria for disease development patterns. The three risk levels are divided into Level I (low risk, score ≤ 30), Level II (medium risk, score 30 < ≤ 70), and Level III (high risk, score > 70). For example, if a crack develops slowly, spreads over a small area, and does not affect critical load-bearing components, it may be assessed as a Level I low risk. If the disease progresses rapidly and affects the core of the structure, the risk level will increase to Level II or Level III.

[0051] A functional impact analysis of defects is conducted on the corridor to assess the extent to which the defects affect its normal functions, such as the structural load-bearing capacity, traffic flow, and ventilation system. The functional impact coefficient is calculated to quantify the specific impact of the defect on the corridor's functions, and the corridor's risk level is weighted and adjusted accordingly. The functional impact coefficient is an indicator that measures the extent of the defect's impact on the corridor's functions. For example, if the defect affects the corridor's core load-bearing structure or causes blockage of important passages, the corresponding functional impact coefficient will be high, resulting in an increase in the risk level.

[0052] The corridor's disease risk level is dynamically adjusted based on the functional impact coefficient, and the corridor's risk level is updated in real time based on changes in disease status. For example, initially, the disease risk in a particular area may be low (Level I). However, as time passes, as the disease spreads and its impact on the corridor's function intensifies, the risk level will increase, potentially reaching Level II or Level III.

[0053] By combining functional impact analysis and risk diffusion coefficient, the actual risk of pipeline corridor diseases can be comprehensively assessed to ensure that the pipeline corridor can be effectively maintained in a timely manner. Based on real-time data, a sensitive response can be made to the health status of the pipeline corridor to prevent the spread of diseases or serious damage to the pipeline corridor structure, making the maintenance strategy of the pipeline corridor more accurate.

[0054] S400: triggering an adaptive early warning strategy according to the pipeline corridor disease risk level, executing the adaptive early warning strategy to generate a decision instruction set, and pushing the decision instruction set to a visual monitoring platform.

[0055] Furthermore, the present application S400 includes: S410: Set a multi-level risk threshold matrix according to the risk level of the pipeline corridor disease, and the multi-level risk threshold matrix includes a first-level threshold and a second-level threshold; S420: Traverse the dynamic perception feature map to extract the pipeline corridor disease index, and match the pipeline corridor disease index with the multi-level risk threshold matrix for judgment; S430: Determine whether the pipeline corridor disease index exceeds the first-level threshold. When the pipeline corridor disease index exceeds the first-level threshold, activate the inspection robot to perform local re-testing and verification on the pipeline corridor, and generate the re-testing and verification results; S440: Update the pipeline corridor disease index based on the re-testing and verification results, determine the pipeline corridor disease update index, and determine whether the pipeline corridor disease update index exceeds the second-level threshold. When the pipeline corridor disease index exceeds the second-level threshold, trigger the adaptive early warning strategy.

[0056] Specifically, a multi-level risk threshold matrix is ​​established to conduct accurate dynamic monitoring and risk warning of pipeline corridor defects, ensuring that potential risks can be identified in a timely manner and appropriate emergency measures can be taken when they occur, thereby improving the intelligence level and safety of pipeline corridor management. A multi-level risk threshold matrix is ​​set according to the risk level of pipeline corridor defects, including a first-level threshold and a second-level threshold. The first-level threshold corresponds to an early warning of local micro-damage, mainly targeting small-scale defects inside the pipeline corridor, such as cracks, minor leakage or local corrosion. Although these defects do not immediately threaten the overall structure of the pipeline corridor, if they are not repaired in time, they will gradually develop into more serious problems over time. The second-level threshold corresponds to the precursors of structural instability, focusing on major damage problems in the core components or key areas of the pipeline corridor structure, such as crack expansion and increased corrosion. These defects will directly affect the stability of the pipeline corridor and may even lead to the risk of structural instability.

[0057] By traversing the dynamic perception feature map, pipeline corridor disease indicators are extracted and matched against a multi-level risk threshold matrix for determination. The dynamic perception feature map is generated using multimodal sensor data (such as temperature, pressure, stress, and leakage), providing comprehensive data on the current health of the pipeline corridor. Key pipeline corridor disease indicators are extracted from the dynamic perception feature map and compared against the set risk thresholds to determine whether the disease exceeds the set risk value. For example, if the crack width or stress in a certain area exceeds the first-level threshold, the area is judged to be in an early stage of disease.

[0058] The system determines whether the corridor's damage indicators exceed the first-level threshold. If so, a patrol robot is automatically activated to retest and verify a portion of the corridor. Equipped with various sensors (such as lidar and infrared thermal imagers), the robot scans the corridor along a pre-set path, acquiring more accurate damage data and generating retest and verification results. For example, if cracks exceeding the first-level threshold are detected in a specific area, the robot will re-measure indicators such as crack width and depth to confirm the crack's actual condition.

[0059] Based on the inspection robot's retest and verification results, the corridor's damage indicators are updated to determine the updated damage index. This updated damage index provides more accurate damage status data. If the corridor's damage update index exceeds the secondary threshold, the damage in that area is considered to have reached a critical stage and may pose a threat to the corridor's structural safety. At this point, the adaptive early warning strategy is triggered, automatically activating the emergency power supply to ensure the normal operation of the corridor's basic functions and cordoning off the affected area to prevent the disease from spreading or affecting other areas. The activation of the adaptive early warning strategy means taking a series of automated measures to control the risks posed by the damage and, based on this, initiate more complex repair or emergency work.

[0060] By triggering an adaptive early warning strategy based on the risk level of pipeline corridor diseases, the risk management capability of the pipeline corridor is improved. Through dynamic multi-level risk threshold analysis and adaptive early warning mechanism, more intelligent and accurate pipeline corridor maintenance and repair are achieved, minimizing the occurrence of pipeline corridor failures and safety accidents, and improving the operational safety and efficiency of the pipeline corridor.

[0061] Furthermore, the present application further comprises the following steps: S450: Execute the adaptive early warning strategy, call the disease development stage of the disease evolution model, match the disease development stage with the three-dimensional point cloud data of the pipeline corridor, and generate a stress concentration area vector diagram; S460: Introduce a historical repair case library, search the historical repair case library for similar diseases, and obtain disease construction record data; S470: Map the disease construction record data to the stress concentration area vector diagram for planning, determine multiple groups of emergency arrival paths, and formulate the decision instruction set based on the multiple groups of emergency arrival paths.

[0062] Specifically, based on the triggered early warning strategy, an adaptive early warning strategy is executed, and the disease development stage in the disease evolution model is called up. The disease development stage is matched with the 3D point cloud data of the tunnel. In other words, the specific location and development level of the disease are compared and analyzed with the 3D model of the tunnel to determine which areas have stress concentrations and generate a vector map of the stress concentration areas. The vector map of the stress concentration areas can intuitively display the 3D location of the damaged areas in the tunnel. Combined with the results of structural mechanics simulation, it can provide a deeper analysis of the mechanical state of the diseased areas. For example, if a crack area in a tunnel section has high stress concentration, this information can be converted into a 3D image, marking the specific location and risk level of the area.

[0063] A historical restoration case database was introduced and searched for similar defects. This database contains various past tunnel defect cases and corresponding restoration construction records, providing a wealth of reference data. By searching for similar defects, restoration construction records were obtained, including information such as past restoration methods, materials used, and restoration results. This provides a valuable reference for current tunnel defect restoration, ensuring the feasibility and efficiency of restoration plans.

[0064] Mapping construction records of defects with vector diagrams of stress concentration areas allows for planning and prioritizes the execution of key steps in the tunnel corridor repair process. For example, the shortest and fastest emergency route is planned based on the location and severity of the stress concentration area to ensure timely repair work. Multiple emergency arrival routes are generated based on the location of the stress concentration area, providing the repair team with multiple routes of action. Each route is designed based on the repair team's arrival speed, tunnel accessibility, and the severity of the defect.

[0065] Based on multiple emergency arrival paths, decision-making instruction sets are developed, including damage location information and repair recommendations. These provide repair personnel or emergency management personnel with clear repair plans and operational guidelines. For example, the instruction set might instruct repair personnel to first reinforce the area with the greatest stress concentration, and then address other localized damage. These instructions ensure that repair work proceeds efficiently and orderly, minimizing the risk of damage spreading.

[0066] By formulating an instruction set, the intelligence level of pipeline corridor defect management has been significantly improved. Through precise data matching and analysis, it can help repair personnel arrive at and deal with defects efficiently, reducing the risk of pipeline corridor failures and safety accidents.

[0067] Furthermore, the present application further comprises the following steps: The decision instruction set is pushed to the visual monitoring platform, and the three-dimensional point cloud data of the pipeline corridor is dynamically annotated based on the decision instruction set to obtain multiple disease location labels; damage repair analysis is performed according to the multiple disease location labels, and a repair priority heat map is constructed; and disease superimposed navigation of the pipeline corridor is performed according to the repair priority heat map through the visual monitoring platform.

[0068] Specifically, a decision instruction set containing the location information of each defect in the tunnel and corresponding repair recommendations is pushed to the visualization monitoring platform. Based on the decision instruction set, the tunnel's 3D point cloud data is dynamically annotated, and the specific locations of the defects are highlighted in the 3D model through visualization. Multiple tags are generated for each defect location. These tags include information such as defect type, severity, and repair priority, indicating the areas in the tunnel requiring repair.

[0069] Based on multiple defect location tags, a defect repair analysis is performed to assess the repair priority of each defect in the tunnel, taking into account the type of defect, its impact on tunnel safety, and the resources and time required for repair. After the defect repair analysis, a repair priority is assigned to each defect, and a repair priority heat map is constructed based on the repair priority of the defect. A heat map is a visualization based on spatial distribution, where the depth of the color represents the repair priority. For example, red areas represent serious defects that need to be repaired urgently, orange indicates a medium repair priority, and yellow or green indicate less serious defects. This clearly reflects the spatial distribution of different defects in the tunnel and their repair priority, helping to identify which areas require priority treatment.

[0070] Through the visual monitoring platform, tunnels are navigated by overlaying defects based on repair priority heat maps. Multiple defect information, repair suggestions, and work paths are superimposed on a single interface, creating an intuitive operational map. This map allows tunnel operators or repair teams to quickly determine work priorities and repair paths. This navigation method allows repair teams to precisely start with the highest-priority defect areas and address them sequentially, avoiding missing critical areas and improving repair efficiency. For example, when a heat map indicates multiple defect areas in a tunnel section, the repair team is guided to begin with the most urgent area, reducing unnecessary time.

[0071] Through dynamic marking and defect labeling, the specific location of each defect in the corridor can be accurately located. Using heat maps and priority analysis, it helps to quickly determine the urgency of repairs and optimize the repair sequence. This improves the visualization and intelligence level of corridor defect management, allowing repair work to be carried out in an orderly manner, significantly improving the safety of corridor operations and maintenance efficiency.

[0072] In summary, the pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data provided by this application has the following beneficial effects: Data is collected through a multimodal sensor network deployed in the underground tunnel to construct a multi-source heterogeneous data set. This data set is then fused based on a graph neural network to generate a dynamic perception feature map for the entire tunnel. Disease evolution patterns are identified on this dynamic perception feature map to determine the tunnel disease risk level. An adaptive early warning strategy is triggered based on the tunnel disease risk level, and the adaptive early warning strategy is executed to generate a decision instruction set, which is then pushed to a visualization monitoring platform. In other words, data is collected through a multimodal sensor network, the collected multi-source heterogeneous data is fused to generate a dynamic perception feature map, identify potential disease evolution paths, trigger an adaptive early warning strategy based on the identified disease risk level, and generate a decision instruction set to promptly identify potential diseases and safety hazards, thereby improving the efficiency and safety of tunnel management.

[0073] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

[0074] Obviously, for those skilled in the art, several improvements and modifications can be made to the present application without departing from the principles of the present application, and these improvements and modifications also fall within the scope of protection of the present application.

Claims

1. A pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data is characterized by: include: Data is collected through a multimodal sensor network deployed in underground pipe corridors to construct a multi-source heterogeneous dataset; The multi-source heterogeneous data sets are fused based on a graph neural network to generate a dynamic perception feature map of the entire pipeline corridor; Performing disease evolution pattern recognition on the dynamic perception feature map to determine the pipeline corridor disease risk level; An adaptive early warning strategy is triggered according to the pipeline corridor disease risk level, the adaptive early warning strategy is executed to generate a decision instruction set, and the decision instruction set is pushed to the visual monitoring platform.

2. The pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data according to claim 1 is characterized in that: The multimodal sensor network includes distributed optical fiber sensors, UWB positioning base stations, and inspection robots. The method for constructing a multi-source heterogeneous data set includes: Collecting data through the distributed optical fiber sensor to generate a first time series data stream; Performing data positioning on the pipeline corridor based on the UWB positioning base station to generate second spatiotemporal trajectory data; Controlling the inspection robot of the laser radar to scan along a preset path to obtain three-dimensional point cloud data of the pipeline corridor; An edge computing node is used to perform standardized preprocessing on the first time series data stream, the second spatiotemporal trajectory data, and the pipeline corridor three-dimensional point cloud data to obtain the multi-source heterogeneous data set.

3. The pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data according to claim 2 is characterized in that: The multi-source heterogeneous datasets are fused based on a graph neural network to generate a dynamic perception feature map of the entire corridor, including: The attention mechanism is used to dynamically assign weights to multi-source data, and spatial correlation features are extracted through convolution calculations. Mapping sensor nodes to graph vertices according to the spatial correlation features and mapping multi-source data weights to dynamic edge weights; Acquire a pipe gallery topology structure, and construct a graph neural network according to the pipe gallery topology structure, the dynamic edge weights, and the graph vertices; The first time series data stream, the second spatiotemporal trajectory data and the three-dimensional point cloud data of the pipeline corridor are fused with spatiotemporal features through the graph neural network to generate the dynamic perception feature map.

4. The pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data according to claim 2 is characterized in that: Performing disease evolution pattern recognition on the dynamic perception feature map includes: Traversing the dynamic perception feature map to perform deep residual analysis to obtain multiple residual data; Performing disease simulation on the pipeline corridor based on the multiple residual data to obtain a simulation evolution path; Degradation is determined according to the simulated evolution path, and the disease development stage is determined based on the determination result; Calculate the life of the pipeline corridor according to the disease development stage to obtain the remaining service life data of the pipeline corridor; The disease development stage and the remaining service life of the pipeline corridor are correlated and integrated to determine the disease evolution pattern.

5. The pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data according to claim 4 is characterized in that: The process of determining the pipeline corridor disease risk level includes: Conduct risk diffusion analysis on the pipeline corridor according to the disease development stage and the remaining service life of the pipeline corridor to obtain a risk diffusion coefficient; Using the risk diffusion coefficient as an index, the disease evolution pattern is searched and three risk levels are set; An analysis of the functional impact of defects on the pipeline corridor is conducted, and the three risk levels are weighted and adjusted according to the functional impact coefficient, so as to dynamically divide the pipeline corridor defect risk levels.

6. The pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data according to claim 1 is characterized in that: The adaptive early warning strategy is triggered according to the pipeline corridor disease risk level, including: A multi-level risk threshold matrix is ​​set according to the pipeline corridor disease risk level, wherein the multi-level risk threshold matrix includes a first-level threshold and a second-level threshold; Traverse the dynamic perception feature map to extract pipeline corridor disease indicators, and match the pipeline corridor disease indicators with the multi-level risk threshold matrix for judgment; Determine whether the corridor disease index exceeds the first-level threshold. When the corridor disease index exceeds the first-level threshold, activate the inspection robot to conduct local retesting and verification of the corridor and generate the retesting and verification results; Based on the retest and verification results, the corridor disease index is updated, the corridor disease update index is determined, and it is judged whether the corridor disease update index exceeds the secondary threshold. When the corridor disease index exceeds the secondary threshold, the adaptive early warning strategy is triggered.

7. The pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data according to claim 4 is characterized in that: The method of triggering an adaptive early warning strategy according to the pipeline corridor disease risk level and executing the adaptive early warning strategy to generate a decision instruction set includes: Executing the adaptive early warning strategy, calling the disease development stage of the disease evolution model, matching the disease development stage with the three-dimensional point cloud data of the pipeline corridor, and generating a vector diagram of stress concentration areas; Introducing a historical restoration case database, searching for similar defects in the historical restoration case database to obtain defect construction record data; The defective construction record data is mapped to the stress concentration area vector diagram for planning, multiple groups of emergency arrival paths are determined, and the decision instruction set is formulated according to the multiple groups of emergency arrival paths.

8. The pipeline corridor disease monitoring and diagnosis method based on multi-source heterogeneous data according to claim 7 is characterized in that: After the decision instruction set is pushed to the visual monitoring platform, it includes: Pushing the decision instruction set to the visual monitoring platform, dynamically annotating the three-dimensional point cloud data of the pipeline corridor based on the decision instruction set to obtain multiple disease location labels; Performing disease repair analysis according to the multiple disease location labels and constructing a repair priority heat map; The visual monitoring platform is used to conduct disease overlay navigation on the pipeline corridor according to the repair priority heat map.

Images