Underground pipeline potential safety hazard assessment system based on geographic information and control method thereof

Abstract

The invention relates to the technical field of underground pipeline safety assessment, and discloses an underground pipeline potential safety hazard assessment system based on geographic information, which comprises a data acquisition module, a preprocessing and management module, a risk assessment module, an early warning and decision module and a display module. Multi-source data are accurately collected based on an underground pipeline BIM model, and a rule engine and machine learning algorithm is used for cleaning and integration, so that abnormal data are eliminated, a reliable database is constructed, accurate data support is provided for a multi-factor coupling risk assessment model, and scientific dynamic assessment of underground pipeline potential safety hazards is realized; meanwhile, an isolated forest algorithm and a reinforcement learning algorithm are adopted for anomaly detection and strategy generation, early warning is triggered according to the risk level, decision suggestions are pushed in a linkage mode, and the early warning accuracy and response efficiency are effectively improved; in addition, a geographic information system is utilized to construct an exclusive three-dimensional visual model and support interaction, so that the pipeline condition can be visually checked.

Description

Technical Field

[0001] This invention relates to the field of underground pipeline safety assessment technology, and more specifically discloses an underground pipeline safety hazard assessment system and control method based on geographic information. Background Technology

[0002] Underground pipelines are a general term for all kinds of municipal utility pipelines and their ancillary facilities built below the ground in cities. They include pipelines and cables for water supply, drainage, gas, heating, electricity, communications, radio and television, and industry, and are important infrastructure to ensure the operation of cities.

[0003] The prior art patent document with authorization announcement number CN116341295B discloses "A Road Stability Assessment Method and System Based on Underground Pipelines", which includes the following steps: obtaining underground pipeline information of the road to be analyzed; building an underground pipeline influence intensity model and an underground pipeline deterioration influence intensity model, and obtaining a first influence intensity value and a second influence intensity value by combining the above models with the underground pipeline information; obtaining the standard bearing capacity of the road and the standard bearing capacity of the underground pipeline of the road to be analyzed; obtaining a first influence degree by combining the first influence intensity value with the standard bearing capacity of the road; obtaining a second influence degree by combining the second influence intensity value with the standard bearing capacity of the underground pipeline; summarizing the first influence degree and the second influence degree to assess the degree of influence of the underground pipeline on road stability.

[0004] The patent document with authorization announcement number CN117610137A discloses "a BIM-based method and system for early warning of underground pipelines", which includes: obtaining the operation data of the target underground pipeline based on the underground pipeline BIM model; determining whether to issue an early warning based on the operation status judgment result; when it is determined to issue an early warning, setting an initial early warning coefficient in advance, determining the final early warning coefficient based on environmental data, determining the final early warning level based on the final early warning coefficient, and issuing an early warning.

[0005] While existing technologies can quantitatively assess the impact of underground pipelines on road stability by building models and can also achieve dynamic early warning of underground pipelines based on BIM models, providing some technical support for underground pipeline safety management, there are still problems such as a single assessment dimension and a lack of real-time dynamic early warning mechanism, making it difficult to capture risks in real time; insufficient coupling with environmental factors, failure to link with engineering standards such as road bearing capacity, and a disconnect between visualization and decision-making, making it impossible to form a complete closed loop for safety hazard assessment and decision-making. Summary of the Invention

[0006] The present invention mainly provides a geographic information-based underground pipeline safety hazard assessment system and control method, which can solve the problems mentioned in the background art.

[0007] To address the aforementioned technical problems, this invention provides the following technical solution, more specifically, a method for controlling safety hazards in underground pipelines based on geographic information, comprising:

[0008] S1. Based on the underground pipeline BIM model, the operation data of the target underground pipeline is collected through 3D laser scanning, ground penetrating radar and IoT sensors, and the environmental data interface is connected to obtain terrain, geological and surrounding environmental data to achieve accurate collection of multi-source data.

[0009] S2. Using the "rule engine + machine learning" algorithm, the collected multi-source data is cleaned, denoised, format converted and coordinate unified, and outliers are removed. The data is then integrated to build an underground pipeline geographic information database, and storage and query are optimized through spatial indexing technology.

[0010] S3. Based on the analytic hierarchy process and Bayesian network, a multi-factor coupled risk assessment model is constructed. The model comprehensively considers factors such as pipeline material, service life, corrosion degree, surrounding construction activities and geological conditions to conduct regular or real-time dynamic risk assessments and determine the risk level.

[0011] S4. Anomaly detection is performed using the isolated forest algorithm and the support vector data description algorithm. Reinforcement learning algorithm is combined to generate maintenance or emergency strategies. Early warning is triggered based on the risk level and decision-making suggestions are pushed in conjunction with the algorithm.

[0012] S5. Utilize the 3D modeling function of the geographic information system to construct a dedicated 3D visualization model of underground pipelines and their surrounding environment, which can also be interactive.

[0013] Furthermore, in S1, the three-dimensional laser scanning parameters are adjusted according to the distribution density and burial depth characteristics of underground pipelines, the ground-penetrating radar signal processing is adapted to the electromagnetic differences between pipeline materials and geological media, the Internet of Things sensors are deployed according to the pipeline operation monitoring needs, and the environmental data interface is based on the standard of the urban geographic information public platform.

[0014] Furthermore, in S2, the rule engine constructs data verification rules based on the spatial continuity and attribute logic of underground pipelines, machine learning trains anomaly recognition models through historical pipeline data, and spatial indexing technology adopts an optimization algorithm that combines R-trees with pipeline network topology.

[0015] Furthermore, in S3, the judgment matrix constructed by the analytic hierarchy process integrates historical accident data of underground pipelines with expert operation and maintenance experience. The Bayesian network training data covers pipeline operation, fault cases and surrounding environmental change information. The periodic evaluation cycle is coordinated with the pipeline inspection plan, and the real-time evaluation triggers the update of IoT sensor data.

[0016] Furthermore, in S4, the anomaly detection thresholds of the Isolation Forest algorithm and the Support Vector Data Description algorithm are determined through historical statistics of the normal operating parameters of underground pipelines; the decision space constructed by the reinforcement learning algorithm includes risk level, maintenance cost, and environmental constraints, and the strategy generation is linked to the maintenance resource scheduling system.

[0017] Furthermore, in S5, the 3D visualization model rendering highlights the differentiated display of pipeline types and risk areas, and the interactive function supports pipeline layer-by-layer viewing, spatial relationship querying, and risk data linkage retrieval.

[0018] According to another aspect of the present invention, a geographic information-based underground pipeline safety hazard assessment system is provided. This system is implemented based on the above-mentioned geographic information-based underground pipeline safety hazard control method, specifically including:

[0019] The data acquisition module, based on the underground pipeline BIM model, collects operational data of the target underground pipeline through 3D laser scanning, ground-penetrating radar, and IoT sensors. It also connects to an environmental data interface to obtain surrounding environmental data, achieving accurate multi-source data collection. This ensures the comprehensiveness and accuracy of the data, providing a reliable foundation for subsequent assessments. The preprocessing and management module utilizes a "rule engine + machine learning" algorithm to clean, denoise, convert formats, and unify coordinates of the multi-source data. Outliers are removed, and the data is integrated to construct a geographic information database. Spatial indexing technology optimizes storage and retrieval, improving data quality and management efficiency, and laying a solid data foundation for system operation. The risk assessment module is based on the analytic hierarchy process (AHP) and Bayesian networks. The multi-factor coupling model comprehensively considers factors such as pipeline material and service life for regular or real-time dynamic assessment to determine risk levels. This enables scientific and accurate evaluation of pipeline safety hazards, providing a basis for preventive measures. The early warning and decision-making module uses the isolated forest algorithm and support vector data description algorithm for anomaly detection, combined with reinforcement learning algorithm to generate maintenance or emergency strategies. It triggers early warnings based on risk levels and pushes decision suggestions in a coordinated manner, enabling timely detection of anomalies and providing scientific decision-making, effectively reducing the probability of safety accidents. The display module utilizes the 3D modeling function of the geographic information system to construct a dedicated 3D visualization model of underground pipelines and their surrounding environment, supporting interaction. This facilitates intuitive viewing of pipeline conditions, improving management efficiency and decision-making accuracy.

[0020] Furthermore, the preprocessing and management module includes: an intelligent cleaning module and a management module;

[0021] Intelligent cleaning module: Utilizes a rule engine to construct data verification rules based on the spatial continuity and attribute logic of underground pipelines. Combined with machine learning algorithms, it trains anomaly recognition models on historical pipeline data to achieve cleaning, noise reduction, format conversion, and coordinate unification of multi-source data, eliminate outliers, ensure data quality, and provide a reliable data foundation for subsequent assessments.

[0022] Management module: The cleaned data is integrated to build an underground pipeline geographic information database. An optimized algorithm combining R-tree and pipeline network topology is used for spatial indexing to achieve efficient data storage and fast query, thereby improving data management efficiency.

[0023] Furthermore, the risk assessment module includes: a dynamic assessment module and a factor construction module;

[0024] Dynamic assessment module: Based on a multi-factor coupled model constructed using the analytic hierarchy process and Bayesian networks, combined with real-time data updates from IoT sensors to trigger real-time assessment, comprehensively determine the pipeline risk level, and achieve dynamic monitoring of the pipeline safety status;

[0025] Factor construction module: By integrating historical accident data of underground pipelines with expert operation and maintenance experience through the analytic hierarchy process, a judgment matrix is ​​constructed to determine the factor weights of pipeline material and service life. At the same time, a Bayesian network is trained using data covering pipeline operation, failure cases and changes in the surrounding environment to build the foundation for a multi-factor coupled risk assessment model.

[0026] Furthermore, the early warning and decision-making module includes: an intelligent early warning module and a strategy generation module;

[0027] Intelligent early warning module: It adopts the isolated forest algorithm and support vector data description algorithm to determine the anomaly detection threshold by the historical statistics of the normal operation parameters of underground pipelines, and monitors the assessed risk data in real time. When the risk exceeds the threshold, an early warning is triggered to improve the accuracy and reliability of the early warning.

[0028] Strategy generation module: Utilizes reinforcement learning algorithms to construct a decision space that includes risk level, maintenance cost, and environmental constraints. Combined with the maintenance resource scheduling system, it generates maintenance or emergency strategies and pushes decision suggestions in a coordinated manner to achieve intelligent response and handling of risks.

[0029] The beneficial effects of this invention, a geographic information-based underground pipeline safety hazard assessment system and its control method, are as follows: By accurately collecting multi-source data based on underground pipeline BIM models and cleaning and integrating it using a "rule engine + machine learning" algorithm, abnormal data is eliminated to construct a reliable database, thus providing accurate data support for multi-factor coupled risk assessment models and enabling scientific and dynamic assessment of underground pipeline safety hazards. Simultaneously, the isolated forest algorithm and reinforcement learning algorithm are used for anomaly detection and strategy generation, triggering early warnings based on risk levels and pushing decision-making suggestions, effectively improving the accuracy of early warnings and response efficiency. Furthermore, a dedicated 3D visualization model is constructed using a geographic information system and supports interaction, facilitating intuitive viewing of pipeline conditions. Ultimately, a complete closed loop from data collection to decision display is formed, achieving accurate assessment and intelligent control of underground pipeline safety hazards, and providing a scientific basis for urban underground pipeline safety management. Attached Figure Description

[0030] The present invention will now be described in further detail with reference to the accompanying drawings and specific implementation methods.

[0031] Figure 1 This is a schematic diagram of the system principle;

[0032] Figure 2 This is a flowchart illustrating the method. Detailed Implementation

[0033] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0034] According to one aspect of the invention, such as Figure 1-2 As shown, a geographic information-based underground pipeline safety hazard assessment system and its control method are provided, including:

[0035] S1, Multi-source data acquisition

[0036] First, based on the underground pipeline BIM model, the target underground pipeline is scanned using a 3D laser scanning device. The scanning parameters are adjusted according to the distribution density and burial depth characteristics of the underground pipeline (e.g., increasing the scanning resolution in high-density areas and adjusting the laser emission angle for deeply buried pipelines) to obtain the spatial location and geometric shape data of the pipeline. Simultaneously, ground-penetrating radar is used to detect the pipeline material and surrounding geological media. Its signal processing flow is adapted to the electromagnetic differences between pipeline materials (e.g., metal, plastic) and geological media (e.g., soil, rock) to ensure the accuracy of information such as pipeline burial depth and direction. In addition, IoT sensors (e.g., pressure sensors, temperature sensors) are deployed at key nodes of the pipeline (e.g., elbows, valves). The deployment principle is based on the pipeline operation monitoring needs (e.g., denser deployment in high-risk sections) to collect operational data such as flow rate, pressure, and temperature in real time. Finally, the system connects to the city's geographic information public platform through an environmental data interface to obtain surrounding environmental data such as topography and geological structure according to standard protocols, achieving comprehensive collection of multi-source data.

[0037] S2, Preprocessing and Management

[0038] The algorithm of "rule engine + machine learning" is used to clean, denoise, convert the format and unify the coordinates of the collected multi-source data, and remove outliers. The data is then integrated to build an underground pipeline geographic information database. The storage and query are optimized by spatial indexing technology. Specifically, the rule engine builds data verification rules based on the spatial continuity and attribute logic of underground pipelines, the machine learning trains an anomaly recognition model through historical pipeline data, and the spatial indexing technology adopts an optimization algorithm that combines R-tree and pipeline network topology.

[0039] The cleaning of multi-source data involves the rule engine first marking preliminary abnormal data with spatial discreteness and contradictory attributes based on the spatial continuity of underground pipelines (the error between adjacent pipeline nodes is ≤10cm) and the logicality of attributes (the matching relationship between pipe diameter, material and operating parameters). The machine learning calls the trained random forest model and inputs feature parameters such as pressure fluctuation amplitude and temperature gradient change rate to perform a second screening of the preliminary abnormal data. Through the dual filtering of "rule verification + model recognition", the accurate cleaning of multi-source data is achieved.

[0040] The denoising and format conversion are implemented as follows: the 3D laser scanning point cloud data adopts the K-neighbor statistical filtering algorithm (calculates the average distance of the 20 nearest neighbors of each point and removes outliers that deviate from the mean by 3 times the standard deviation). The ground penetrating radar signal is decomposed into a high-frequency noise layer and a low-frequency effective layer through wavelet transform. The low-frequency band is retained for reconstruction and denoising. For format conversion, a multi-source data field mapping table is established to uniformly parse laser scanning (.las), ground penetrating radar (.dzt), and IoT sensor (.json) data into GeoJSON format, standardizing core fields such as "spatial coordinates, material type, and operating parameters" to achieve compatibility of heterogeneous data formats.

[0041] Furthermore, the details of coordinate unification and database construction are as follows: Coordinate unification is based on a seven-parameter transformation model (using regional control points to solve transformation parameters), converting local coordinate system and engineering coordinate system data into the WGS84 coordinate system to ensure that the coordinate error of pipeline nodes is ≤ ±5cm; When constructing the database, R-tree indexes pipeline spatial nodes, and combines the "main line-branch line" topology relationship for hierarchical storage, while recording data confidence levels to provide a basis for subsequent weighted analysis of risk assessment, further optimizing data storage and query efficiency;

[0042] This dual data cleaning mechanism of "rule engine + machine learning" effectively removes abnormal data such as "brushing" and spatially discrete points, improving data accuracy and providing a reliable data foundation for subsequent risk assessment. The unified format of multi-source data and accurate coordinate conversion eliminate compatibility barriers between heterogeneous data, enabling seamless integration of data from laser scanning, radar detection, etc. The indexing technology based on R-tree and pipeline topology shortens the response time for large-scale data queries to the millisecond level. At the same time, the hierarchical recording of data confidence provides a quantitative basis for the weighted analysis of risk assessment, ultimately forming a closed loop of "high-quality data - efficient management - accurate assessment," significantly improving the efficiency and reliability of underground pipeline safety hazard assessment.

[0043] S3, Risk Assessment

[0044] A multi-factor coupled risk assessment model is constructed based on the analytic hierarchy process and Bayesian network. It integrates factors such as pipeline material, service life, corrosion degree, surrounding construction activities and geological conditions to conduct regular or real-time dynamic risk assessments and determine the risk level.

[0045] The judgment matrix constructed by the analytic hierarchy process integrates historical accident data of underground pipelines with expert operation and maintenance experience. The Bayesian network training data covers pipeline operation, fault cases and surrounding environmental changes. The regular evaluation cycle is coordinated with the pipeline inspection plan, and the real-time evaluation triggers rely on IoT sensor data updates.

[0046] Specifically, when executing the Analytic Hierarchy Process (AHP), first extract underground pipeline accident data from the past 5 years (e.g., in leakage accidents, material degradation accounted for 35% and surrounding construction disturbance accounted for 28%). Combined with industry experts' weighted scores for five factors—material, age, corrosion, construction, and geology—using a 1-9 scale, a judgment matrix is ​​constructed. Simultaneously, the maximum eigenvalue λmax of the matrix is ​​calculated using the eigenvalue method, and the weight vector W is derived. A consistency check is then introduced to calculate the consistency index.

[0047]

[0048] Then compare the average random consistency index RI (which is 1.12). If the consistency ratio is:

[0049]

[0050] The matrix then passes the consistency check, ensuring that the weight allocation is reasonable;

[0051] In the construction of the Bayesian network, "pipeline material, service life, corrosion degree, surrounding construction activities, and geological conditions" are first defined as the parent nodes of the network, and "risk level" is defined as the child nodes. A directed acyclic graph (DAG) is established to describe the causal relationship (such as "surrounding construction activities" → "increased pipeline corrosion degree"). Then, using pipeline operation data (such as pressure and flow time series data), failure cases (including leakage, rupture, etc.) and surrounding construction activity data (such as excavation distance and frequency), the maximum likelihood estimation method is used to train the conditional probability table (CPT).

[0052] By constructing a multi-factor coupled risk assessment model using the Analytic Hierarchy Process (AHP) and Bayesian networks, a combination of expert experience and data-driven approaches is achieved. This model can quantify the weights of each factor using historical accident data and expert knowledge, and dynamically update the risk assessment results using real-time monitoring data. This significantly improves the scientific rigor and accuracy of underground pipeline safety hazard assessment, providing a more reliable basis for risk warning and maintenance decisions.

[0053] S4. Early Warning and Decision-Making

[0054] Anomaly detection is performed using the Isolation Forest algorithm and the Support Vector Data Description algorithm. Reinforcement learning algorithm is combined to generate maintenance or emergency strategies. Early warnings are triggered based on the risk level and decision-making suggestions are pushed in conjunction with them.

[0055] The anomaly detection thresholds for the isolated forest algorithm and the support vector data description algorithm are determined through historical statistics of normal operating parameters of underground pipelines; the decision space constructed by the reinforcement learning algorithm includes risk level, maintenance cost, and environmental constraints, and the strategy generation is linked to the maintenance resource scheduling system.

[0056] Specifically, the anomaly detection thresholds for the Isolation Forest algorithm and the Support Vector Data Description algorithm are determined through historical statistics of normal operating parameters of underground pipelines, as shown below:

[0057] The Isolation Forest algorithm uses parameters such as pressure, flow rate, and temperature of underground pipelines during normal operation as a training set to construct an "isolated forest" composed of multiple binary trees. For each sample, the average path length in the tree is calculated; the shorter the path, the higher the probability of an anomaly. The anomaly detection threshold is determined by the interquartile range (IQR) of historical data. When the anomaly score of a sample meets the following conditions:

[0058] s(x) > Q3 + 1.5 × IQR

[0059] Then it is determined to be abnormal, where s(x) is:

[0060]

[0061] Where h(x) is the sample path length, and c(n) is the average path length correction value of the tree;

[0062] Support Vector Data Description (SVDD) uses radial basis function (RBF) kernels to map normal operation parameters to a high-dimensional space, constructing a minimum hypersphere to enclose the normal data. The optimization objective of the hypersphere is:

[0063]

[0064] Where R is the radius of the hypersphere, ξ i Let ω be the center coordinate of the hypersphere and the slack variable. The constraint condition is:

[0065]

[0066] Where x i For the original sample of the i-th normal operating parameter, when the test data exceeds the hypersphere boundary, it is judged as abnormal. The threshold is determined by the hypersphere radius R of the historical normal data.

[0067] The decision space constructed by the reinforcement learning algorithm includes risk level, maintenance cost, and environmental constraints. The strategy generation is linked to the maintenance resource scheduling system, as shown below:

[0068] Decision space construction:

[0069] State S includes the current risk level (Y1-Y4), maintenance cost budget, and surrounding environmental constraints (such as construction restricted areas). Action A includes strategies such as "emergency maintenance," "planned maintenance," and "monitoring and observation." The reward function R is defined as:

[0070] R = -(α × Risk Level + β × Maintenance Cost + γ × Risk Reduction)

[0071] Where α, β, and γ are weighting coefficients, which are determined through training with historical maintenance data;

[0072] Strategy optimization: Employ the Q-learning algorithm to iteratively update the Q-value table.

[0073]

[0074] Where Q(S) t A t ) is in state S t Take action A t The expected cumulative reward value at time η is the learning rate, R t For state S t Take action A t The immediate reward obtained afterward, with γ as the discount factor, ultimately generates a maintenance strategy that maximizes long-term cumulative rewards, and automatically matches the nearest maintenance team and equipment inventory with the maintenance resource scheduling system.

[0075] Using the above technologies, risk data can be monitored in real time. When an anomaly is detected by the isolated forest or support vector data description (SVDD), the corresponding early warning level (Y1-Y4) is triggered. The optimal maintenance strategy is generated through reinforcement learning, realizing an intelligent closed loop from anomaly detection to decision response.

[0076] S5, Visualization and Interaction

[0077] Using the 3D modeling capabilities of a geographic information system, a custom 3D visualization model of underground pipelines and their surrounding environment can be constructed, and interactive features are also available.

[0078] The 3D visualization model rendering highlights the differentiated display of pipeline types and risk areas, and the interactive functions support pipeline layer-by-layer viewing, spatial relationship query, and risk data linkage retrieval.

[0079] Specifically, the 3D visualization model rendering uses GIS 3D modeling technology to construct a unique model of underground pipelines and their surrounding environment, including spatial coordinates, material properties, and risk levels. The pipeline type is displayed by color and material differentiation (e.g., water supply pipelines use blue semi-transparent material, gas pipelines use orange bright material), and risk areas (e.g., corroded sections, high-pressure areas) are added with red grid textures or flashing effects.

[0080] The interactive features include:

[0081] Layered viewing: Three-dimensional spatial slicing is performed based on burial depth thresholds (such as 0.5-meter intervals) or pipeline type (water supply / drainage / gas). Users can dynamically separate pipeline layers at different depths using slider controls.

[0082] Spatial Relationship Query: Click on any pipeline node with the mouse to calculate and display the minimum distance (based on the Euclidean distance formula) and topological connection relationship with surrounding buildings and other pipelines in real time;

[0083] Risk data linkage retrieval: When a risk area is selected, it automatically links to the underground pipeline geographic information database and displays data reports on the pipeline section, including historical maintenance records, current corrosion rate, and pressure monitoring values.

[0084] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention are also within the protection scope of the present invention.

Claims

1. A method for controlling safety hazards in underground pipelines based on geographic information, characterized in that, The method comprises: S1. Based on the underground pipeline BIM model, the operation data of the target underground pipeline is collected through 3D laser scanning, ground penetrating radar and IoT sensors, and the environmental data interface is connected to obtain terrain, geological and surrounding environmental data to achieve accurate collection of multi-source data. S2. Using the "rule engine + machine learning" algorithm, the collected multi-source data is cleaned, denoised, format converted and coordinate unified, and outliers are removed. The data is then integrated to build an underground pipeline geographic information database, and storage and query are optimized through spatial indexing technology. S3. Based on the analytic hierarchy process and Bayesian network, a multi-factor coupled risk assessment model is constructed. The model comprehensively considers factors such as pipeline material, service life, corrosion degree, surrounding construction activities and geological conditions to conduct regular or real-time dynamic risk assessments and determine the risk level. S4. Anomaly detection is performed using the isolated forest algorithm and the support vector data description algorithm. Reinforcement learning algorithm is combined to generate maintenance or emergency strategies. Early warning is triggered based on the risk level and decision-making suggestions are pushed in conjunction with the algorithm. S5. Utilize the 3D modeling function of the geographic information system to construct a dedicated 3D visualization model of underground pipelines and their surrounding environment, which can also be interactive.

2. The method for controlling underground pipeline safety hazards based on geographic information according to claim 1, characterized in that: In S1, the three-dimensional laser scanning parameters are adjusted according to the distribution density and burial depth characteristics of underground pipelines, the ground-penetrating radar signal processing is adapted to the electromagnetic differences between pipeline materials and geological media, the Internet of Things sensors are deployed according to the pipeline operation monitoring needs, and the environmental data interface is based on the standard of the urban geographic information public platform.

3. The method for controlling underground pipeline safety hazards based on geographic information according to claim 1, characterized in that: In S2, the rule engine constructs data verification rules based on the spatial continuity and attribute logic of underground pipelines, machine learning trains anomaly identification models through historical pipeline data, and spatial indexing technology adopts an optimization algorithm that combines R-trees with pipeline network topology.

4. The method for controlling underground pipeline safety hazards based on geographic information according to claim 1, characterized in that: In S3, the judgment matrix constructed by the analytic hierarchy process integrates historical accident data of underground pipelines with expert operation and maintenance experience. The Bayesian network training data covers pipeline operation, fault cases and surrounding environmental changes. The periodic evaluation cycle is coordinated with the pipeline inspection plan, and the real-time evaluation triggers the update of IoT sensor data.

5. The method for controlling underground pipeline safety hazards based on geographic information according to claim 1, characterized in that: In S4, the anomaly detection thresholds of the Isolation Forest algorithm and the Support Vector Data Description algorithm are determined by historical statistics of the normal operating parameters of underground pipelines; the decision space constructed by the reinforcement learning algorithm includes risk level, maintenance cost, and environmental constraints, and the strategy generation is linked to the maintenance resource scheduling system.

6. The method for controlling underground pipeline safety hazards based on geographic information according to claim 1, characterized in that: In S5, the 3D visualization model rendering highlights the differentiated display of pipeline types and risk areas, and the interactive function supports pipeline layer-by-layer viewing, spatial relationship query, and risk data linkage retrieval.

7. A geographic information-based underground pipeline safety hazard assessment system, characterized in that, This system is based on a geographic information-based method for controlling underground pipeline safety hazards, as described in any one of claims 1-6. Specifically, it includes: a data acquisition module that uses a BIM model of the underground pipeline to collect operational data of the target underground pipeline through 3D laser scanning, ground-penetrating radar, and IoT sensors, and connects to an environmental data interface to obtain surrounding environmental data, achieving accurate collection of multi-source data. This ensures the comprehensiveness and accuracy of the data, providing a reliable foundation for subsequent assessments. A preprocessing and management module utilizes a "rule engine + machine learning" algorithm to clean, denoise, convert formats, and unify coordinates of the multi-source data, removing outliers and integrating them to construct a geographic information database. Spatial indexing technology optimizes storage and retrieval, improving data quality and management efficiency, and laying a solid data foundation for system operation. The basic risk assessment module uses a multi-factor coupled model based on the analytic hierarchy process (AHP) and Bayesian networks to conduct regular or real-time dynamic assessments of pipeline materials and service life to determine risk levels. This allows for a scientific and accurate assessment of pipeline safety hazards, providing a basis for preventative measures. The early warning and decision-making module employs the isolated forest algorithm and support vector data description algorithm for anomaly detection. Combined with reinforcement learning algorithms, it generates maintenance or emergency strategies, triggers early warnings based on risk levels, and pushes decision-making suggestions in a timely manner. This enables timely detection of anomalies and provides scientific decision-making, effectively reducing the probability of safety accidents. The display module utilizes the 3D modeling capabilities of a geographic information system to construct a dedicated 3D visualization model of underground pipelines and their surrounding environment, supporting interactivity. This facilitates intuitive viewing of pipeline conditions, improving management efficiency and decision-making accuracy.

8. The underground pipeline safety hazard assessment system based on geographic information according to claim 7, characterized in that: The preprocessing and management module includes: an intelligent cleaning module and a management module; Intelligent cleaning module: Utilizes a rule engine to construct data verification rules based on the spatial continuity and attribute logic of underground pipelines. Combined with machine learning algorithms, it trains anomaly recognition models on historical pipeline data to achieve cleaning, noise reduction, format conversion, and coordinate unification of multi-source data, eliminate outliers, ensure data quality, and provide a reliable data foundation for subsequent assessments. Management module: The cleaned data is integrated to build an underground pipeline geographic information database. An optimized algorithm combining R-tree and pipeline network topology is used for spatial indexing to achieve efficient data storage and fast query, thereby improving data management efficiency.

9. The underground pipeline safety hazard assessment system based on geographic information according to claim 7, characterized in that: The risk assessment module includes: a dynamic assessment module and a factor construction module; Dynamic assessment module: Based on a multi-factor coupled model constructed using the analytic hierarchy process and Bayesian networks, combined with real-time data updates from IoT sensors to trigger real-time assessment, comprehensively determine the pipeline risk level, and achieve dynamic monitoring of the pipeline safety status; Factor construction module: By integrating historical accident data of underground pipelines with expert operation and maintenance experience through the analytic hierarchy process, a judgment matrix is ​​constructed to determine the factor weights of pipeline material and service life. At the same time, a Bayesian network is trained using data covering pipeline operation, failure cases and changes in the surrounding environment to build the foundation for a multi-factor coupled risk assessment model.

10. The underground pipeline safety hazard assessment system based on geographic information according to claim 7, characterized in that: The early warning and decision-making module includes: an intelligent early warning module and a strategy generation module; Intelligent early warning module: It adopts the isolated forest algorithm and support vector data description algorithm to determine the anomaly detection threshold by the historical statistics of the normal operation parameters of underground pipelines, and monitors the assessed risk data in real time. When the risk exceeds the threshold, an early warning is triggered to improve the accuracy and reliability of the early warning. Strategy generation module: Utilizes reinforcement learning algorithms to construct a decision space that includes risk level, maintenance cost, and environmental constraints. Combined with the maintenance resource scheduling system, it generates maintenance or emergency strategies and pushes decision suggestions in a coordinated manner to achieve intelligent response and handling of risks.

Images