Urban lifeline-oriented underground pipe network spatio-temporal data mining and risk warning method

Abstract

This invention provides a method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines. It relates to the field of urban pipeline network data processing technology. The method divides the underground pipeline network into a known monitoring area grid containing multiple pipeline topology nodes and a spatiotemporal monitoring blind zone grid. Multimodal time-series data is collected from IoT sensor data to construct a spatiotemporal stress field flow matrix for the known monitoring area. By calculating the spatial topological impedance coefficient between the known and blind zone grids and performing attenuation extrapolation, virtual sensor data streams for each node in the spatiotemporal monitoring blind zone grid are obtained. The data are merged to generate a global spatiotemporal stress dynamic evolution matrix. A spatiotemporal influence sequence list is extracted, and cross-domain anomaly characteristic spatiotemporal trajectory chains are identified. Finally, trend extrapolation mapping is used for time-series prediction and compared with the safety envelope boundary. When conditions are met, early warning signals for hidden pipeline leakage and secondary road collapse risks are collaboratively output, achieving proactive and accurate perception of blind zone risks across the entire network.

Description

Technical Field

[0001] This invention relates to the field of urban pipeline network data processing technology, specifically to a method for spatiotemporal data mining and risk early warning of underground pipeline networks, which are vital lifelines for cities. Background Technology

[0002] Urban underground pipe networks, as a vital component of the city's lifeline, bear the core functions of ensuring urban operation, such as water, gas, and power supply. Due to their long-term exposure to complex soil and geological environments, these networks are highly susceptible to hidden leaks caused by factors such as pipe aging, external construction disturbances, and uneven geological settlement. As fluids continuously erode the surrounding soil, secondary road collapses can easily occur, posing a significant threat to urban operational safety. Currently, spatiotemporal data mining and risk warning for underground pipe networks mainly rely on IoT sensors (such as pressure gauges, flow meters, and settlement meters) deployed on certain pipe sections, or on a combination of machine algorithms for risk prediction. However, existing data processing and early warning technologies suffer from the following two technical problems in practical applications: Problem 1: Due to limitations in underground installation environment, construction costs, and network transmission conditions, IoT sensors cannot achieve full network segment and full node coverage. A large amount of pipeline space is in a spatiotemporal monitoring blind zone grid state where no sensors are deployed. Existing technologies cannot effectively mine the actual operating data of each pipeline topology node within the spatiotemporal monitoring blind zone grid, resulting in the active risk warning system being in a state of failure within the blind zone. The second problem is that existing risk warning methods often treat fluid IoT data (pressure, flow rate) and geological monitoring data (micro-displacement) as independent indicators for separate judgment, or directly feed them into machine algorithm models for statistical fitting. This approach ignores the causal transmission mechanism of underground fluid dynamics and solid mechanics in space, and is easily interfered with by occasional noise from sensors, resulting in high false alarm and false negative rates under extreme conditions and a lack of industrial interpretability. Summary of the Invention

[0003] To achieve the above objectives, this invention provides the following technical solution: a method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines, the method comprising: Acquire pre-stored three-dimensional geographic information data, geological property parameters, and pipe material performance parameters of urban underground pipe networks, and divide the underground pipe network into known monitoring area grids and spatiotemporal monitoring blind area grids. Each grid contains multiple pipe network topology nodes. Collect multimodal time-series data within the known monitoring area grid, and construct the spatiotemporal stress field flow matrix of the known monitoring area; Calculate the spatial topological impedance coefficient between the known monitoring area grid and the spatiotemporal monitoring blind zone grid, and perform attenuation extrapolation by combining geological physical parameters and spatiotemporal stress field flow matrix. Calculate the virtual stress extrapolation value of each pipeline topology node in the spatiotemporal monitoring blind zone grid and convert it into a blind zone virtual sensor data stream. The spatiotemporal stress field flow matrix is ​​merged with the data stream of the virtual sensor in the blind zone to generate a global spatiotemporal stress dynamic evolution matrix. The spatiotemporal influence sequence table is extracted to identify the spatiotemporal trajectory chain of cross-domain anomaly characteristics. The spatiotemporal trajectory chain of cross-domain anomaly characteristics is predicted in a forward time series to generate a stress evolution curve. The stress evolution curve is compared in real time with the safety envelope boundary constructed based on pipe performance parameters. When the triggering condition is met, a dual collaborative early warning signal is output.

[0004] This invention provides a method for spatiotemporal data mining and risk early warning of underground pipeline networks, which are essential for urban lifelines. It offers the following advantages: 1. This invention utilizes three-dimensional geographic information data of urban underground pipe networks to divide the pipe network space into a known monitoring area grid nested with multiple pipe network topology nodes and a spatiotemporal monitoring blind zone grid. By calculating the spatial topological impedance coefficient between the known and blind zone grid nodes, and combining the energy dissipation coefficient of the underground medium characterized by soil saturation and density, a rigorous non-uniform conduction mapping of the underground medium is established. By adaptively attenuating the stress values ​​of the known monitoring area grid nodes along the complex physical topology path of the pipe network, the virtual stress estimation values ​​of each pipe network topology node in the spatiotemporal monitoring blind zone grid are reconstructed with high precision without relying on physical sensors. These values ​​are then reverse-interpreted and converted into a virtual sensor data stream for the blind zone, enabling spatial continuity in data mining. This fundamentally overcomes the technical bottleneck of the inability to detect hidden dangers in blind zones, realizing proactive perception and early warning in sensorless areas, eliminating monitoring blind zones, and providing comprehensive spatiotemporal data support for proactive prevention of sensorless coverage segments of urban lifelines.

[0005] 2. This invention performs equivalent transformation on the collected instantaneous pressure time-series data and vertical micro-displacement time-series data, fusing them into a spatiotemporal stress field flow matrix of the known monitoring area that includes spatial adjacency relationships. This achieves a unified multi-physics field representation of fluid dynamic stress and solid static compressive stress in the spatiotemporal grid dimension. Based on fluid stress changes and causal verification, it captures the causal time-series relationship between "dynamic fluid stress drop events" and "static compressive stress anomaly fluctuation events" in the surrounding soil along the topological path, accurately identifying the spatiotemporal trajectory chain of cross-domain anomaly characteristics. Utilizing the physical transmission mechanism between fluid and solid, data processing is performed, with a clear and verifiable logical chain and high interpretability. It can accurately identify and automatically filter out occasional noise caused by sensor hardware failures and false exceedance noise caused by severe weather. By comparing the trend extrapolation prediction curve with the safety envelope boundary constructed based on pipe material performance and geological properties point by point, it achieves dual collaborative proactive risk output for leakage and collapse, effectively reducing the false alarm rate and missed alarm rate, and improving the early warning accuracy under complex operating conditions of urban lifelines. Attached Figure Description

[0006] Figure 1 This is a flowchart of the spatiotemporal data mining and risk warning method for underground pipeline networks in urban lifelines, based on the present invention. Detailed Implementation

[0007] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0008] like Figure 1 A spatiotemporal data mining and risk early warning method for underground pipeline networks, a vital lifeline for cities, including: Step S100 involves acquiring pre-stored 3D geographic information data, geological property parameters, and pipe material performance parameters of the urban underground pipe network. The underground pipe network is then divided into known monitoring area grids and spatiotemporal monitoring blind area grids, with each grid containing multiple pipe network topology nodes. Since the underground pipe network is intricate and buried underground, full-line sensor coverage is impossible. Grid division allows limited IoT hardware resources to be focused on known monitoring areas, while clearly defining spatiotemporal monitoring blind areas lacking direct data sources, thus guiding subsequent data flow completion. Under this architecture, each pipe network topology node serves as a digital mapping of key physical locations such as pipe intersections, diameter changes, bends, valves, and monitoring points. It not only connects physical pipe relationships but also serves as a discretized computational point for stress wave propagation, hydrodynamic amplitude changes, and settlement displacement transmission, used to spatially anchor and transmit high-dimensional pipe physical state data.

[0009] Step S200: Collect multimodal time-series data within the known monitoring area grid and construct the spatiotemporal stress field-flow matrix of the known monitoring area. By constructing the spatiotemporal stress field-flow matrix, heterogeneous data from different types of sensors, reflecting different physical processes, are integrated into a unified, structured, and computable multiphysics state representation. This reflects the comprehensive mechanical coupling state of each pipeline topology node within the known monitoring area at each historical sampling moment, simultaneously subjected to dynamic pressure from the fluid inside the pipeline and static extrusion pressure from the soil outside the pipeline. It reveals how stress diffuses, attenuates, or superimposes in the pipeline space with fluid movement and time progression, reflecting the dynamic energy conservation and transformation laws of the pipeline network caused by severe disturbances in the internal and external environments in the physical world.

[0010] Step S300: Calculate the spatial topological impedance coefficient between the known monitoring area grid and the spatiotemporal monitoring blind zone grid. Combine this with geological property parameters and the spatiotemporal stress field flow matrix to perform attenuation extrapolation. Calculate the virtual stress extrapolation value of each pipeline topology node in the spatiotemporal monitoring blind zone grid and convert it into a blind zone virtual sensor data stream. By calculating the spatial topological impedance coefficient, the resistance encountered by stress waves propagating along the physical path of the pipeline network between the known monitoring area and the spatiotemporal monitoring blind zone is quantitatively characterized. This reflects the hindering characteristics of the pipeline network's physical structure on energy transfer. A high impedance coefficient means a large physical connection barrier between the two nodes, making it difficult for stress to be effectively transmitted. The blind zone virtual sensor data stream derives continuous state data in a vacuum grid area without any physical hardware monitoring, using the transmission laws of known boundaries. It reflects the timely "simulated" physical response of the blind zone pipeline to fluid and stratum stress waves from adjacent known areas, compensating for the sensing gaps caused by insufficient hardware coverage and piecing together the fragmented monitoring network into a spatiotemporally continuous and extrapolable complete physical state.

[0011] Step S400 involves merging the spatiotemporal stress field flow matrix with the data stream from the virtual sensor in the blind zone to generate a global spatiotemporal stress dynamic evolution matrix. A spatiotemporal influence sequence list is extracted, and the spatiotemporal trajectory chain of cross-domain anomaly characteristics is identified. The global spatiotemporal stress dynamic evolution matrix reflects the continuous evolution history and current distribution of internal and external stresses at any location and time within the entire pipeline network. By identifying the spatiotemporal trajectory chain of cross-domain anomaly characteristics, spatial tracing and future path locking are performed on the "originating point, propagation path, evolution speed, and impact range" of abnormal stress waves. This reflects the transmission mechanism of hidden defects within the pipeline network (such as soil erosion caused by micro-leakage and stress concentration in the pipe body) in both spatiotemporal dimensions, enabling maintenance personnel to clearly see how anomalies propagate across lines and revealing the dynamic evolution trajectory of the disaster chain.

[0012] Step S500 involves forward temporal evolution prediction of the spatiotemporal trajectory chain of cross-domain anomaly characteristics to generate a stress evolution curve. This curve is then compared in real-time with the safety envelope boundary constructed based on pipe performance parameters. When triggering conditions are met, a dual-coordinated early warning signal is output. By calculating the stress evolution curve, historical and current static and dynamic stress anomaly trends are quantitatively and visually extended into the future, thus predicting the trajectory of risk development. This reflects how the stress values ​​of identified anomalies will change in the future if left unchecked. The dual-coordinated early warning signal, through bidirectional verification of "intra-pipe fluid dynamic anomalies" and "extra-geological micro-displacement mutations," eliminates the interference of false alarms from single indicators. It reflects the deep resonance mechanism of pipeline structural safety being jointly induced by internal factors (water hammer, overpressure) and external factors (ground subsidence, external construction pressure). Through bidirectional verification, the issued early warning instructions are ensured to have extremely high physical support certainty and disaster-specificity.

[0013] In this embodiment, the underground pipe network is divided into a known monitoring area grid and a spatiotemporal monitoring blind area grid. The implementation steps for collecting multimodal time series data within the known monitoring area grid include: Step S101: Acquire pre-stored 3D geographic information data of the urban underground pipeline network, parse out the spatial geometric network of all pipelines, valve nodes, and the installation locations of IoT sensors, and divide the underground pipeline network into multiple grids. Based on whether IoT sensors are deployed, the grids are classified into known monitoring area grids and spatiotemporal monitoring blind zone grids. During grid division, first, the spatial geometric network of all pipelines and the installation coordinates of all IoT sensors are loaded from the 3D geographic information data. Then, with a sensor as the center, an initial envelope area is defined on the 2D horizontal projection plane based on the pipeline direction. All pipes and nodes within this area are defined as a known monitoring area grid. This process is repeated for all sensors to obtain all known monitoring area grids. Finally, the entire urban pipeline network is traversed, and all continuous pipeline areas not covered by any known monitoring area grids are divided into one or more independent spatiotemporal monitoring blind zone grids. Through grid division, data-complete areas and data-vacuum areas are explicitly separated, enabling the system to perform targeted conventional data governance on known areas and deduce the data flow of virtual sensors in blind zones.

[0014] When dividing underground pipe networks into grids, adaptive spatial partitioning is performed based on the spatial density of pipelines, the physical deployment locations of IoT sensors, and the distribution boundaries of urban geological risk levels. The grid size is not fixed; for example, it can be a rectangle or irregular polygon ranging from 50m×50m to 200m×200m. In dense main pipe networks and high-risk areas, the grid size is generally set to a refined cubic space of 50m×50m. In secondary branches or areas with sparse sensors, the grid size is generally expanded to a coarse space of 200m×200m. Multiple pipeline topology nodes within each grid are defined as all three-dimensional turning points, branching and merging points, diameter change points, and hardware deployment points of pipelines falling within the boundary of the grid's spatial cube. Specifically, these include key feature points such as valves, elbows, tees, diameter change points, endpoints, and sensor installation points on all pipelines. Each node includes its three-dimensional spatial coordinates, node type (such as valve or elbow), and connection relationship with adjacent nodes. Depending on the grid size and pipeline complexity, the number of nodes contained in each grid is not fixed. A typical fine grid generally contains 5 to 15 pipeline topology nodes.

[0015] The three-dimensional geographic information data, acquired from urban surveying and planning departments, includes the spatial geometric network of the entire city's pipelines, valve node coordinates, and sensor installation locations. This data is used to establish a global spatial geometric benchmark, providing a physical carrier for the spatial anchoring of all stress and fluid data. Geological property parameters, obtained from geological exploration reports or soil databases, include soil elastic modulus, saturation, density, and shear strength limit. These are used to calculate the static and dynamic resistance of the soil to the pipelines, providing stratigraphic constraint parameters for assessing external compressive stress. Pipe material performance parameters, obtained from pipeline design and construction drawings and material specifications, include pipe diameter, material, deformation limit, and internal wall friction coefficient. These parameters are used to establish the physical stiffness and strength boundaries of the pipeline under stress, determining whether stress is excessive.

[0016] The spatial geometric network of the entire city's pipeline system refers to a three-dimensional topological skeleton of the pipeline network centerline, constructed based on three-dimensional geographic information data and free from unnecessary physical details. Line segments represent pipe segments, and intersections represent pipeline connections, forming a closed, connected, and directional directed spatial graph network. Valve nodes are defined as key hard feature points in the spatial geometric network of the entire city's pipeline system that actively control fluid flow, regulate flow and pressure, and exhibit sudden changes in local head loss. Physically, they act as checkpoints for regulating and blocking fluid energy; in data processing, they serve as control inflection points for pressure jump boundaries and changes in fluid flow direction.

[0017] Step S102 involves introducing urban surveying control points as a unified three-dimensional rectangular coordinate system benchmark. The installation spatial coordinates of the IoT sensors are obtained and statically aligned with the pipe performance parameters, including pipe diameter, material parameters, and elevation information. During static spatial benchmark alignment, the precise three-dimensional coordinates of the control points in the CGCS2000 coordinate system or a local independent coordinate system are first obtained from the urban surveying department. Then, multiple urban surveying control points closest to the pipeline network are retrieved, and the translation, rotation, and scaling transformation factors from the relative construction coordinate system to the absolute urban three-dimensional rectangular coordinate system are calculated. All pipeline topology node coordinates, sensor spatial locations, and geological structure layer boundary data are multiplied by these transformation factors, and a spatial coordinate projection transformation is performed. Resampling verification is then conducted using control points to correct projection distortions caused by elevation differences and Earth curvature, ensuring all data is seamlessly aligned to the same three-dimensional rectangular coordinate system. Finally, the corrected sensor spatial coordinates are associated and bound one-to-one or one-to-many with the pipe performance parameters (such as pipe diameter and material) and pipe elevation information extracted from the design drawings, to ensure that when calculating the stress at a specific coordinate point, the physical anomalies captured by the sensor can be accurately located to the specific physical pipe section and geological layer.

[0018] Among them, urban surveying control points are basic urban surveying facilities. They are permanent markers on the ground with precise and unified national or local plane coordinates and elevations, such as GPS control stakes along main urban roads. These are obtained from the urban surveying authorities and serve as a unified three-dimensional rectangular coordinate system benchmark. This eliminates systematic errors between data from different sources (such as pipeline drawings and sensor installation coordinates from different years and construction units), ensuring that all pipeline nodes, sensors, and geological data are in the same coordinate system, thus achieving precise spatial alignment of the data.

[0019] Step S103: All IoT sensors within the known monitoring area grid acquire multimodal time-series data using a unified timestamp reference and sampling frequency. IoT sensors are used to capture transient physical quantities of the internal operating status of the pipeline network and external environmental disturbances in real time, converting the physical evolution of the physical world into continuous digital signals. These sensors include, but are not limited to, instantaneous pressure sensors and electromagnetic flowmeters installed inside the pipeline, and vertical micro-displacement sensors and fiber optic vibration sensors attached to the pipe wall or buried around the pipeline. To accurately capture the transient characteristics of fluid flow and minute vibrations in the pipe body, the sampling frequency of instantaneous pressure sensors and vertical micro-displacement sensors is generally set to a high frequency of 50Hz to 100Hz. For flow data with relatively gentle amplitude variations, the sampling frequency of electromagnetic flowmeters can be appropriately reduced to 1Hz to 5Hz. Alternatively, depending on the speed of dynamic changes in the pipeline network, the sampling frequency can be uniformly set to once every 5 to 30 seconds. For main pipelines with frequent pressure fluctuations, a high-frequency sampling of 1 to 5 seconds can be used, while for secondary or branch pipelines with slow changes, a sampling frequency of 30 to 60 seconds can be used. In actual use, the selection and setting should be made according to the situation.

[0020] At the same sampling time, instantaneous pressure time-series data uploaded by the instantaneous pressure sensor, instantaneous flow time-series data uploaded by the electromagnetic flowmeter, and continuously changing vertical micro-displacement time-series data uploaded by the vertical micro-displacement sensor are synchronously read, together forming multimodal time-series data. Instantaneous pressure time-series data is used to capture energy fluctuations in the fluid within the pipe caused by pump start-up and shutdown, valve opening and closing, or pipe bursts, reflecting the transient pressure wave transmission of fluid dynamics within the pipe network, and can sensitively expose endogenous water hammer destructive forces.

[0021] Instantaneous flow time-series data is used to calculate the total amount of fluid transported and the flow velocity and direction at a specific cross-section of the pipeline network. It reflects the smoothness of material flow and the balance of flow distribution within the pipeline network. Flow step data can directly reflect whether there are hidden leaks or illegal theft in the pipeline. Vertical micro-displacement time-series data is used to monitor the absolute three-dimensional displacement of the pipe body caused by external stratum disturbance. It reflects the external morphological damage to the pipe body's structural stiffness caused by external soil subsidence, uneven uplift, or lateral compression, and directly reflects the severity of the pipeline network's threat from external geological disasters.

[0022] In this embodiment, the steps for constructing the spatiotemporal stress field flow matrix of a known monitoring area include: Step S201: Combine the pipe cross-sectional area parameters to perform stress equivalence transformation on the instantaneous pressure time series data, and extract the dynamic fluid stress vector borne by the pipe inner wall. Calculate the product of the specific pipe cross-sectional area and the instantaneous pressure time series data within the known monitoring area grid to extract the dynamic fluid stress vector borne by the pipe inner wall. The dynamic fluid stress vector refers to the set of axial frictional tensile stress and radial shear compressive stress generated on the pipe inner wall by the fluid flowing inside the pipe and experiencing pressure pulsation. It includes instantaneous head hydrostatic stress and momentum impact stress caused by changes in fluid velocity, and is used to quantitatively assess the fatigue failure and damage risk caused by internal fluid loads on the pipe wall. During the calculation, real-time instantaneous pressure and velocity data at a specific pipe cross-section are obtained. Based on the specific pipe cross-sectional area, the instantaneous pressure is applied to the inner surface area of ​​the pipe wall to obtain the radial tensile stress borne by the pipe wall. Using the velocity data and fluid viscosity, the frictional resistance at the contact surface between the fluid and the inner wall of the pipe is calculated to obtain the shear stress along the pipe axis. The radial tensile stress and axial shear stress are then combined into a spatial vector to output the dynamic fluid stress vector at that specific pipe cross-section. For common fluids, such as water and natural gas, the viscosity can be obtained from the corresponding physical property manual or by consulting a database. For mixed fluids, such as sewage and gas-liquid mixtures, a viscometer can be installed inside the pipeline to directly monitor and obtain the viscosity.

[0023] Soil elastic modulus corresponding to known monitoring area grids is extracted from geological physical parameters. Vertical micro-displacement time-series data are multiplied and transformed with the soil elastic modulus to obtain the static compressive stress vector on the pipeline outer wall caused by uneven soil settlement. This static compressive stress vector, representing the spatial constraint stress directly acting on the pipeline outer wall due to the weight of the overlying soil, surface traffic live load, and soil deformation, is used to assess whether the pipeline will experience buckling, flattening deformation, or fracture under external loads.

[0024] Step S202 involves using the dynamic fluid stress vector and static compressive stress vector as dynamic features of the corresponding pipeline network topology nodes. Combined with pipe performance parameters, and using three-dimensional spatial coordinates as index keys, a mesh adjacency reorganization is performed to construct a spatiotemporal stress field flow matrix for the known monitoring area. Each element in the matrix represents the known coupling state value of the internal and external bidirectional mechanics of the corresponding mesh node at a specific sampling time. First, based on a set high-frequency sampling timestamp, the dynamic fluid stress vector and static compressive stress vector read from all known monitoring points within all meshes are aligned by a time step on a second-by-second basis. Then, within each known monitoring area mesh, the aligned dynamic fluid stress vector and static compressive stress vector are spatially synthesized to obtain the total resultant force vector of each pipeline network topology node within that mesh at the current time step. Next, a new three-dimensional spatial matrix is ​​created, with its rows, columns, and height corresponding to the three-dimensional spatial grid indices of the underground pipe network. Each cell of the matrix acts as a storage bucket, filling the corresponding matrix cell storage bucket with the calculated total resultant force vector from each known grid, along with the directional flux of the fluid flow. For cells corresponding to blind zone grids, initial empty values ​​are temporarily filled in (to be filled in later with topological impedance calculations). As the time step progresses, the spatial grid force matrices generated by each discrete timestamp are concatenated and assembled along the time axis, ultimately generating a multidimensional continuous spatiotemporal stress-flow matrix in memory. The spatiotemporal stress-flow matrix is ​​a multidimensional, time-series nested spatial grid data structure, including the combined superposition value of dynamic fluid stress vectors and static compressive stress vectors on each pipe network topology node and pipe segment in the entire grid at a specific time axis, as well as the flux direction and rate of stress transmission between grid cells.

[0025] The specific cross-sectional area of ​​the pipe is obtained from the pipe material performance parameters bound to the pipe network topology node. The specific pipe is defined as the specific section of pipe with a clearly defined diameter to which the pipe network topology node is attached. For example, at “DN400 steel pipe - node A”, DN400 is the specification of the specific pipe.

[0026] Soil elastic modulus refers to the proportionality coefficient of soil when stress and strain are linearly related during the compression stage. It includes soil compression modulus, deformation modulus and shear modulus. It is obtained based on field load test, triaxial shear test data or in-situ standard penetration test data in the engineering geological survey report. It is used to measure the ability of the soil around the pipeline to resist deformation and determines how much compressive force or loss of support the soil will transmit to the buried pipeline when the stratum settles.

[0027] In this embodiment, the steps for calculating the spatial topological impedance coefficient between the known monitoring area grid and the spatiotemporal monitoring blind zone grid include: Step S301: In the spatially integrated network constructed based on 3D geographic information data, the pipeline topology nodes of the spatiotemporal monitoring blind zone grid to be simulated are taken as the target endpoints, and the pipeline topology nodes of the adjacent known monitoring area grids are taken as the search starting points. All interconnected pipe segments and fitting sequences between the search starting point and the target endpoint are traversed, and disconnected or closed non-connected branches are eliminated. For each retrieved sequence, the attribute parameters of each pipe segment and node along the path are extracted according to the order of physical connection, and their cumulative physical length is recorded. The extracted complete set of attribute parameters is encapsulated as a specific connected path between the search starting point and the search endpoint. All connected paths between the two are retrieved from the 3D geographic information data. A connected path refers to a physically accessible fluid and stress transmission channel connecting the pipeline topology nodes of the known monitoring area grid and the pipeline topology nodes of the spatiotemporal monitoring blind zone grid in the underground pipeline spatial geometric network. This includes the logical arrangement order of all single pipe segments, tees, elbows, reducers, and valve nodes along the path, establishing a clear physical energy transmission channel for the data simulation of the blind zone.

[0028] Pipeline connectivity parameters are extracted from the connected path. These parameters are then used to calculate the kinetic energy loss ratio of the fluid during transmission within the path, which is then used as the spatial topological impedance coefficient. Pipeline connectivity parameters include the physical length of the pipeline, the number of local bends, the number of reducing valves, the internal friction coefficient of the pipeline wall, and the elevation change slope. First, the physical length of the pipeline in the connected path is calculated and linearly weighted to obtain a distance-weighted value. The physical length is then multiplied by the internal friction coefficient to obtain the frictional resistance contribution. Next, the sum of the local resistance coefficients for all local bends and reducing valves is calculated to obtain the local resistance contribution. Finally, the elevation change slope value in the connected path is obtained. If the slope is upward from the starting point, the elevation contribution is positive, equal to the increase in potential energy per unit weight of fluid; if the slope is downward from the starting point, the elevation contribution is negative. Finally, the contributions of frictional resistance, local resistance, elevation, and distance are algebraically summed to obtain a total impedance base within the range of 0 to 1. This base is then mapped to the final spatial topological impedance coefficient; the closer the value is to 1, the higher the proportion of kinetic energy loss. The spatial topological impedance coefficient quantitatively characterizes the total proportion of kinetic energy loss when fluid pressure waves (or stress waves) propagate from a known monitoring zone node along the physical topology of the pipeline network to a spatiotemporal monitoring blind zone node, reflecting the comprehensive hindering ability of pipeline connectivity parameters on energy transmission.

[0029] When applying distance-weighted linear weighting to the physical length of a pipeline, a baseline physical length (e.g., 100 meters) is first set, and the physical length of the pipeline along the connecting path is obtained. Next, the physical length of the pipeline is divided by the baseline physical length to obtain a length factor. Finally, the length factor is multiplied by a preset base weight value representing the "impedance contribution per unit length" to obtain the distance-weighted value. For example, if the impedance contribution per unit length is set to 0.002, and the actual physical length is 500 meters, then the length-weighted contribution is (500 / 100) × 0.002 = 0.01. In the complete impedance coefficient calculation, the length-weighted contribution is only one component, and its weight is typically set to account for 30% to 50% of the spatial topological impedance coefficient. The longer the distance, the greater the weight, but it needs to be normalized and balanced with other factors such as local resistance to ensure that the spatial topological impedance coefficient does not exceed 1.

[0030] The physical length of the pipeline is directly obtained from the distance field in the pipeline design as-built drawings or GIS data, reflecting the fundamental contribution of the conduction distance to energy attenuation. The number of local bends in the pipeline is calculated based on the number of bends (including those at different angles) recorded in the connected path. As a core input for the accumulation of local resistance, each bend causes additional energy loss due to fluid pressure and stress waves. The number of reducing valves is obtained by statistically analyzing the diameter changes or valve nodes in the path, reflecting the concentrated energy dissipation caused by sudden changes in pipe diameter or valve throttling. The friction coefficient of the pipeline inner wall is selected based on the material (e.g., cast iron, PE, steel) and the roughness of the pipeline inner wall in the pipe material performance parameters, used to quantify the friction resistance between the fluid and the pipe wall. The elevation change slope value is calculated based on the ratio of the elevation difference between the start and end points of the connected path to the horizontal distance, used to incorporate the influence of gravitational potential energy changes on pressure wave conduction into the impedance calculation; an uphill slope increases impedance, and a downhill slope decreases impedance.

[0031] The local resistance coefficient is used to quantify the additional resistance that a single pipe component (such as an elbow, tee, reducer, or valve) exerts on the fluid. This coefficient is obtained from fluid mechanics engineering handbooks or pipe design specifications and is set according to the specific type, shape, and size ratio of the pipe component. For example, the local resistance coefficient of a 90° standard elbow is approximately 0.5 to 1.0, that of a fully open gate valve is approximately 0.1 to 0.2, and that of a reducer with a sudden narrowing is approximately 0.3 to 0.5. Its value range is generally from 0 to 10. For example, a downstream tee is generally 0.2 to 0.5, while a half-open gate valve may reach 5.0 to 8.0.

[0032] In this embodiment, the steps of calculating the virtual stress projection values ​​of each pipeline topology node in the spatiotemporal monitoring blind zone grid and converting them into a blind zone virtual sensor data stream include: Step S302: Obtain the known coupling state values ​​of each pipe network topology node within the adjacent known monitoring area grid; calculate the three-dimensional spatial straight-line distance between a specific pipe network topology node in the spatiotemporal monitoring blind zone grid to be simulated and the pipe network topology nodes in the known monitoring area grid; extract the pre-stored soil saturation and soil density of adjacent areas from geological property parameters. The known coupling state value refers to the comprehensive stress scalar value output by the pipe network topology node within the known monitoring area grid at the current time step, after the interaction between the internal fluid dynamic vector and the external stratum compressive stress vector reaches equilibrium. This includes the interweaving result of the dynamic fluid stress component generated by the fluid pressure at the node and the static compressive stress component generated by the settlement of the surrounding soil, serving as the physical source strength benchmark for simulating the unmeasured state of the blind zone. By using a known monitoring area grid's pipe network topology node that is spatially adjacent (e.g., closest in a straight line) to a specific pipe network topology node within the spatiotemporal monitoring blind zone grid to be simulated, and based on the current simulation time (usually the latest time), using the spatial coordinates of the pipe network topology node of that known monitoring area grid as the index key, the dynamic fluid stress vector and static extrusion stress vector corresponding to that moment can be directly queried and extracted from the constructed spatiotemporal stress field flow matrix. These are the known coupling state values.

[0033] By combining known coupling state values ​​and spatial topological impedance coefficients with soil saturation, soil density, and three-dimensional spatial straight-line distance, numerical attenuation calculations are performed to obtain virtual stress projection values ​​for specific pipeline topology nodes within the spatiotemporal monitoring blind zone grid. The underground medium energy dissipation coefficient is calculated based on soil saturation and soil density. Using the known coupling state value as the base, this is multiplied by the negative exponent of the product of the underground medium energy dissipation coefficient and the three-dimensional spatial straight-line distance, and then multiplied by the complementary angle ratio of the spatial topological impedance coefficient to calculate the virtual stress projection values ​​for the pipeline topology nodes within the spatiotemporal monitoring blind zone grid. The virtual stress projection value refers to the estimated comprehensive stress state of the pipe body simulated within the spatiotemporal monitoring blind zone grid using physical conduction laws. It reflects the actual stress deformation trend of the pipeline in the blind zone under the energy wave influence of adjacent known areas, and the expected stress level after being affected by the surrounding real stress field, filling the sensing gap in sensorless areas. The specific calculation formula is as follows: ; in, This represents the virtual stress projection value; This indicates a known coupling state value; The base of the natural logarithm is a constant. It represents the energy dissipation coefficient of the underground medium, which is determined by the soil saturation and soil density of the adjacent area; Represents straight-line distance in three-dimensional space; This represents the spatial topological impedance coefficient. Among them, the energy dissipation coefficient of the underground medium... The value range is typically between 0.001 and 1. For dry, dense sandy soil, energy transfer is relatively good. The values ​​are relatively small, approximately 0.01 to 0.1; for saturated, loose silty clay, energy dissipates extremely quickly. The values ​​are relatively large, ranging from 0.5 to 1.0. They are generally obtained by consulting an empirical table of geological physical parameters that match the soil type in the area, or by conducting active source stress tests (such as micro-excitation) at known points on site and then inverting and fitting the results.

[0034] Among them, the three-dimensional spatial straight-line distance refers to the geometric spatial Euclidean distance between the pipeline topology nodes of the known monitoring area and the pipeline topology nodes of the spatiotemporal monitoring blind area in the absolute three-dimensional rectangular coordinate system of the city. It includes three spatial projection displacement components along the east-west direction (X-axis), the north-south direction (Y-axis), and the elevation direction (Z-axis). By obtaining the coordinates of the X-axis, Y-axis, and Z-axis of the two nodes, the coordinate difference in each direction is calculated, the square of the coordinate difference is summed, and then the square root is taken to obtain the three-dimensional spatial straight-line distance between the two nodes.

[0035] Soil saturation reflects the degree to which the pores of the underground medium are filled with water. Its role, along with soil density, is to determine the propagation speed and energy dissipation rate of stress waves in the soil. Excessively high soil saturation (close to 100%) or too low soil saturation (close to 0%) will lead to an increase in the energy dissipation coefficient α. Higher soil saturation indicates stronger soil fluidity, increasing the risk of lateral compression and erosion of the pipeline. Soil density reflects the compactness of soil particle packing and, together with saturation, describes the mechanical stiffness of the soil. Higher soil density indicates harder soil and faster stress wave propagation, but non-uniformity can also lead to increased energy scattering and dissipation. Lower soil density suggests the soil may be loose, porous, or uncompacted, easily leading to localized suspension and uneven settlement of the pipeline. Both saturation and density reflect the soil's ability to absorb energy as a propagation medium.

[0036] Step S303: The virtual stress projection values ​​are reverse-interpreted to obtain a virtual sensor data stream, which includes virtual pressure time-series data and virtual instantaneous flow rate data. The virtual pressure time-series data simulates the function of a pressure gauge in the blind zone, used to determine whether a sudden pressure drop or abnormal fluctuation has occurred, reflecting the fluid kinetic energy state within the blind zone pipe section. The virtual instantaneous flow rate time-series data simulates the function of a flow meter, used to verify the rationality of pressure changes (e.g., a pressure drop accompanied by a sudden increase in flow rate indicates a very high probability of leakage), reflecting the volumetric flux change of the fluid within the blind zone pipe section.

[0037] The reverse interpretation of virtual stress projection values ​​is primarily based on the hydrostatic equilibrium relationship and energy conservation principle (such as the core idea of ​​Bernoulli's equation) of fluid flowing in a pipe. The aim is to transform this single stress value into a more intuitive and physically meaningful sensor data stream. The process of generating the virtual sensor data stream involves first obtaining the virtual stress projection values ​​calculated from the pipe network topology nodes of the spatiotemporal monitoring blind zone grid, and then obtaining pipe material performance parameters such as pipe diameter and medium density for the pipe segment where the topology node is located. Next, using the hydrostatic relationship that pressure equals the force exerted by stress on a unit area, the virtual pressure value of the node is calculated by dividing the virtual stress projection value by the cross-sectional area of ​​the pipe. Then, using the fluid continuity equation and energy conservation relationship, combined with the flow or pressure boundary conditions of known upstream and downstream areas (e.g., attenuation projection from instantaneous flow data of adjacent known nodes), the virtual flow velocity passing through the node is estimated, multiplied by the pipe cross-sectional area, to calculate the virtual instantaneous flow rate. The calculated virtual pressure and virtual flow rate are arranged in chronological order and packaged into a virtual sensor data stream consistent with the format of real sensor data.

[0038] When performing reverse interpretation of virtual stress projection values, select a pipeline topology node in the spatiotemporal monitoring blind zone grid that carries the main flow and has a relatively large surrounding geological displacement amplitude, and select the geometric inflection point of this node for interpretation. The geometric inflection point is the intersection point of the axes of pipelines with different directions in three-dimensional space. Geometric inflection points (such as elbows and deformation junctions) are the "stress-sensitive weak points" in the pipeline physical structure where stress concentration is most severe and the impact of fluid flow direction change is the greatest. The stress change amplification effect is most obvious at the inflection point. If the spatiotemporal monitoring blind zone grid is damaged or destroyed by external forces, the virtual stress projection value at the inflection point will generate a step response immediately. Deploying the interpretation point at this location can greatly improve the system's sensitivity to capturing weak disturbance signals and the sensing signal-to-noise ratio.

[0039] The combined application of hydrostatic equilibrium in reverse interpretation: Hydrostatic equilibrium refers to the physical laws of energy conservation and momentum balance that strictly govern the pressure difference, elevation difference, head loss, and kinetic energy change of fluid flow between any two points in a physical pipe network. In reverse interpretation, the system first uses this equilibrium relationship, taking the actual pressure data of the upstream known area as the initial head boundary, and then corrects for gravitational potential energy by combining the elevation difference between the two points (provided by the absolute coordinates in three-dimensional space). Then, the virtual pressure of the blind zone obtained by reverse interpretation is compared and verified with the theoretical reference head derived from this equilibrium relationship. The resistance loss term is adjusted through iterative adjustment, thereby ensuring that the interpreted virtual pressure and flow rate can not only reasonably explain the virtual stress of the pipe wall, but also fully satisfy the global hydraulic equilibrium of the entire pipe network system.

[0040] In this embodiment, the steps of merging the spatiotemporal stress field flow matrix with the blind zone virtual sensor data stream to generate a global spatiotemporal stress dynamic evolution matrix and extracting the spatiotemporal influence sequence table include: Step S401 involves spatially recombining the spatiotemporal stress field flow matrix of the known monitoring area with the virtual sensor data stream of the blind area using three-dimensional Cartesian coordinates to generate a global spatiotemporal stress dynamic evolution matrix. By initializing a spatial multidimensional grid tensor covering the entire underground pipe network and containing continuous time steps, the data from the spatiotemporal stress field flow matrix generated in step S202 is filled into the corresponding coordinate positions according to the grid and node. The virtual stress extrapolation values ​​generated in step S303 and the interpreted virtual sensor data stream are filled into the corresponding grid cells according to spatial coordinates. A spatial boundary smoothing operation is initiated to check the data continuity at the grid boundary between the known area and the blind area, correcting isolated spatial anomalies, and stitching together the data from all grids at the same timestamp, then horizontally assembling them along the time sliding axis to finally output the global spatiotemporal stress dynamic evolution matrix. The global spatiotemporal stress dynamic evolution matrix is ​​composed of elements such as the magnitude of stress, principal stress direction, shear deformation, and stress change rate over time for each spatial grid unit in the global domain. It reflects the spatial transfer and temporal cumulative evolution trend of the stress state of the entire underground pipeline system under the combined action of internal and external loads, as well as the complete evolution life cycle of the pipeline structure from healthy to fatigue, yielding, and finally failure.

[0041] Step S402: Set a sliding monitoring time window, continuously calculate the stress gradient change rate of any pipeline network topology node in the global spatiotemporal stress dynamic evolution matrix between adjacent sampling times, and extract the absolute value of the stress gradient change rate. The stress gradient change rate refers to the ratio of the change in stress value between two adjacent sampling times at the same pipeline network topology node to the time interval. It is used to quantify the severity of stress change over time and is a core indicator for determining whether a "sudden change" has occurred, reflecting the instantaneous dynamic stability of the stress at that node. For any node, obtain its stress value at time t. and Stress value at time ,in Given the sampling interval, the rate of change of the stress gradient is... The calculation is as follows Finally, the absolute value of the stress gradient change rate is compared with the background anomaly threshold.

[0042] The sliding monitoring time window is set based on the propagation speed of pipeline strain and the physical delay required for event causality verification. It should be short enough to quickly capture anomalies, but long enough to filter out random noise. It is generally set to 10 minutes to 2 hours, with a typical value of 30 minutes. The range can vary from 5 minutes (for pressure-sensitive main pipelines) to 24 hours (for slowly changing suburban pipelines) depending on the importance and dynamic characteristics of the pipeline network.

[0043] The stress gradient change rate is compared with a pre-stored background anomaly threshold. When the stress gradient change rate is greater than or equal to the background anomaly threshold, a spatiotemporal heterogeneous mutation event is determined to have occurred at the pipeline network topology node. The occurrence timestamp, GPS 3D coordinates, stress change direction, and energy change amplitude of the spatiotemporal heterogeneous mutation event are recorded and sequentially written into the cache to generate a spatiotemporal impact sequence table. The spatiotemporal impact sequence table is a list of detailed information for all events determined to be "spatiotemporal heterogeneous mutation events" in chronological order. Among the parameters included in the table, the occurrence timestamp is directly obtained from the sampling time; the GPS 3D coordinates are obtained from the attributes of the pipeline network topology node; the stress change direction is obtained by comparing the stress value at the current time with the stress value at the previous time, with positive values ​​indicating an increase and negative values ​​indicating a decrease; the energy change amplitude is estimated by multiplying the absolute value of the stress change by a preset conversion coefficient, reflecting all spatiotemporal points in the entire pipeline network that require further analysis and their change characteristics.

[0044] The direction of stress change is used to distinguish the type of abnormal event, based on the amount of stress change. The value determination, for example, stress drop is usually related to leakage, while stress increase may be related to blockage or external compression. If, then the direction is increasing; if If, then it decreases, if If the stress is constantly changing, the direction will remain unchanged. Therefore, the stress difference between two nodes is generally positive or negative.

[0045] The amplitude of energy change is used to quantitatively describe the intensity of abnormal events and provides a basis for subsequent risk level classification. This is based on multiplying the stress change by an energy conversion coefficient related to the pipe volume and medium density at that node. First, the absolute value of the stress change is calculated. , Then, the pipe inner diameter r and the length L of the pipe segment are obtained from the pipe material parameters, and the pipe internal volume V is calculated as a preset conversion factor. Finally, the magnitude of the energy change is estimated. , That is, the stress change is multiplied by the volume of the affected fluid to obtain an approximate energy change dimension.

[0046] The background anomaly threshold is set based on the statistical distribution of the historical stress gradient change rate of the pipeline network topology nodes under normal operating conditions (e.g., a historical period with no alarm records in the past 7 to 30 days). For example, it can be set as "historical mean plus 3 standard deviations" or "95th percentile of historical data". The default setting is generally 1.5 to 2 times the maximum normal fluctuation value, with a typical value of 0.05 MPa / minute or a dynamic value calculated based on a large amount of historical data.

[0047] In this embodiment, the implementation steps for identifying the spatiotemporal trajectory chain of cross-domain anomaly characteristics include: Step S403: Along the pipeline topology path of the global spatiotemporal stress dynamic evolution matrix, perform spatiotemporal correlation retrieval on the spatiotemporal heterogeneous mutation events in the spatiotemporal influence sequence table. From the nodes where spatiotemporal heterogeneous mutation events occurred, select the node where a dynamic fluid stress drop event occurred as the starting point, and expand the search outward along the pipeline topology path where that node is located. The search range is a preset radius centered on the starting point, and the search time is within a preset physical causality verification time window. Through spatiotemporal correlation retrieval, in the recorded global spatiotemporal influence sequence table, according to the actual physical connection relationship and temporal order of the pipeline network, search and determine whether there is an inherent, physically explainable causal relationship between multiple heterogeneous mutation events that are spatially adjacent and temporally sequential, thereby filtering out isolated and unrelated noise events.

[0048] When any pipeline topology node is found to trigger a dynamic fluid stress drop event due to a decrease in dynamic fluid stress at a specific time, and within a preset physical causality verification time window, the soil layer within a preset radius around the pipeline topology node experiences a static extrusion stress anomaly fluctuation event within a preset physical time delay, it is determined that there is a cross-domain anomaly feature. The spatiotemporal coordinates, event types, and associated times of the dynamic fluid stress drop event and the static extrusion stress anomaly fluctuation event are spliced ​​together to obtain the spatiotemporal trajectory chain of the cross-domain anomaly feature, and then step S501 is executed.

[0049] If no matching "static compressive stress anomaly fluctuation event" is found within the preset physical causality verification time window, preset radius, and preset physical time delay, it is determined that the sudden drop in dynamic stress is not caused by pipeline leakage, but is highly likely a pseudo-anomaly caused by non-destructive factors such as occasional sensor noise, communication interruption, pump start-up / stop, or sudden changes in user water usage. In this case, instead of generating a cross-domain anomaly characteristic spatiotemporal trajectory chain, the following steps are executed:

[0050] The record of this dynamic fluid stress drop event in the spatiotemporal impact sequence table is marked as an "isolated noise event" and retained in the cache for subsequent statistical analysis and threshold self-learning, but it does not participate in any risk warning process; The search results are logged, and the search window is automatically slid to the next time segment. Then, return to step S402 to continue to monitor the dynamic evolution matrix of spatiotemporal stress across the entire domain and continuously monitor new spatiotemporal heterogeneous mutation events. If such isolated dynamic stress drop events occur frequently at the same node within a short period of time (e.g., within 1 hour), the system will issue a "sensor abnormality or frequent fluctuations in operating conditions" prompt to the operation and maintenance platform, suggesting manual verification of the sensor status or pipeline operation conditions near the node.

[0051] Among them, the pipeline topology path refers to the physical path connecting two or more pipeline nodes in the spatial geometric network of underground pipelines, which is composed of a series of pipe segments and pipeline topology nodes (such as pipes, elbows, tees, valves) connected end to end. It includes the types of all intermediate nodes that constitute the path (such as ordinary nodes, valve nodes), the connection relationship between nodes, the direction of each pipe segment, and the properties of the pipe (such as pipe diameter, material).

[0052] A dynamic fluid stress drop event refers to a significant downward jump in the dynamic fluid stress value of a pipeline network topology node within a very short period of time (usually within 1 to several sampling periods), exceeding the normal fluctuation range. This indicates a sharp and abnormal drop in fluid pressure within the pipeline, and is the most direct and significant hydraulic characteristic signal of pipeline leakage or bursting. Specifically, the dynamic fluid stress gradient change rate of this node is negative, and its absolute value exceeds a high threshold specifically set for the "drop" direction (e.g., this threshold is more than 5 times the average normal drop rate), serving as the starting point of the entire cross-domain anomaly characteristic causal chain. A static compressive stress anomaly event refers to a significant and non-stationary continuous increase or violent oscillation in the static compressive stress value of the soil surrounding a pipeline network topology node acting on the pipeline within a preset physical causal verification time window. This indicates a change in the mechanical state of the soil around the pipeline (such as erosion, a sharp increase in water content, settlement, or displacement), usually a secondary result of underground fluid seepage acting on the soil. The definition is based on the fact that the absolute value of the static extrusion stress gradient change rate of the node exceeds the preset static anomaly threshold, and the change has a certain degree of persistence (e.g., all three consecutive sampling points exceed the limit), which serves as the basis for verifying whether the dynamic stress drop is a real leak rather than other interference.

[0053] The preset physical causality verification time window is set based on the maximum possible physical delay of stress waves (pressure waves) and seepage water propagating in the soil to reach the adjacent area and cause a static stress response. The typical value ranges from 30 seconds to 30 minutes, with a typical value of 5 minutes. The preset radius range is set based on the soil area affected by a single leak point and the node distribution density of the pipeline network. The typical value ranges from 10 meters to 100 meters, with a typical value of 30 meters. The preset physical delay is set based on the shortest possible time required for seepage water to infiltrate the surrounding soil from the leak point and cause a change in static compressive stress measurable by a micro-displacement settlement gauge. The typical value ranges from 10 seconds to 5 minutes, with a typical value of 1 minute. All three parameters can be dynamically adjusted according to pipe material, burial depth, and geological conditions.

[0054] In this embodiment, the implementation steps for real-time comparison of the stress evolution curve with the boundary of the safety envelope constructed based on pipe performance parameters include: Step S501: Extrapolate stress evolution based on the spatiotemporal trajectory chain of cross-domain anomaly characteristics to obtain the stress evolution curve within a preset future time period. First, extract historical stress data of nodes in the spatiotemporal trajectory chain of cross-domain anomaly characteristics that exhibit dynamic fluid stress drop events, forming a historical multivariate time series. Then, perform trend analysis on the latest historical sequence (e.g., data from the past 30 minutes) to calculate the current stress change rate, i.e., the slope. Finally, obtain the stress value of the node at the current moment. Starting from the stress value at the current moment, and using the stress change rate as the step size, accumulate and extrapolate the stress value for each minute (or each sampling interval) in the future until the future time T. Connect all the extrapolated points to obtain the stress evolution curve. The stress evolution curve is a two-dimensional curve depicting the stress change trend from the current moment to the preset future time period (e.g., the next 24 hours), with future time as the horizontal axis and the predicted stress value as the vertical axis. It reflects how the mechanical impact of the currently identified cross-domain anomaly characteristics (e.g., leakage) will evolve in the future if no intervention is taken.

[0055] The pipe deformation limit is extracted from pipe performance parameters, and the soil shear strength limit is extracted from geological properties. These two limits are combined and encapsulated into a dynamic safety envelope boundary. The soil shear strength limit is converted into an equivalent additional pipe stress. The minimum value between the additional pipe stress and the pipe deformation limit is selected as the safety envelope boundary. The dynamic safety envelope boundary is a dynamic threshold curve (or straight line) in a stress-time coordinate system, defining the upper limit (or lower limit, for negative changes) of the stress value considered "safe" at each future prediction time. This includes the permissible stress values ​​calculated based on the pipe deformation limit and the soil shear strength limit for different prediction times. The dynamic safety envelope boundary can be dynamically adjusted; that is, whenever the system obtains the latest geological properties or pipe performance parameters, the baseline value of this boundary can be recalculated and updated.

[0056] Among them, the historical multivariate time series is a dataset containing multiple physical quantities extracted from the source node in the cross-domain anomaly characteristic spatiotemporal trajectory chain and arranged in chronological order. Specifically, it includes the dynamic fluid stress value and static compressive stress value of the node in the past period (such as the past 2 hours). Using the spatial coordinates of the node in the cross-domain anomaly characteristic spatiotemporal trajectory chain as the index, the historical data of a specified length is queried and extracted from the global spatiotemporal stress dynamic evolution matrix in reverse chronological order.

[0057] The rate of stress change refers to the average change in stress per unit time (e.g., per minute), describing how quickly the current stress value increases or decreases. The faster the rate, the more rapidly the risk deteriorates. The rate of stress change is obtained by subtracting the stress value of the previous adjacent timestamp from the current stress value in the extracted historical multivariate time series, and then dividing by the time difference between the current moment and the previous adjacent moment.

[0058] The pipe deformation limit refers to the maximum allowable strain or yield stress intensity that the pipe material can withstand without permanent plastic deformation or fracture, including yield strength and tensile strength. It is obtained from material handbooks or design specifications based on the material type (e.g., ductile iron, PE, steel pipe) and specifications in the pipe performance parameters. The soil shear strength limit refers to the maximum critical stress value that the surrounding soil can withstand against shear failure, including soil cohesion and internal friction angle. It is obtained from soil mechanics parameter tables based on the soil type and compaction state of the corresponding area in the geological physical properties parameters.

[0059] Step S502: Compare the values ​​of the stress evolution curve with the boundary of the dynamic safety envelope point by point; When the value of the stress evolution curve is greater than or equal to the boundary of the dynamic safety envelope at any predicted future time, the triggering condition is met, generating a dual-coordinated early warning signal. The difference between this predicted future time and the current time is defined as the risk countdown. The dual-coordinated early warning signal includes a pipeline hidden leakage risk warning signal and a secondary road collapse risk warning signal. The pipeline hidden leakage risk warning signal accurately locates the specific location of the leak (source coordinates, pipeline number, burial depth) and provides quantitative information on the leak (leakage rate value) to guide engineers in targeted repairs and proactive pressure reduction control of upstream valves. The secondary road collapse risk warning signal is used to delineate dangerous areas (horizontal surface impact areas) that may lead to road collapse due to underground leakage in advance, and provides a risk countdown and cavity evolution rate to guide municipal maintenance departments in taking preventative measures such as surface enclosure, traffic diversion, and cavity backfilling and reinforcement.

[0060] When the stress value at each predicted future moment on the stress evolution curve is strictly less than the threshold at the corresponding moment on the boundary of the dynamic safety envelope, i.e., it has never touched or exceeded the safety boundary, it is determined that the risk corresponding to the currently identified cross-domain anomaly characteristic spatiotemporal trajectory chain is still within a controllable range, and there is no need to immediately trigger the dual collaborative early warning signal. At this time, the following steps are performed: Record the current stress evolution curve and its difference from the safety envelope (e.g., minimum margin value) in the tracking log of the trajectory chain, continue to execute step S501, take the current moment as the new starting point, continuously update the historical multivariate time series, recalculate the stress change rate, and continuously predict the future stress evolution curve.

[0061] No alarms are issued, but continuous monitoring of this trajectory chain is maintained. Once it is found that the gap between the predicted curve and the safety boundary continues to narrow, or the rate of stress change shows an accelerating trend, the system will increase the monitoring frequency (e.g., shorten the sliding time window by half) and generate an internal prompt of "increased risk trend" for the system administrator to review. Only when the predicted stress value first touches or exceeds the safety boundary will the system proceed to the second half of step S502 and output a dual collaborative early warning signal.

[0062] In this embodiment, the steps for generating a hidden leakage risk warning signal in the pipeline network include: When the triggering conditions are met, the source of the cross-domain anomaly characteristic spatiotemporal trajectory chain is traced, and a hidden leakage risk warning signal for the pipeline network is output. The hidden leakage risk warning signal includes the GPS 3D coordinates of the source, the involved pipe number, pipe material, pipe depth, leakage rate value, and the stress gradient reduction at the current moment. The source refers to the location of the first node marked as a "dynamic fluid stress drop event" in the cross-domain anomaly characteristic spatiotemporal trajectory chain, including the node's 3D spatial coordinates (longitude, latitude, burial depth), the attached pipe number, and the pipe material. The GPS 3D coordinates of the source are obtained from the attribute field of this node in the global spatiotemporal stress dynamic evolution matrix, used to accurately locate the leak point. The involved pipe number and pipe material are obtained from the fields bound to this node in the pipe material performance parameters, used to locate specific assets and assess the leakage risk level. The pipe depth is obtained from the burial depth field of this node in the 3D geographic information data, used to guide excavation operations. The leakage rate value is calculated and used to quantify the severity of the leak. The stress gradient reduction at the current moment is obtained from the record of this abrupt event in the spatiotemporal impact sequence table and is used to quickly assess the situation.

[0063] Based on the early warning signal of hidden leakage risk in the pipeline network, the valves of the valve nodes in the nearest known monitoring area grid upstream of the source origin grid are regulated. The early warning signal of hidden leakage risk in the pipeline network is sent to the main control interface of the urban lifeline monitoring terminal for a highlighted pop-up display. At the same time, the system retrieves the smart control valve in the nearest known monitoring area upstream of the source origin coordinates in the underground pipeline network topology and issues a preset percentage valve opening slight reduction control command to this smart control valve to achieve proactive pressure reduction and leakage prevention control of the pipeline network. The preset percentage issued by the smart control valve should effectively slow down the leakage rate at the leak point by reducing upstream pressure and flow to achieve proactive pressure reduction and leakage prevention; however, it should also avoid excessive valve movement that could cause a sudden drop in pipeline pressure, affecting the normal water supply of downstream users or triggering new water hammer risks. It is generally set to a slight reduction of 5% to 20%, with a typical value of 10%.

[0064] The leakage rate value is used to quantitatively describe the amount of water lost from the leak point per unit time. It is an important basis for judging the severity of the leak and adjusting the valve opening. It is calculated based on the physical relationship between the rate of decrease of dynamic fluid stress and pipeline parameters. The calculation formula is as follows: Where Q represents the leakage rate, measured in cubic meters per second. The flow coefficient is selected based on the shape and edge sharpness of the leak hole, ranging from 0.6 to 0.8. In this invention, 0.7 is used by default. It can be obtained from the empirical coefficient table accompanying the pipe material performance parameters, depending on the pipe type (e.g., a larger value for metal pipes and a smaller value for plastic pipes). A represents the estimated leak area, in square meters. This represents the effective pressure difference, i.e., the stress gradient decrease at the current moment, obtained from the record of this abrupt event in the spatiotemporal influence sequence table. This represents the density of the fluid; for urban water supply networks, it refers to the density of the water drawn. For sewage or other special fluids, the data is obtained from the pipe performance parameters or operational data.

[0065] In this embodiment, the steps for generating a secondary road surface collapse risk warning signal include: While generating early warning signals for hidden leakage risks in pipeline networks, the spatial distribution point set of abnormal fluctuation events of static extrusion stress in the spatiotemporal trajectory chain of cross-domain anomaly characteristics is extracted.

[0066] A closed boundary region encompassing the spatially distributed point set is delineated. This closed boundary region is then projected onto the horizontal surface in two dimensions. Each node experiencing an abnormal fluctuation in static compressive stress is designated as a vertex. The smallest convex polygon that can contain all vertices is selected, and this convex polygon is buffered outward by a preset safety distance, such as 5 to 10 meters. This determines the horizontal surface area affected by secondary road collapse. Combining the energy accumulation rate of static compressive stress, the energy accumulation rate is multiplied by the energy-volume conversion coefficient to estimate the evolution rate of soil cavitation, expressed in cubic meters per hour. The evolution rate of soil cavitation reflects the rate at which the volume of underground cavities increases over time.

[0067] The energy accumulation rate of static compressive stress refers to the sum of the increases in all static compressive stress values ​​within the affected area per unit time, used to quantify the energy input rate of soil erosion and cavity development. For each node in the spatial distribution point set, the sequence of its static compressive stress changes over time is obtained. The increase in the latest stress value of each node at the current moment relative to a baseline value (e.g., the average value one hour ago) is calculated. Finally, the increases of all nodes are summed and divided by the time interval to obtain the energy accumulation rate.

[0068] The energy-to-volume conversion coefficient is set based on the soil type of the region in the geological physical parameters. This coefficient represents the volume of soil that can be eroded by a unit of energy, expressed in cubic meters per joule. Different soil types have different value ranges; for example, when the soil type is dense sand, its value range is... to When the soil type is silt or silty sand, the value range is: to Generally, the energy-to-volume conversion factor is obtained from geotechnical engineering handbooks or local long-term monitoring data for each soil type. For miscellaneous fill or silty clay where urban underground pipe networks are located, the value is taken as... .

[0069] The GPS coordinates of the inflection points of the horizontal surface impact area, the risk countdown, and the evolution rate are standardized and encapsulated to generate a secondary road subsidence risk warning signal. This warning signal is then sent to the municipal maintenance terminal, where the horizontal surface impact area is displayed with differentiated color rendering on a digital twin map, completing proactive risk collaborative warning. Specifically, the GPS coordinates of the inflection points of the horizontal surface impact area in the secondary road subsidence risk warning signal are obtained from the calculated vertex coordinates of the smallest convex polygon; the risk countdown is obtained from the difference between the future prediction time calculated in step S502 and the current time.

[0070] In the color rendering display, on a digital twin map system, such as a City Information Modeling (CIM) platform, a closed vector surface is drawn on the map layer based on the coordinates of the inflection points of the horizontally affected land area. The color of this filled surface is dynamically set according to the risk countdown (e.g., 24 hours, 12 hours, 6 hours) and the rate of evolution (fast or slow). For example, yellow is used when the countdown is long and the rate is slow; red is used when the countdown is short and the rate is fast. This vector surface is overlaid on the map base and rendered with semi-transparent filling and a highlighted border, providing municipal maintenance personnel with intuitive, vivid, and easy-to-understand visual alerts, enabling them to quickly locate risk areas and assess risk levels.

[0071] In this embodiment, the entire technical solution requires at least the following hardware devices in actual use:

[0072] IoT sensors are deployed within a known monitoring grid to collect real-world multimodal time-series data. Data acquisition units (RTUs / DTUs) and industrial gateways are deployed to aggregate sensor signals and upload the data to a central server via 4G / 5G or wired networks.

[0073] High-performance computing servers are used to run core algorithms such as stress extrapolation, matrix construction, and trend prediction; database servers are used to store 3D GIS data, geological physical parameters, pipe performance parameters, and various generated time-series matrix data.

[0074] Urban lifeline monitoring terminals (usually high-performance workstations or multi-screen display systems, used to display warning signals in a high-brightness pop-up window on the main control interface) and municipal maintenance terminals (can be fixed PCs, tablets, or large-screen display systems, used to display digital twin maps and risk area renderings).

[0075] Intelligent control valves, along with their matching electric actuators and valve controllers, are installed at key locations in the pipeline network to receive commands and perform valve opening regulation.

[0076] In this embodiment, the entire technical solution, in its actual implementation, requires at least the following technical and software architecture: Three-dimensional Geographic Information System (GIS) technology: used to store, manage, query and visualize the spatial geometric network, node coordinates and attribute information of urban underground pipelines; implemented based on existing computer software platforms such as ArcGIS, SuperMap or open source software such as QGIS, PostGIS, etc., this application utilizes them for grid division, topological path retrieval and spatial distance calculation.

[0077] Time-series data processing and stream computing technology: used to process high-frequency sampled multimodal sensor data and construct and update the spatiotemporal stress field flow matrix; it can be implemented based on real-time stream processing frameworks such as Apache Kafka (message queue), Apache Flink or Spark Streaming. This solution uses them to perform sliding window calculation, state management and time-series alignment on the data.

[0078] Database and data warehouse technology: used to store massive amounts of historical time-series matrix data, parameter data, and early warning event records. Based on databases such as TimeScaleDB (time-series database), InfluxDB, or traditional PostgreSQL, this solution utilizes them for efficient data writing, querying, and time-series analysis.

[0079] Numerical computation and spatial analysis engine: Used to perform core algorithms such as stress equivalent transformation, attenuation extrapolation, trend extrapolation, and convex hull calculation. It can be implemented using computing environments such as Python (NumPy, SciPy, Shapely libraries), MATLAB, or Java (JTS topology suite). This solution utilizes these environments for all nonlinear, custom physical model calculations.

[0080] Digital twin and visualization technology: used to visually display early warning results (such as horizontal surface impact areas) in the form of 2D / 3D map rendering. It can be implemented based on digital twin platforms such as Unity 3D, Unreal Engine, CesiumJS, or ThingJS, or WebGL graphics libraries. This solution utilizes them for differentiated color rendering and highlight pop-ups for risk areas.

[0081] Industrial automation control technology: used to issue control commands to remote intelligent valves. It can be implemented through SCADA (Supervisory and Data Acquisition) system or PLC (Programmable Logic Controller) programming based on industrial communication protocols such as Modbus and OPC UA. This solution utilizes it to perform active pressure reduction and leakage prevention control.

[0082] In this embodiment, using three-dimensional geographic information data of urban underground pipe networks, the pipe network space is divided into a known monitoring area grid nested with multiple pipe network topology nodes and a spatiotemporal monitoring blind zone grid. By calculating the spatial topological impedance coefficient between the known and blind zone grid nodes, and combining the energy dissipation coefficient of the underground medium characterized by soil saturation and density, a rigorous non-uniform conduction mapping of the underground medium is established. By adaptively attenuating the stress values ​​of the known monitoring area grid nodes along the complex physical topology path of the pipe network, the virtual stress estimation values ​​of each pipe network topology node in the spatiotemporal monitoring blind zone grid are reconstructed with high precision without relying on physical sensors. These values ​​are then reverse-interpreted and converted into a virtual sensor data stream for the blind zone, enabling data mining to have spatial continuity. This fundamentally overcomes the technical bottleneck of the inability to perceive hidden dangers in blind zones, realizes proactive perception and early warning in sensorless areas, eliminates monitoring blind zones, and provides comprehensive spatiotemporal data support for proactive prevention of sensorless coverage segments of urban lifelines.

[0083] This invention performs equivalent transformation on the collected instantaneous pressure time-series data and vertical micro-displacement time-series data, fusing them into a spatiotemporal stress field flow matrix of the known monitoring area that includes spatial adjacency relationships. This achieves a unified multi-physics field characterization of fluid dynamic stress and solid static compressive stress in the spatiotemporal grid dimension. Based on fluid stress changes and causal verification, it captures the causal time-series relationship between "dynamic fluid stress drop events" and "static compressive stress anomaly fluctuation events" in the surrounding soil along the topological path, accurately identifying the spatiotemporal trajectory chain of cross-domain anomaly characteristics. Utilizing the physical transmission mechanism between fluid and solid, data processing is performed, with a clear and verifiable logical chain and high interpretability. It can accurately identify and automatically filter out occasional noise caused by sensor hardware failures and false exceedance noise caused by severe weather. By comparing the trend extrapolation prediction curve with the safety envelope boundary constructed based on pipe material performance and geological properties point by point, it achieves dual collaborative proactive risk output for leakage and collapse, effectively reducing the false alarm rate and missed alarm rate, and improving the early warning accuracy under complex operating conditions of urban lifelines.

[0084] This application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable code, which, when executed by the one or more processors, can perform the spatiotemporal data mining and risk warning method for urban lifeline underground pipeline networks as described above.

[0085] The method or system according to the embodiments of this application can also be implemented using the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or hard disk, may store the spatiotemporal data mining and risk warning method for underground pipeline networks for urban lifelines provided in this application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in this application is merely exemplary; when implementing different devices, one or more components in the electronic device shown in this application may be omitted according to actual needs.

[0086] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0087] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines, characterized in that, The method includes: Acquire pre-stored three-dimensional geographic information data, geological property parameters, and pipe material performance parameters of urban underground pipe networks, and divide the underground pipe network into known monitoring area grids and spatiotemporal monitoring blind area grids. Each grid contains multiple pipe network topology nodes. Collect multimodal time-series data within the known monitoring area grid, and construct the spatiotemporal stress field flow matrix of the known monitoring area; Calculate the spatial topological impedance coefficient between the known monitoring area grid and the spatiotemporal monitoring blind zone grid, and perform attenuation extrapolation by combining geological physical parameters and spatiotemporal stress field flow matrix. Calculate the virtual stress extrapolation value of each pipeline topology node in the spatiotemporal monitoring blind zone grid and convert it into a blind zone virtual sensor data stream. The spatiotemporal stress field flow matrix is ​​merged with the data stream from the virtual sensor in the blind zone to generate a global spatiotemporal stress dynamic evolution matrix. The spatiotemporal influence sequence table is extracted, and the spatiotemporal trajectory chain of cross-domain anomaly characteristics is identified. The spatiotemporal trajectory chain of the cross-domain anomaly features is predicted by forward time-series evolution to generate a stress evolution curve. The stress evolution curve is compared in real time with the safety envelope boundary constructed based on the pipe performance parameters. When the triggering condition is met, a dual collaborative early warning signal is output.

2. The method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines as described in claim 1, characterized in that, The process involves dividing the underground pipeline network into a known monitoring area grid and a spatiotemporal monitoring blind area grid, and collecting multimodal time-series data within the known monitoring area grid, including: Acquire pre-stored 3D geographic information data of urban underground pipe network, analyze the spatial geometric network of the entire city's pipelines, valve nodes and the installation location of IoT sensors, divide the underground pipe network into multiple grids, and classify the grids into known monitoring area grids and spatiotemporal monitoring blind area grids according to whether IoT sensors are deployed; Urban surveying control points are introduced as a unified three-dimensional rectangular coordinate system reference to obtain the installation spatial coordinates of IoT sensors and align them with the static spatial reference of pipe performance parameters, including pipe diameter, pipe material parameters and elevation information. All IoT sensors within the known monitoring area grid acquire multimodal time-series data using a unified timestamp reference and sampling frequency. The multimodal time-series data includes instantaneous pressure time-series data, instantaneous flow time-series data, and vertical micro-displacement time-series data.

3. The method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines as described in claim 2, characterized in that, The construction of the spatiotemporal stress field flow matrix of the known monitoring area includes: By combining the pipe cross-sectional area parameters with the stress equivalent transformation of the instantaneous pressure time series data, the dynamic fluid stress vector borne by the inner wall of the pipe is extracted. The soil elastic modulus corresponding to the known monitoring area grid is extracted from the geological physical parameters. The vertical micro-displacement time series data is multiplied and transformed with the soil elastic modulus to obtain the static compressive stress vector of the pipeline outer wall caused by uneven soil settlement. The dynamic fluid stress vector and static extrusion stress vector are used as the dynamic features of the corresponding pipeline topology nodes. Combined with the pipe performance parameters, the three-dimensional spatial coordinates are used as the index key to reorganize the mesh adjacency and construct the spatiotemporal stress field flow matrix of the known monitoring area.

4. The method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines as described in claim 1, characterized in that, The calculation of the spatial topological impedance coefficient between the known monitoring area grid and the spatiotemporal monitoring blind zone grid includes: Using the pipeline topology nodes of the spatiotemporal monitoring blind zone grid to be simulated as the target endpoint and the pipeline topology nodes of the adjacent known monitoring area grid as the search starting point, all connected paths between the two are retrieved in the three-dimensional geographic information data. Extract the pipe connectivity parameters in the connected path, calculate the kinetic energy loss ratio of the fluid during transmission in the connected path based on the pipe connectivity parameters, and use the kinetic energy loss ratio as the spatial topological impedance coefficient.

5. The method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines according to claim 1, characterized in that, The virtual stress projection values ​​of each pipeline topology node in the spatiotemporal monitoring blind zone grid are calculated and converted into a blind zone virtual sensor data stream, including: Obtain the known coupling state values ​​of each pipeline topology node in the adjacent known monitoring area grid, and calculate the three-dimensional spatial straight-line distance between a specific pipeline topology node in the spatiotemporal monitoring blind zone grid to be simulated and the pipeline topology node in the known monitoring area grid. The soil saturation and soil density of the adjacent area are extracted from the geological physical parameters. The known coupling state value and spatial topological impedance coefficient are combined with the soil saturation, soil density and three-dimensional spatial straight distance to perform numerical attenuation calculation to obtain the virtual stress estimation value of a specific pipeline topology node in the spatiotemporal monitoring blind zone grid. The virtual stress projection values ​​are interpreted in reverse to obtain a virtual sensor data stream, which includes virtual pressure time series data and virtual instantaneous flow data.

6. The method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines according to claim 1, characterized in that, The step of merging the spatiotemporal stress field flow matrix with the blind zone virtual sensor data stream to generate a global spatiotemporal stress dynamic evolution matrix and extracting a spatiotemporal influence sequence table includes: The spatiotemporal stress field flow matrix of the known monitoring area and the virtual sensor data flow of the blind area are combined with three-dimensional rectangular coordinates to spatially reorganize and generate a global spatiotemporal stress dynamic evolution matrix. Set a sliding monitoring time window and continuously calculate the rate of change of stress gradient between adjacent sampling times for any pipeline topology node in the global spatiotemporal stress dynamic evolution matrix; The stress gradient change rate is compared with the pre-stored background anomaly threshold. When the stress gradient change rate is greater than or equal to the background anomaly threshold, it is determined that a spatiotemporal heterogeneous mutation event has occurred at the pipeline topology node. The timestamp of the spatiotemporal heterogeneous mutation event, GPS three-dimensional coordinates, stress change direction and energy change amplitude are recorded and written into the cache to generate a spatiotemporal impact sequence table.

7. The method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines as described in claim 6, characterized in that, The identification of the spatiotemporal trajectory chain of cross-domain anomaly features includes: Along the pipeline topology path of the global spatiotemporal stress dynamic evolution matrix, spatiotemporal correlation retrieval is performed on the spatiotemporal heterogeneous mutation events in the spatiotemporal influence sequence table; When any pipeline topology node is found to trigger a dynamic fluid stress drop event due to a decrease in dynamic fluid stress at a specific time, and within a preset physical causality verification time window, the soil layer within a preset radius around the pipeline topology node experiences an abnormal fluctuation event in static compressive stress within a preset physical time delay, it is determined that there is a cross-domain anomaly characteristic. The spatiotemporal coordinates, event types, and associated times of the dynamic fluid stress drop event and the abnormal fluctuation event in static compressive stress are spliced ​​together to obtain the spatiotemporal trajectory chain of the cross-domain anomaly characteristic.

8. The method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines according to claim 7, characterized in that, The real-time comparison of the stress evolution curve with the boundary of the safety envelope constructed based on pipe performance parameters includes: Stress evolution extrapolation is performed based on the spatiotemporal trajectory chain of cross-domain anomaly characteristics to obtain the stress evolution curve within a future preset time period; Extract the pipe deformation limit from the pipe performance parameters, extract the soil shear strength limit from the geological properties parameters, and combine the pipe deformation limit and the soil shear strength limit into a dynamic safety envelope boundary. The value of the stress evolution curve is compared point by point with the boundary of the dynamic safety envelope. When the value of the stress evolution curve is greater than or equal to the boundary of the dynamic safety envelope at any predicted time in the future, the triggering condition is determined to be met, and a dual-coordinated early warning signal is generated. The difference between the predicted time in the future and the current time is defined as the risk countdown. The dual-coordinated early warning signal includes a pipeline hidden leakage risk early warning signal and a secondary road collapse risk early warning signal.

9. The method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines according to claim 1, characterized in that, The steps for generating the hidden leakage risk warning signal in the pipeline network include: When the triggering condition is met, the source and starting point of the spatiotemporal trajectory chain of the cross-domain anomaly characteristics are traced, and a warning signal of hidden leakage risk in the pipeline network is output. Based on the early warning signal of hidden leakage risk in the pipeline network, the valves of the valve nodes of the nearest known monitoring area grid upstream of the grid where the source origin is located are regulated; The hidden leakage risk warning signal of the pipeline network includes the GPS three-dimensional coordinates of the source and starting point, the pipeline number involved, the pipe material, the pipeline depth, the leakage rate value, and the stress gradient reduction at the current moment.

10. The method for spatiotemporal data mining and risk early warning of underground pipeline networks for urban lifelines according to claim 8, characterized in that, The steps for generating the secondary road collapse risk warning signal include: While generating early warning signals for hidden leakage risks in the pipeline network, the spatial distribution point set of abnormal fluctuation events of static extrusion stress in the spatiotemporal trajectory chain of cross-domain anomaly features is extracted. The boundary closed region covering the spatially distributed point set was delineated, the horizontal surface impact area affected by secondary road collapse was determined, and the evolution rate of soil cavitation was estimated by combining the energy accumulation rate of static compressive stress. The GPS coordinates of the inflection points, risk countdown, and evolution rate of the horizontal surface impact area are standardized and encapsulated to generate a secondary road collapse risk warning signal.

Images