Modulus grading and classifying efficient organization and management method for railway full life cycle

Abstract

The invention provides a modulus grading and classification efficient organization and management method for a railway full life cycle, and the method comprises the steps: obtaining three-dimensional model data corresponding to a target railway scene, and carrying out the fusion construction of a geographic space fusion expression model; jointly querying real-time monitoring data on the basis of timestamps and spatial position codes; obtaining model feature data associated with each three-dimensional model; updating a knowledge graph corresponding to the target railway scene based on the model feature data and the real-time monitoring data, constructing graph structure data, inputting the graph structure data into a graph neural network for risk mining, determining risk nodes, and indicating the risk nodes in a geographic space fusion expression model; and matching the risk node with a preset normative rule, generating an emergency management scheme, performing quantitative evaluation through a decision tree model, and determining a final management scheme. The problem of low organization and management efficiency in the railway infrastructure management process can be solved, and the organization and management efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of railway engineering management technology, and in particular to a modular hierarchical classification and efficient organization and management method for the entire life cycle of railways. Background Technology

[0002] The modular hierarchical classification and efficient organization and management method for the entire railway lifecycle refers to a method for standardized organization and collaborative management of model data and business information of railway infrastructure in the planning, design, construction, operation, and decommissioning stages by constructing a unified modular classification standard, component hierarchical system, and classification coding rules. This method can provide basic data support and organizational assurance for various business scenarios such as spatial modeling of railway facilities, component positioning, operation and maintenance scheduling, equipment management, and intelligent control. After acquiring the 3D model or business data of railway infrastructure, technologies such as model aggregation, semantic disambiguation, coding standardization, and component classification archiving are typically used to ensure data consistency and accessibility across multiple stages and systems. Currently, forward modeling, reverse modeling, and various acquisition methods such as UAVs, laser scanning, and track trolleys are widely used for model data acquisition and have achieved preliminary integration in some projects.

[0003] However, existing railway model organization techniques typically rely on manual component identification, classification, and data archiving in specialized modeling software. Meanwhile, mainstream model data update methods also largely depend on remodeling and reclassification, lacking a unified modular organization structure and component indexing mechanism. This fails to meet the needs of efficient, unified, and dynamic management throughout the entire lifecycle of railway infrastructure. In the process of railway infrastructure management, there are problems such as inconsistent modular division, chaotic component classification, low organizational management efficiency, and data inconsistency across stages, making it difficult to share model information, which leads to difficulties in railway infrastructure management. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a modular hierarchical classification and efficient organization and management method for the entire life cycle of railways, to eliminate or improve one or more defects existing in the prior art. It can solve the problems of low organizational management efficiency caused by inconsistent modular division, chaotic component classification, inconsistent cross-stage data, and difficulty in sharing model information during railway infrastructure management.

[0005] One aspect of the present invention provides a modular hierarchical classification and efficient organization and management method for the entire life cycle of railways, the method comprising the following steps: Acquire the 3D model data corresponding to the target railway scene, fuse and construct a multi-scale geospatial fusion representation model and display it; the 3D model data includes geographic information system data and building information model data; Based on the current timestamp and the spatial location code corresponding to each 3D model, a joint query is performed to obtain and display the real-time monitoring data associated with each 3D model; the real-time monitoring data includes time-series data or equipment status data collected in real time by each monitoring device in the target railway scene; Obtain the model feature data associated with each 3D model; the model feature data includes image feature data or text feature data; the image feature data is obtained by target recognition and feature extraction from image data collected in the target railway scene; the text feature data is obtained by entity recognition and semantic vectorization processing from the operation and maintenance record text corresponding to each 3D model. The knowledge graph corresponding to the target railway scenario is updated based on model feature data and real-time monitoring data; the knowledge graph includes monitoring equipment entities, building facility entities, and the relationships between these entities. Entities in the knowledge graph are converted into nodes, and the relationships between entities are converted into edges between nodes to construct graph structure data. The graph structure data is then input into a pre-trained graph neural network for risk mining, to identify risk nodes, and to indicate risk nodes in the geospatial fusion representation model. Risk nodes are matched with preset normative rules to generate emergency management plans. These plans are then quantitatively evaluated using a pre-trained decision tree model to determine the final management plan.

[0006] In some embodiments of the present invention, the three-dimensional model data is model data under a unified coordinate system; the geographic information system data also includes digital elevation model data corresponding to the terrain surface model of the target railway scene; The model integrates and constructs a multi-scale geospatial fusion representation model, including: For each building information model, determine the geometry of the model's base, as well as the elevation values ​​and Z-axis coordinates of the vertices in the geometry in a unified coordinate system; Adjust the Z-axis coordinates of the vertex so that the elevation value of the vertex is consistent with the elevation value of the corresponding position of the building information model on the terrain surface model; Identify the boundary area between the building information model and the terrain surface model, and delete the triangulation control points of the terrain surface model located in the boundary area; The local triangulation is reconstructed based on the intersection vertices between the Building Information Model (BIM) and the terrain surface model to achieve seamless integration of the BIM and the terrain surface model, resulting in a geospatial fusion representation model.

[0007] In some embodiments of the present invention, the three-dimensional model data further includes oblique photogrammetry model data corresponding to various building and facility entities in the target railway scene; the fusion and construction of a multi-scale geospatial fusion representation model includes: Traverse the coordinates of building vertices in the oblique photogrammetry model and perform geometric corrections to ensure that the oblique photogrammetry model and the corresponding building information model remain spatially consistent. By dynamically rendering and adjusting the height information of the oblique photogrammetry model using the GPU vertex shader, real-time fusion and visual alignment with the corresponding building information model plot boundaries are achieved.

[0008] In some embodiments of the present invention, the three-dimensional model data is pre-stored in a model database; before obtaining the three-dimensional model data corresponding to the target railway scene, the method further includes: A seven-parameter Bursa model was used to perform a three-dimensional coordinate system transformation on each three-dimensional model, transforming it to the geocentric coordinate system; Based on the actual location information, professional type and component type of each 3D model in the target railway scene, a spatial location code corresponding to each 3D model is generated. Each 3D model is associated with its corresponding spatial location code and stored in the model database.

[0009] In some embodiments of the present invention, before jointly querying and obtaining the real-time monitoring data associated with each 3D model based on the current timestamp and the spatial location code corresponding to each 3D model, the method further includes: Interpolation methods are used to resample time-series data collected by different monitoring devices at different frequencies to the same frequency; According to the preset time window, the time series data after unifying the frequency are divided into time series segments of fixed length; The dynamic time warping algorithm is used to perform time synchronization calibration on each time segment to obtain time-aligned time series data.

[0010] In some embodiments of the present invention, before jointly querying and obtaining the real-time monitoring data associated with each 3D model based on the current timestamp and the spatial location code corresponding to each 3D model, the method further includes: Determine the data missing rate of the time-series data collected by each monitoring device within a preset time window; For time-series data with a missing rate less than a preset missing rate threshold, the Kalman filter algorithm is used for interpolation repair.

[0011] In some embodiments of the present invention, the real-time monitoring data further includes railway operation data corresponding to the target railway scene; before obtaining the real-time monitoring data associated with each three-dimensional model through joint query based on the spatial location codes corresponding to each three-dimensional model, the method further includes: The real-time monitoring data is processed to ensure a standardized format. The real-time monitoring data, after being standardized in format, is associated with the spatial location code of its data source and stored in a preset database. The preset database includes a document database for storing railway operation data and equipment status data, and a time-series database for storing time-series data.

[0012] In some embodiments of the present invention, the image data includes image data collected by image sensors deployed in the target railway scene, inspection image data collected by drones, and satellite image data collected by satellite platforms.

[0013] In some embodiments of the present invention, the inspection image data includes railway track image data corresponding to the railway track in the target railway scene; the image feature data includes railway track image feature data corresponding to the railway track image data. After obtaining the model feature data associated with each 3D model, the following is also included: Based on railway track image feature data, the structural state of key components in the railway track is determined and indicated in the geospatial fusion representation model.

[0014] In some embodiments of the present invention, the satellite image data includes building and facility image data corresponding to the target building in the target railway scene; before obtaining the model feature data associated with each 3D model, the method further includes: Obtain a pre-trained building facility evaluation model; By inputting building facility image data into the building facility assessment model, the structural assessment value corresponding to the target building can be obtained; The structural assessment value is compared with a preset anomaly threshold. If the structural assessment value is greater than or equal to the preset anomaly threshold, a prompt message is generated and indicated in the geospatial fusion expression model.

[0015] This invention presents a modular hierarchical classification and efficient organization management method for the entire railway lifecycle. It addresses the low organizational management efficiency caused by inconsistent modular divisions, chaotic component classification, discontinuous cross-stage data, and difficulties in sharing model information during railway infrastructure management. Combining the integrated needs of railway engineering in planning, design, construction, and operation and maintenance stages for multi-dimensional model and business data organization and management, it establishes spatial location coding rules based on a pre-set database, component-level information modeling, a hierarchical classification rule system, and modular coding standards. This constructs a bidirectional association mechanism between the 3D model and data, achieving unique identification and precise binding between the 3D model and data. Furthermore, it introduces high-precision timestamps to record data changes, enabling time-dimensional traceability and supporting efficient... The system integrates source tracing and data reverse lookup with automatic change monitoring to ensure the consistency and integrity of multi-source data in terms of space, attributes, and time. This provides efficient, unified, and intelligent organizational management methods for facility data organization, component retrieval, and operation and maintenance control throughout the entire railway lifecycle. Simultaneously, it constructs a multi-source heterogeneous data model for railway operations based on a knowledge graph, integrating real-time monitoring data, historical fault records, and standardized rules to achieve a structured expression of multi-dimensional information. Through graph neural network technology, it performs deep learning and feature propagation on nodes and relationships in the knowledge graph, uncovering potential correlations and risks. This assists in generating precise safety control and emergency dispatch plans, and combines decision tree algorithms to quantitatively evaluate emergency strategies, improving the scientific nature and efficiency of emergency response.

[0016] In addition, the 3D model is standardized according to the industrial basic class standard and the railway attributes are extended. The geographic information system data is standardized to GeoJSON or TopoJSON format, and the integrated model and digital data is standardized to JSON or CSV format. This achieves standardization of multi-source data formats and semantics, which can effectively integrate heterogeneous data and support the integrated management and intelligent application of railway spatial information.

[0017] In addition, for sensor data of different frequencies, a dynamic time warping algorithm is adopted to achieve precise synchronous calibration of multi-source time series data using a preset time window, which can effectively improve the accuracy of time series alignment. At the same time, the Kalman filter algorithm is combined to interpolate and repair data with a missing rate lower than a preset missing rate threshold to ensure data continuity and integrity. In practical applications, it can accurately identify the correlation between multi-source monitoring data and improve the reliability and timeliness of data fusion.

[0018] In addition, pixel-level semantic segmentation of inspection image data is performed using deep learning models to achieve automatic identification and localization of multi-category targets. Image features are extracted through convolutional neural networks, which can improve the accuracy and robustness of target detection. At the same time, entity recognition and semantic vectorization of operation and maintenance record text are performed using natural language processing models to realize the association between text information and 3D models, enhance the semantic fusion capability of multi-source heterogeneous data, and thus improve the overall intelligent analysis level of data.

[0019] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the description and drawings.

[0020] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, are not intended to limit the scope of the invention. In the drawings: Figure 1 A flowchart of a modular hierarchical classification and efficient organization and management method for the entire life cycle of railways, provided as an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram illustrating the data categorized by professional type and equipment type according to an embodiment of the present invention.

[0023] Figure 3 Another flowchart of a modular hierarchical classification and efficient organization and management method for the entire life cycle of railways, provided as an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0025] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0026] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0027] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.

[0028] In the following description, embodiments of the invention will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.

[0029] The following is a detailed introduction to the modular hierarchical classification and efficient organization and management method for the entire life cycle of railways provided in this application.

[0030] Optionally, the implementing entity of the modular hierarchical classification and efficient organization and management method for the entire life cycle of railways provided in this application is an electronic device. The electronic device can be a terminal such as a computer, mobile phone, or tablet computer, or it can be a server. This embodiment does not limit the implementation method of the electronic device.

[0031] This embodiment provides a modular hierarchical classification and efficient organization and management method for the entire life cycle of railways, such as... Figure 1 As shown, the method includes at least the following steps: Step S101: Obtain the 3D model data corresponding to the target railway scene, fuse and construct a multi-scale geospatial fusion expression model and display it.

[0032] The target railway scenario refers to a specific railway infrastructure area that requires 3D modeling, data fusion, status monitoring, or visualization management, including but not limited to key geographical and engineering units such as stations, bridges, tunnels, track sections, signal equipment areas, substations, and dispatch centers.

[0033] Three-dimensional model data refers to the digital representation obtained after geometric modeling and attribute description of the building facilities entities (such as buildings, tracks, bridges, vehicles, and equipment) in the target railway scene.

[0034] The 3D model data is model data under a unified coordinate system. After geometrically modeling and describing the attributes of the building facilities entities in the target railway scene to obtain the 3D model data, in order to ensure spatial consistency between the geographic information system data and the building information model data, it is also necessary to perform coordinate system transformation on the 3D model data, converting the 3D model data under different coordinate systems to a unified coordinate system.

[0035] In some embodiments of the present invention, the three-dimensional model data includes Geographic Information System (GIS) data and Building Information Modeling (BIM) model data. Accordingly, the unified coordinate system can be the GIS coordinate system to which the GIS data belongs or the BIM design coordinate system to which the BIM model belongs.

[0036] Based on this, before acquiring the 3D model data corresponding to the target railway scene, coordinate system transformation is performed on the 3D model data to convert the 3D model data under different coordinate systems to a unified coordinate system. This includes: converting the geographic information system data to the building information model design coordinate system to which the building information model data belongs, so as to make the space between the geographic information system data and the building information model data consistent; or, converting the model data of the building information model to the geographic information system coordinate system to which the geographic information system data belongs, so as to make the space between the geographic information system data and the building information model data consistent.

[0037] Alternatively, the unified coordinate system can also be a three-dimensional geocentric coordinate system (World Geodetic System 1984, WGS84). When the unified coordinate system is WGS84, a seven-parameter transformation method is used to convert GIS data and BIM data to the WGS84 coordinate system, achieving high-precision alignment between different spatial coordinate systems and ensuring spatial consistency between GIS data and BIM data. The seven parameters include the X-axis translation, Y-axis translation, and Z-axis translation between the origins of the two coordinate systems; the rotation angles (Euler angles) along the X-axis, Y-axis, and Z-axis; and the scale factor between the two coordinate systems.

[0038] Specifically, before obtaining the 3D model data corresponding to the target railway scene, the process includes: using the seven-parameter Bursa-Wolf Model to perform a 3D coordinate system transformation on each 3D model, transforming it to the geocentric coordinate system; generating a spatial location code corresponding to each 3D model based on the actual location information, professional type, and component type of each 3D model in the target railway scene; and associating and storing each 3D model with its corresponding spatial location code in the model database.

[0039] In some embodiments of this invention, the model database is built upon a document database, enabling unified storage and efficient retrieval of various types of 3D models. By employing the CouchDB document database to store 3D model metadata and binary files in JSON format, combined with professional type classification and logical storage area design, it supports the categorized management of various types of 3D models. Simultaneously, the construction of a full-text index and design document view enables keyword fuzzy search and customized retrieval, significantly improving the efficiency of 3D model retrieval to meet the application needs of complex railway engineering projects.

[0040] In spatial location coding, actual location information refers to the geographical location of the building entity represented by the 3D model in the real world. For example, 3D coordinates (X, Y, Z) or latitude, longitude, and altitude (Lon, Lat, Alt) information in the geocentric coordinate system.

[0041] Specialty types are used to distinguish different business systems, indicating the professional field or system type to which the building facility entity represented by the 3D model belongs. Specialty types include, but are not limited to, engineering, power supply, communication, signaling, land use, building construction, passenger service, or water supply and drainage types.

[0042] Railway engineering includes infrastructure such as railway lines, bridges, and tunnels. Power supply includes power supply systems, such as overhead contact lines and substations. Communication includes railway communication network facilities, such as communication towers and fiber optic cables. Signaling includes railway train control systems, signal lights, switches, and transponders. Land use includes railway land and station layouts. Buildings include station buildings and dispatching towers. Passenger services include passenger service facilities such as restaurants and public restrooms. Water supply and drainage includes water supply and drainage facilities.

[0043] In some embodiments of the present invention, different professional type codes correspond to different professional types. The professional type code is used to uniquely indicate the professional type and can be numbers, letters, or a combination of numbers and letters. This embodiment does not limit the implementation method of the professional type code.

[0044] Component type refers to the specific facility type, equipment type, or sub-component type corresponding to the 3D model. Among them, component types include, but are not limited to, track type, signal type, building type, bridge type, etc.

[0045] Track types include, but are not limited to, turnouts, curved sections, straight sections, or sleepers. Signal types include, but are not limited to, traffic lights, transponders, and track circuits. Building types include, but are not limited to, station buildings, control rooms, or power distribution rooms. Bridge types include, but are not limited to, main beams, supports, or piers.

[0046] In some embodiments of the present invention, different components have different component codes. The component type code is used to uniquely indicate the component type and can be numbers, letters, or a combination of numbers and letters. This embodiment does not limit the implementation method of the device type code.

[0047] Specifically, spatial location coding can be represented as: [Professional Type Code]_[3D Coordinate Code or Latitude / Longitude Code]_[Component Type Code] Alternatively, the actual location information can also be the linear location information of the building entity represented by the 3D model within the railway line, including absolute coordinates or offset distance relative to the starting point of the railway line. The linear location information uses the mileage representation method of railway engineering, such as KXXXYYY, where KXXX represents kilometers and YYY represents meters.

[0048] Correspondingly, spatial location coding can also be represented as: [Professional Type Code]_[Mileage Information Code]_[Component Type Code] Furthermore, to effectively integrate heterogeneous data and achieve integrated management and intelligent application of data within the target railway scenario, it is also necessary to standardize the data format of 3D model data. This involves integrating 3D model data from different sources to reduce complexity and redundancy caused by differences in data formats, thereby indirectly promoting data lightweighting. Simultaneously, using a unified standardized format makes it easier to perform automated processing, analysis, and visualization operations on 3D model data. This is particularly important for long-term monitoring and maintenance planning, providing comprehensive support for subsequent decision-making.

[0049] Specifically, for modeling information model data, the unified Industry Foundation Classes (IFC) standard is used to standardize the data format and extend railway attributes.

[0050] The methods for expanding railway attributes include, but are not limited to: Railway-specific views can be defined using Model View Definition (MVD); or railway-specific attributes can be added using Property Set (Pset) or Quantity Take-off (Qto); or railway facility parameters (such as track type, gauge, power supply method, etc.) can be defined through the Engineering Basic Class Schema Extension Mechanism (IFC Schema), Construction Operations Building Information Exchange (COBie), CityGML bridge extension, etc.

[0051] By converting the building information model (BIM) to the IFC standard, interoperability between different software can be ensured, allowing the BIM to be used and shared across various tools and platforms. Furthermore, for railway projects, extending specific railway attributes can better adapt the BIM to industry needs.

[0052] For geographic information system (GIS) data, a unified geographic JavaScript Object Notation (GeoJSON) or topology JSON (TopoJSON) is used for data format standardization to simplify the storage and exchange of GIS data and facilitate seamless integration between various GIS platforms.

[0053] Specifically, the process involves acquiring 3D model data corresponding to the target railway scene, fusing and constructing a multi-scale geospatial fusion representation model, and displaying it. This includes: extracting geographic information system (GIS) data and building information model (BIM) data from the model database; visualizing the GIS data and BIM data based on a unified spatial reference coordinate system; and fusing the GIS data and BIM data to obtain the geospatial fusion representation model.

[0054] In practical implementation, a multi-level detail (LOD) strategy can be combined to generate model versions with different precision to adapt to different rendering needs. During the geospatial fusion representation model display process, a layered loading mechanism is adopted to achieve multi-scale, high-performance visualization. Simultaneously, view-first layered loading and GPU-accelerated rendering technologies are utilized to improve model loading efficiency and interactive smoothness, achieving efficient visualization of large-scale 3D models.

[0055] In other embodiments of the present invention, the geographic information system data also includes digital elevation model (DEM) data corresponding to the terrain surface model of the target railway scene. The DEM data can be data acquired through satellite remote sensing, aerial photography (e.g., drone aerial photography), or data acquired using lidar technology, etc. This embodiment does not limit the acquisition method of the DEM data.

[0056] In some embodiments of this invention, a precise fusion of the Building Information Model (BIM) and the terrain surface model is achieved based on elevation adaptation and triangulation reconstruction methods. By extracting the elevation of the BIM's bottom surface and performing linear interpolation adjustment with the terrain surface model, the coordinates in the vertical direction (Z-axis direction) are unified, ensuring that the bottom surface of the BIM fits the surface of the terrain surface model. Then, the intersection line between the BIM and the terrain surface model is calculated, the terrain triangulation control points within the intersection line are deleted, and the local triangulation is reconstructed using the intersection line vertices to eliminate penetration conflicts. This effectively solves the spatial mismatch problem between the BIM and the complex terrain surface model, improving simulation accuracy and visual continuity.

[0057] Specifically, the fusion construction of a multi-scale geospatial fusion representation model also includes: for each building information model, determining the geometry of the model's base surface, as well as the vertex elevation values ​​and Z-axis coordinates of the vertices in the geometry in a unified coordinate system; adjusting the vertex Z-axis coordinates so that the vertex elevation values ​​are consistent with the elevation values ​​of the corresponding positions of the building information model and the terrain surface model; identifying the boundary area between the building information model and the terrain surface model, and deleting the triangulation control points of the terrain surface model located in the boundary area; and reconstructing the local triangulation based on the intersection vertices between the building information model and the terrain surface model to achieve seamless fusion of the building information model and the terrain surface model, thereby obtaining the geospatial fusion representation model.

[0058] In some embodiments of the present invention, the 3D model data also includes oblique photogrammetry model data corresponding to each building and facility entity in the target railway scene. The oblique photogrammetry model data is obtained by a drone equipped with an oblique photogrammetry camera.

[0059] In some embodiments of this invention, Building Information Modeling (BIM) data and oblique photogrammetry model data are fused. By iterating and modifying the coordinates of each vertex of the oblique photogrammetry model, accurate geometric fitting with the BIM model is achieved, ensuring spatial alignment and seamless texture integration. Simultaneously, for large-scale plots, the display height of the oblique photogrammetry model is dynamically adjusted using GPU vertex shaders, achieving real-time synchronous fusion with the plot boundaries of the BIM model. This improves the matching accuracy and visual consistency of multi-source 3D models, making it more suitable for scenarios such as railway station simulation and tunnel visualization.

[0060] Specifically, the integration and construction of a multi-scale geospatial fusion representation model also includes: traversing the building vertex coordinates in the oblique photogrammetry model and performing geometric corrections to ensure that the oblique photogrammetry model and the corresponding building information model maintain spatial consistency; and dynamically rendering and adjusting the height information of the oblique photogrammetry model through the GPU vertex shader to achieve real-time fusion and visual alignment with the plot boundaries of the corresponding building information model.

[0061] In practice, the 3D model data can also be LiDAR point cloud data, computer-aided design (CAD) models, or voxel models, etc. This embodiment does not limit the type of 3D model data.

[0062] In practice, the building information model can also be geometrically simplified based on the edge folding algorithm, removing redundant triangular faces while keeping the key structures unchanged, thereby significantly reducing the number of faces and volume of the model.

[0063] Step S102: Based on the current timestamp and the spatial location code corresponding to each 3D model, perform a joint query to obtain and display the real-time monitoring data associated with each 3D model.

[0064] The current timestamp refers to the current time identifier obtained through the system timestamp (e.g., Date.now()), used to synchronize the time dimension of real-time monitoring data. Real-time monitoring data includes time-series data or equipment status data collected in real-time by various monitoring devices in the target railway scenario.

[0065] Among them, time-series data refers to dynamic monitoring data that changes continuously over time and is collected by monitoring equipment, including but not limited to the track surface temperature curve collected by the track temperature sensor (such as recording the temperature value once per minute); the time sequence of track vibration amplitude changes over time monitored by the accelerometer; and the time series of environmental parameters such as wind speed and rainfall collected by the meteorological station.

[0066] Equipment status data refers to the static or discrete status information of the monitoring equipment at the current timestamp, including but not limited to the current status of the signal (red / green / yellow light on), the opening direction of the turnout (left / right / straight), or whether rail cracks are detected by the flaw detector (Boolean value or fault code), etc.

[0067] By obtaining the current time, the system ensures that data collected by all monitoring devices is queried under the same time reference. Simultaneously, it utilizes pre-assigned spatial location codes within the 3D model to quickly locate the associated positions of the monitoring devices within the model. Using the current timestamp as a filter, the system extracts real-time time-series data collected by the monitoring devices and the device status data of each device from a pre-set database. This data is then linked to the 3D model using spatial location codes. The locations of the monitoring devices are highlighted in the 3D model, and real-time monitoring data (e.g., temperature values ​​presented as a color-coded heatmap) is overlaid. For abnormal status data (such as overheating or switch malfunctions), early warning information is generated and output (e.g., pop-up text messages, flashing icons, or warning sounds) to alert the user.

[0068] For example, taking a 3D model of a railway bridge as an example; the railway bridge is equipped with temperature sensors, vibration sensors, and waterproof gates; the temperature sensors output real-time time-series data including temperature values ​​and corresponding timestamps; the vibration sensors output real-time time-series data including amplitudes and corresponding timestamps; the waterproof gates output real-time equipment status data including open or closed status; the system uses the current timestamp and the spatial location code of the railway bridge to jointly query from a preset database to obtain the real-time temperature value output by the temperature sensors, the real-time amplitude output by the vibration sensors, and the status of the waterproof gates at the current timestamp, and displays them in the model.

[0069] In some embodiments of the present invention, the time-series data collected by the monitoring device and the corresponding device status data are stored in a preset database. The preset database includes a document database (e.g., a MongoDB document database) for storing device status data and a time-series database (InfluxDB time-series database) for storing time-series data. For the time-series data collected by the monitoring device, the time-series database is used to achieve efficient time window queries. For the device status data of the monitoring device, a modular and digital integrated model is managed through the document database, while also supporting dynamic field expansion to meet the flexible storage requirements of structured and semi-structured data.

[0070] In practice, to balance storage costs and data read speeds, a hot and cold partitioning mechanism can be used to deploy frequently accessed data in high-speed storage devices (such as solid-state drives) to ensure low-latency response, while low-frequency accessed data is archived in low-frequency storage devices (such as hard disk drives or tape libraries) to optimize storage resource allocation and costs.

[0071] Specifically, after the monitoring equipment collects the time-series data corresponding to the current moment, the pre-assigned spatial location code of the monitoring equipment, the timestamp corresponding to the current moment, and the time-series data collected at the current moment are associated and stored in a preset database. Alternatively, after the monitoring equipment status is updated, the pre-assigned spatial location code of the monitoring equipment, the timestamp corresponding to the moment of status change, and the updated equipment status data are associated and stored in a document database.

[0072] Among them, the pre-assigned spatial location code of the monitoring equipment refers to the identifier used to uniquely indicate the monitoring equipment, which is generated based on the professional type, equipment type and location information of the building facility entity to which each monitoring equipment belongs in the target railway scenario.

[0073] The professional type is used to distinguish different business systems, indicating the professional field or system type to which the building facility entity represented by the 3D model belongs. Professional types include, but are not limited to, engineering, power supply, communication, signaling, land use, building construction, passenger service, or water supply and drainage types.

[0074] Railway engineering includes infrastructure such as railway lines, bridges, and tunnels. Power supply includes power supply systems, such as overhead contact lines and substations. Communication includes railway communication network facilities, such as communication towers and fiber optic cables. Signaling includes railway train control systems, signal lights, switches, and transponders. Land use includes railway land and station layouts. Buildings include station buildings and dispatching towers. Passenger services include passenger service facilities such as restaurants and public restrooms. Water supply and drainage includes water supply and drainage facilities.

[0075] Equipment type is used to distinguish the monitoring functions of different monitoring devices, including but not limited to temperature sensors, vibration sensors, displacement sensors, current sensors, etc.

[0076] In some embodiments of the present invention, different device codes correspond to different device types. The device type code is used to uniquely indicate the device type and can be numbers, letters, or a combination of numbers and letters. This embodiment does not limit the implementation method of the device type code.

[0077] Specifically, the spatial location code of the monitoring equipment can be represented as: [Professional Type Code]_[Mileage Information Code]_[Equipment Type Code] In practical implementation, spatial location coding may also include other codes, such as company codes, line codes, railway bureau codes, station codes, asset codes, or equipment model numbers. This embodiment does not limit the implementation method of spatial location coding.

[0078] Among them, the company code is a unique identifier used to identify the company to which the monitoring equipment or building facility belongs, supporting cross-company asset management and facilitating group-wide data integration and access control; the line code is a unique identifier used to identify linear projects such as railways and highways; the railway bureau code is used to identify the railway bureau to which the asset belongs in the railway system, enabling regional management; the station / section code is used to identify a specific station or section; the asset code is a unique asset identifier for each monitoring equipment or building facility, ensuring the uniqueness and traceability of assets and supporting full lifecycle management; and the equipment model number is used to record the specific model and specifications of the equipment, facilitating accurate matching during maintenance and replacement and reducing compatibility issues.

[0079] Furthermore, in order to effectively integrate heterogeneous data and achieve integrated management and intelligent application of data in the target railway scenario, it is also necessary to standardize the data format of real-time monitoring data before storing it in the preset database. This involves integrating data from different monitoring devices to reduce complexity and redundancy caused by differences in data formats, thereby indirectly promoting data lightweighting.

[0080] In some embodiments of the present invention, real-time monitoring data is standardized using a unified JSON or CSV format to simplify data storage and exchange.

[0081] Specifically, before obtaining the real-time monitoring data associated with each 3D model through joint query based on the spatial location codes corresponding to each 3D model, the process also includes: unifying the format of the real-time monitoring data; and associating the unified real-time monitoring data with the spatial location codes of the data sources to be stored in a preset database.

[0082] In practice, data from different monitoring devices often have different sampling frequencies and time offsets. For example, a vibration sensor may sample at 100Hz, while a temperature sensor may only update at 1Hz. This asynchronous characteristic can cause multi-source time-series data to be misaligned in the time dimension, thus affecting the accuracy of data analysis and status judgment.

[0083] To address the aforementioned issues, in this embodiment, after the time-series data acquired by the monitoring device is collected, a Dynamic Time Warping (DTW) algorithm is also required to perform non-linear time-series alignment of the multi-source sensor data. By finding the optimal matching path within a preset time window (including but not limited to 10ms, 20ms, or 30ms), precise synchronization calibration between data of different frequencies is achieved, thereby eliminating time deviations, improving data alignment accuracy, and providing a high-quality, spatiotemporally consistent data foundation for subsequent intelligent analysis, anomaly detection, and fusion visualization.

[0084] Specifically, before obtaining the real-time monitoring data associated with each 3D model through joint query based on the current timestamp and the spatial location code corresponding to each 3D model, the process includes: resampling time-series data collected by different monitoring devices at different frequencies to the same frequency using interpolation methods; dividing the time-series data after unifying the frequency into time-series segments of fixed length according to a preset time window; and performing time synchronization calibration on each time-series segment using a dynamic time warping algorithm to obtain time-aligned time-series data.

[0085] In addition, to address the issue of partial missing data in multi-source sensor data, when the data missing rate is less than 15%, the Kalman filter algorithm is used to dynamically interpolate and repair the missing segments. This can effectively fill the gaps while preserving the data trend characteristics, ensuring the continuity and integrity of time-series data.

[0086] Specifically, before obtaining the real-time monitoring data associated with each 3D model through joint query based on the current timestamp and the spatial location code corresponding to each 3D model, the process also includes: determining the data missing rate of the time series data collected by each monitoring device within a preset time window; and using the Kalman filter algorithm for interpolation repair for time series data with a data missing rate less than a preset missing rate threshold.

[0087] Furthermore, in some embodiments of the present invention, a three-level modeling method of "entity-table-field" is adopted in the preset database to achieve hierarchical management and efficient integration of multi-source heterogeneous data in the target railway scenario. Specifically, the entity layer defines core objects and associates them with spatial location codes; the table layer constructs attribute tables and associates them with entities; and the field layer refines specific data items and supports dynamic expansion. Simultaneously, a data platform engine based on SQL-on-Hadoop supports cross-database joint queries to ensure rapid response to multi-source heterogeneous data, significantly improving the speed of data location and analysis, thereby enhancing overall business response speed and decision-making accuracy.

[0088] In practice, the document database stores not only structured data collected by various monitoring devices, but also unstructured railway operation data, such as maintenance records, user data, and equipment technical specifications. The default database also includes a document database for storing railway operation data.

[0089] like Figure 2 As shown, in some embodiments of the present invention, structured data (including real-time monitoring data, equipment status data, etc.) and unstructured data (including technical history, documents or engineering archives, such as engineering documents, as-built drawings, maintenance and repair records, technical specifications, professional regulations, etc.) are classified according to professional types, and then further divided according to equipment level. The equipment level includes unit level, component level, asset level, equipment level, etc. The data of each level will be recorded in the corresponding ledger (such as fixed asset cards, ledger record tables, etc.), thereby constructing a master data ontology model.

[0090] The master data ontology model is the core of the entire data management system, encompassing information on various professional asset data, including fixed assets, asset consolidation, transfer, and liquidation. Within this model, data is subdivided into multiple parts, such as fixed assets, equipment ledgers, technical documents, and engineering archives. These parts are connected through relationships and compositional relationships to ensure data consistency and integrity.

[0091] Step S103: Obtain the model feature data associated with each 3D model.

[0092] In some embodiments of the present invention, the model feature data includes image feature data or text feature data.

[0093] Among them, the image feature data is obtained by target recognition and feature extraction from image data collected in the target railway scene.

[0094] In some embodiments of the present invention, a pre-trained target recognition model is used to perform pixel-level semantic segmentation on image data to achieve automatic recognition and localization of multi-category targets and obtain target recognition results; then, a pre-trained feature extraction model is used to extract image features to obtain image feature data, thereby improving the accuracy and robustness of target detection.

[0095] Specifically, target recognition and feature extraction are performed on image data collected in the target railway scene, including: inputting the collected image data into a pre-trained target recognition model to obtain target recognition results corresponding to the image data, wherein the target recognition results include the identified targets and their corresponding image patches; and inputting the target recognition results into a pre-trained feature extraction model to obtain image feature data corresponding to each target. The target recognition model is trained using training data on a pre-set deep learning model; the feature extraction model is trained using training data on a pre-set neural network model.

[0096] In some embodiments of the present invention, the training data includes sample image data, sample target recognition results corresponding to the sample image data, and sample image features.

[0097] Accordingly, the training process of the target recognition model includes: inputting sample image data into a preset deep learning model to obtain training results; inputting the training results and sample target recognition results into a loss function to obtain a loss result; training the deep learning model based on the loss result to reduce the difference between the training results and the corresponding sample target recognition results, until the deep learning model converges to obtain the target recognition model.

[0098] The deep learning model can be a general object detection network architecture, such as Faster R-CNN or YOLO series. This embodiment does not limit the implementation method of the deep learning model.

[0099] The training process of the feature extraction model includes: inputting the output of the target recognition model into a preset neural network model to obtain the training result; inputting the training result and sample image features into a loss function to obtain the loss result; training the neural network model based on the loss result to reduce the difference between the training result and the corresponding sample image features until the neural network model converges to obtain the feature extraction model.

[0100] The neural network model can be a convolutional neural network (CNN), a recursive neural network (RNN), or a feedforward neural network (FNN). This embodiment does not limit the implementation method of the neural network model.

[0101] In some embodiments of the present invention, the image data includes image data collected by image sensors deployed within the target railway scene, inspection image data collected by drones, and satellite image data collected by satellite platforms, and is stored in a pre-defined data lake. This pre-defined data lake is built on Apache Hudi (Hadoop Upserts Deletes and Incrementals) to enable large-scale archiving and long-term management.

[0102] refer to Figure 3 The system retrieves 3D model data corresponding to the target railway scene from a pre-defined model library and basic imagery data from a data lake. The acquired 3D model data is optimized using lightweight processing techniques to reduce data volume and improve processing efficiency. Then, a layered loading mechanism is employed, loading different levels of data progressively as needed to ensure efficient operation during rendering and dynamic updating of the geospatial fusion representation model. The lightweight-processed and layered-loading data is used to construct the geospatial fusion representation model, which is then rendered and dynamically updated. Simultaneously, structured and unstructured data are tagged with spatial location encoding and timestamps to facilitate association and retrieval with the corresponding 3D models, enabling accurate positioning and display of the data in 3D space. Finally, all processed data (including the geospatial fusion representation model and data associated with the 3D models) is integrated and provided to users for data retrieval and visualization. Users can view, analyze, and manipulate data through this process, achieving a comprehensive understanding and management of complex systems such as railway infrastructure.

[0103] In some embodiments of the present invention, the inspection image data includes railway track image data corresponding to the railway track in the target railway scene, covering key components of the track structure, such as rails, ballast, fasteners, sleepers, and turnout areas. Correspondingly, the image feature data includes railway track image feature data extracted based on the railway track image data. These features are obtained through deep learning models or traditional image processing algorithms and can reflect key information such as the geometric shape, texture changes, and surface damage of track components, providing data support for subsequent structural condition identification.

[0104] Specifically, after obtaining the model feature data associated with each 3D model, the process also includes: determining the structural state of key components in the railway track based on railway track image feature data and indicating it in the geospatial fusion expression model.

[0105] The structural status includes, but is not limited to, indicators such as wear level, crack distribution, offset, and loosening status. Subsequently, the identified structural status information is spatially bound to the corresponding 3D model, and visualized and dynamically indicated in the geospatial fusion representation model. For example, through color coding, flashing prompts, or pop-up displays, the operating status and health level of each component are intuitively displayed, thereby achieving real-time perception and intelligent early warning of the railway track facility status.

[0106] In practice, the inspection image data can also be collected by vehicle-mounted cameras or trackside vision acquisition devices. This embodiment does not limit the scope and method of collecting inspection image data.

[0107] In some embodiments of the present invention, the satellite imagery data includes imagery data of the building structures corresponding to the target building in the target railway scene. The satellite imagery data is typically acquired by high-resolution satellite sensors, covering the entire building structure and its surrounding environment, providing fundamental information for accurately assessing the condition of the building structure.

[0108] Specifically, before acquiring the model feature data associated with each 3D model, the process also includes: acquiring a pre-trained building facility evaluation model; inputting building facility image data into the building facility evaluation model to obtain the structural evaluation value corresponding to the target building; comparing the structural evaluation value with a preset anomaly threshold, and generating a prompt message and indicating it in the geospatial fusion expression model if the structural evaluation value is greater than or equal to the preset anomaly threshold.

[0109] The building facilities assessment model is obtained by training a pre-set neural network model using training data.

[0110] In some embodiments of the present invention, the training data includes sample satellite image data and corresponding sample evaluation values.

[0111] Accordingly, the training process of the building facility assessment model includes: inputting sample satellite image data into a preset neural network model to obtain training results; inputting the training results and sample evaluation values ​​into a loss function to obtain loss results; training the neural network model based on the loss results to reduce the difference between the training results and the corresponding sample evaluation values ​​until the neural network model converges to obtain the building facility assessment model.

[0112] The neural network model can be a convolutional neural network (CNN), a recursive neural network (RNN), or a feedforward neural network (FNN). This embodiment does not limit the implementation method of the neural network model.

[0113] In some embodiments of the present invention, the text feature data is obtained by performing entity recognition and semantic vectorization processing on the operation and maintenance record text corresponding to each 3D model. Specifically, entity recognition and semantic vectorization are performed on the operation and maintenance record text using a Natural Language Processing (NLP) model to realize the association between text information and 3D models, thereby enhancing the semantic fusion capability of multi-source heterogeneous data and improving the overall intelligent analysis level of the data.

[0114] Step S104: Update the knowledge graph corresponding to the target railway scenario based on model feature data and real-time monitoring data.

[0115] The knowledge graph includes monitoring equipment entities, building facility entities, and the relationships between these entities.

[0116] In some embodiments of the present invention, a multi-source heterogeneous data model corresponding to the target railway environment is constructed based on a knowledge graph, and real-time monitoring data, operation and maintenance record text and normative rules are integrated to realize the structured expression of multi-dimensional information.

[0117] Through the construction and updating mechanism of knowledge graph, multi-source heterogeneous data is integrated. The multi-source heterogeneous data includes real-time monitoring data (such as track displacement, vibration frequency, temperature changes, etc. collected by monitoring equipment), operation and maintenance record text (such as maintenance logs, fault reports, etc.), and normative rules (such as the "Railway Engineering Design Specification" and "Equipment Maintenance Standard").

[0118] In updating the knowledge graph, an incremental learning strategy is adopted. By comparing the differences between old and new data (such as changes in equipment status or rule revisions), only the affected entities and their relationships are updated, avoiding the performance loss caused by a full reconstruction of the knowledge graph. Simultaneously, a confidence weighting mechanism is introduced to assign dynamic weights to data from different sources (e.g., sensor data has a higher weight than manually recorded data) to ensure the reliability of the knowledge graph. Furthermore, implicit relationships (e.g., "abnormal track displacement may lead to switch misalignment") are established through a semantic reasoning engine (such as the OWL ontology), expanding the reasoning capabilities of the knowledge graph.

[0119] In some embodiments of the present invention, the knowledge graph adopts a layered modeling approach, including an infrastructure layer (describing physical entities such as railway tracks, bridges, and tunnels and their attributes, including materials, construction dates, etc.), a dynamic monitoring layer (associating real-time monitoring data with physical entities), and a rule constraint layer (embedding normative rules as logical constraints).

[0120] Step S105: Convert entities in the knowledge graph into nodes, convert the relationships between entities into edges between nodes, construct graph structure data, and input the graph structure data into a pre-trained graph neural network for risk mining, identify risk nodes, and indicate risk nodes in the geospatial fusion expression model.

[0121] This involves converting entities in the knowledge graph into nodes, including converting equipment entities into equipment nodes (e.g., converting track sensor entities into track sensor nodes, signal light entities into signal light nodes, etc.) and building facility entities into building facility entity nodes (e.g., converting bridge entities into bridge nodes, catenary support entities into catenary support nodes, etc.).

[0122] Subsequently, a pre-trained graph neural network model (such as GraphSAGE, GAT, GCN) is used to input graph structure data and combine it with node features (such as device status values ​​and historical failure frequencies) for feature propagation and aggregation. Node embeddings are calculated iteratively through multi-layer GNNs to capture potential relationships between nodes. High-risk nodes are identified based on clustering analysis (such as DBSCAN) or threshold determination (such as the Euclidean distance between the node embedding and the normal pattern being greater than a preset value). Risk nodes are then visualized and annotated in a 3D geospatial fusion model (e.g., marked with flashing red markers), and a pop-up window displays the risk level, associated equipment, and recommended handling measures.

[0123] In practical implementation, the attention mechanism of GNNs can be used to analyze edge weights and locate key risk propagation paths (such as "bridge settlement → track displacement → train derailment"). This way, after a user clicks on a risk node, its relational chain in the knowledge graph can be displayed simultaneously to help the user quickly locate the root cause of the problem.

[0124] Step S106: Match the risk nodes with preset normative rules to generate an emergency management plan and perform quantitative evaluation through a pre-trained decision tree model to determine the final management plan.

[0125] Specifically, the key attributes of the risk nodes (such as equipment type, abnormal parameters, and scope of impact) are first matched with the logical conditions in the normative rule base (such as "speed must be limited immediately if track displacement exceeds 5mm") to generate a preliminary emergency management plan.

[0126] This process is implemented through a rule parsing engine, which supports two modes: exact matching and fuzzy matching. When a risk node fully matches a certain rule, the corresponding handling measures are directly output (such as "replace worn parts"). When the matching degree is low, approximate rule suggestions are generated through similarity calculation (such as cosine similarity or Euclidean distance) (such as "it is recommended to observe and inspect once per hour").

[0127] Subsequently, the emergency management plan is input into a pre-trained decision tree model for multi-dimensional quantitative evaluation. The decision tree model is pre-trained based on historical decision data (such as "the percentage decrease in failure rate after equipment replacement" and "the economic cost of speed-limited operation") and is used to comprehensively analyze key indicators such as the priority of response measures, resource requirements (such as manpower and spare parts inventory), expected risk reduction (such as the percentage decrease in failure probability), and execution time window.

[0128] Finally, a structured emergency management plan is generated based on the quantitative assessment results, including specific handling actions (such as "dispatching a maintenance team to track section K12+345"), timeline planning (such as "completing equipment replacement within 1 hour"), and resource allocation (such as "calling up maintenance car X and spare sleeper Y").

[0129] Furthermore, during the implementation of the plan, real-time monitoring data from the target railway scenario and model feature data corresponding to each 3D model are collected in real time to monitor the progress of the response (e.g., "track displacement returned to normal range after replacement"), and the equipment status and risk node markers in the knowledge graph are automatically updated. If the response effect does not meet expectations (e.g., "displacement is still deteriorating after replacement"), secondary risk mining and plan iteration are triggered, forming a closed-loop management of the entire process from risk identification to response feedback, significantly improving the scientific nature and dynamic adaptability of railway emergency response.

[0130] In summary, the modular hierarchical classification and efficient organization and management method for the entire railway lifecycle provided in this embodiment can solve the problems of low organizational management efficiency caused by inconsistent modular division, chaotic component hierarchical classification, data inconsistency across stages, and difficulty in sharing model information in the process of railway infrastructure management. Combining the integrated needs of multi-dimensional model and business data organization and management in railway engineering at the planning, design, construction, and operation and maintenance stages, it formulates spatial location coding rules based on component-level information modeling, hierarchical classification rule system, and modular coding standards. It constructs a bidirectional association mechanism between 3D models and data, achieving unique identification and precise binding of 3D models and data, and introduces high-precision timestamps to record data changes, enabling time dimension traceability and supporting high... It enables efficient source tracing and data reverse lookup, integrates automatic change monitoring functions, and ensures the consistency and integrity of multi-source data in terms of space, attributes, and time. This provides efficient, unified, and intelligent organizational management methods for facility data organization, component retrieval, and operation and maintenance control throughout the entire railway lifecycle. Simultaneously, it constructs a multi-source heterogeneous data model for railway operations based on knowledge graphs, integrating real-time monitoring data, historical fault records, and normative rules to achieve structured expression of multi-dimensional information. Through graph neural network technology, it performs deep learning and feature propagation on nodes and relationships in the knowledge graph, uncovering potential correlations and risks, assisting in the generation of precise safety control and emergency dispatch plans, and combining decision tree algorithms to quantitatively evaluate emergency strategies, improving the scientific nature and efficiency of emergency response.

[0131] In addition, the 3D model is standardized according to the industrial basic class standard and the railway attributes are extended. The geographic information system data is standardized to GeoJSON or TopoJSON format, and the integrated model and digital data is standardized to JSON or CSV format. This achieves standardization of multi-source data formats and semantics, which can effectively integrate heterogeneous data and support the integrated management and intelligent application of railway spatial information.

[0132] In addition, for sensor data of different frequencies, a dynamic time warping algorithm is adopted to achieve precise synchronous calibration of multi-source time series data using a preset time window, which can effectively improve the accuracy of time series alignment. At the same time, the Kalman filter algorithm is combined to interpolate and repair data with a missing rate lower than a preset missing rate threshold to ensure data continuity and integrity. In practical applications, it can accurately identify the correlation between multi-source monitoring data and improve the reliability and timeliness of data fusion.

[0133] In addition, pixel-level semantic segmentation of inspection image data is performed using deep learning models to achieve automatic identification and localization of multi-category targets. Image features are extracted through convolutional neural networks, which can improve the accuracy and robustness of target detection. At the same time, entity recognition and semantic vectorization of operation and maintenance record text are performed using natural language processing models to realize the association between text information and 3D models, enhance the semantic fusion capability of multi-source heterogeneous data, and thus improve the overall intelligent analysis level of data.

[0134] Corresponding to the above method, the present invention also provides a modular hierarchical classification and efficient organization and management device for the entire life cycle of railways. The device includes a computer device, which includes a processor and a memory. The memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the device implements the steps of the modular hierarchical classification and efficient organization and management method for the entire life cycle of railways as described above.

[0135] This invention also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the steps of the aforementioned modular hierarchical classification and efficient organization and management method for the entire lifecycle of railways. The computer-readable storage medium can be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disks, removable storage disks, CD-ROMs, or any other form of storage medium known in the art.

[0136] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0137] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0138] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0139] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A modular hierarchical classification and efficient organization and management method for the entire life cycle of railways, characterized in that, The method includes the following steps: Acquire the 3D model data corresponding to the target railway scene, fuse and construct a multi-scale geospatial fusion representation model and display it; wherein, the 3D model data includes geographic information system data and building information model data; Based on the current timestamp and the spatial location code corresponding to each 3D model, a joint query is performed to obtain and display the real-time monitoring data associated with each 3D model; the real-time monitoring data includes time-series data or equipment status data collected in real time by each monitoring device in the target railway scene; The model feature data associated with each 3D model is obtained; the model feature data includes image feature data or text feature data; the image feature data is obtained by target recognition and feature extraction of image data collected in the target railway scene; the text feature data is obtained by entity recognition and semantic vectorization processing of the operation and maintenance record text corresponding to each 3D model. The knowledge graph corresponding to the target railway scenario is updated based on the model feature data and the real-time monitoring data; the knowledge graph includes monitoring equipment entities, building facility entities, and the relationships between these entities. The entities in the knowledge graph are converted into nodes, and the relationships between entities are converted into edges between nodes to construct graph structure data. The graph structure data is then input into a pre-trained graph neural network for risk mining, risk nodes are identified, and the risk nodes are indicated in the geospatial fusion expression model. The risk nodes are matched with preset normative rules to generate emergency management plans. These plans are then quantitatively evaluated using a pre-trained decision tree model to determine the final management plan.

2. The method according to claim 1, characterized in that, The three-dimensional model data is model data under a unified coordinate system; the geographic information system data includes digital elevation model data corresponding to the terrain surface model of the target railway scene; The fusion constructs a multi-scale geospatial fusion representation model, including: For each building information model, determine the geometry of the model's bottom surface, as well as the elevation values ​​and Z-axis coordinates of the vertices in the geometry in the unified coordinate system; Adjust the Z-axis coordinate of the vertex so that the elevation value of the vertex is consistent with the elevation value of the building information model at the corresponding position on the terrain surface model; Identify the boundary area between the building information model and the terrain surface model, and delete the triangulation control points of the terrain surface model located in the boundary area; Based on the intersection vertices between the building information model and the terrain surface model, a local triangulation network is reconstructed to seamlessly integrate the building information model and the terrain surface model, thereby obtaining the geospatial fusion representation model.

3. The method according to claim 1, characterized in that, The 3D model data also includes oblique photogrammetry model data corresponding to each building and facility entity in the target railway scene; The fusion constructs a multi-scale geospatial fusion representation model, including: The coordinates of the building vertices in the oblique photogrammetry model are traversed and geometric corrections are performed to ensure that the oblique photogrammetry model and the corresponding building information model remain spatially consistent. The height information of the oblique photogrammetry model is dynamically rendered and adjusted by the GPU vertex shader to achieve real-time fusion and visual alignment with the corresponding building information model plot boundary.

4. The method according to claim 1, characterized in that, The 3D model data is pre-stored in a model database; before obtaining the 3D model data corresponding to the target railway scene, the process also includes: A seven-parameter Bursa model was used to perform a three-dimensional coordinate system transformation on each three-dimensional model, transforming it to the geocentric coordinate system; Based on the actual location information, professional type and component type of each 3D model in the target railway scene, a spatial location code corresponding to each 3D model is generated. Each 3D model is associated with its corresponding spatial location code and stored in the model database.

5. The method according to claim 1, characterized in that, Before obtaining the real-time monitoring data associated with each 3D model through joint querying based on the current timestamp and the spatial location code corresponding to each 3D model, the process also includes: Interpolation methods are used to resample time-series data collected by different monitoring devices at different frequencies to the same frequency; According to the preset time window, the time series data after unifying the frequency are divided into time series segments of fixed length; The dynamic time warping algorithm is used to perform time synchronization calibration on each time segment to obtain time-aligned time series data.

6. The method according to claim 1, characterized in that, Before obtaining the real-time monitoring data associated with each 3D model through joint querying based on the current timestamp and the spatial location code corresponding to each 3D model, the process also includes: Determine the data missing rate of the time-series data collected by each monitoring device within a preset time window; For time-series data with a missing rate less than a preset missing rate threshold, the Kalman filter algorithm is used for interpolation repair.

7. The method according to claim 1, characterized in that, The real-time monitoring data also includes railway operation data corresponding to the target railway scenario; before obtaining the real-time monitoring data associated with each 3D model through joint query based on the spatial location codes corresponding to each 3D model, the process further includes: The real-time monitoring data is processed to unify its format; The real-time monitoring data, after being standardized in format, is associated with the spatial location code of its data source and stored in a preset database; the preset database includes a document database for storing the railway operation data and the equipment status data, and a time-series database for storing the time-series data.

8. The method according to claim 1, characterized in that, The image data includes image data collected by image sensors deployed in the target railway scene, inspection image data collected by drones, and satellite image data collected by satellite platforms.

9. The method according to claim 8, characterized in that, The inspection image data includes railway track image data corresponding to the railway track in the target railway scene; the image feature data includes railway track image feature data corresponding to the railway track image data; After obtaining the model feature data associated with each 3D model, the process also includes: Based on the railway track image feature data, the structural state of key components in the railway track is determined and indicated in the geospatial fusion representation model.

10. The method according to claim 8, characterized in that, The satellite imagery data includes building and facility imagery data corresponding to the target building in the target railway scene; before acquiring the model feature data associated with each 3D model, the process also includes: Obtain a pre-trained building facility evaluation model; The building facility image data is input into the building facility evaluation model to obtain the structural evaluation value corresponding to the target building; The structural evaluation value is compared with a preset anomaly threshold. If the structural evaluation value is greater than or equal to the preset anomaly threshold, a prompt message is generated and indicated in the geospatial fusion expression model.

Images