Rail transit infrastructure intelligent operation and maintenance method and system based on digital twinborn technology

Abstract

The invention relates to a rail transit infrastructure intelligent operation and maintenance method and system based on a digital twinborn technology. According to the method, a finite element and multi-body dynamics joint simulation method is adopted, a vehicle-road coupling calculation model is established, and the calculation result of the vehicle-road coupling calculation model and real-time data of a sensor are integrated to form a disease database; based on the dynamic characteristics in the disease database, training to obtain a disease damage identification model; secondly, digital twinning modeling is conducted on the rail transit infrastructure, then a rail transit digital twinning intelligent operation and maintenance front end is constructed, and in the operation and maintenance process, the operation and maintenance front end sends a disease prediction request to the disease damage identification model; and after the request is received, the model reads real-time data of the sensor in the disease database, performs damage identification to generate a real-time prediction result, and returns the real-time prediction result to the operation and maintenance front end. Compared with the prior art, the method has the advantages of being stable, high in adaptability, good in prediction capability, high in working efficiency and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of rail transit, and in particular, to an intelligent operation and maintenance method and system for rail transit infrastructure based on digital twin technology. Background Art

[0002] As an important part of urban public transportation, the safe and efficient operation of rail transit is crucial for the stability of the urban transportation system. The spatial span of urban rail transit infrastructure is large, there are many management objects, and the occurrence and distribution of diseases are random. The complexity of its operation and maintenance increases with the continuous expansion of traffic demand and scale. In practice, rail transit operation entities usually conduct preventive maintenance activities on track infrastructure regularly, usually every day. This operation and maintenance method usually relies on regular planned inspections and manual maintenance, with low efficiency and easy omission of potential structural problems.

[0003] As an emerging technical means, digital twin realizes real-time interaction and data synchronization between the physical and virtual worlds by constructing a virtual digital model of physical infrastructure, providing strong support for equipment operation status monitoring, fault prediction, and operation and maintenance optimization.

[0004] Currently, the infrastructure visualization platform based on digital twin focuses on information display and data query, but is insufficient in intelligent fault diagnosis, risk warning, and decision support, and is difficult to meet the actual needs of intelligent operation and maintenance. There have been research on machine learning algorithms in the fields of airborne equipment fault prediction, track facility fault prediction, train scheduling optimization, etc., but these algorithms lack transferable interfaces and cannot be integrated into the information management system to play a wide range of autonomous decision-making functions. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned defects existing in the prior art and provide an intelligent operation and maintenance method and system for rail transit infrastructure based on digital twin technology.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] According to one aspect of the present invention, there is provided an intelligent operation and maintenance method for rail transit infrastructure based on digital twin technology, characterized in that the method steps include:

[0008] S1. Collect physical entity information and deploy sensors within the operation interval of the rail transit infrastructure to collect real-time sensor data;

[0009] S2. Perform digital twin modeling on the rail transit infrastructure to obtain a three-dimensional digital model of the rail transit infrastructure;

[0010] S3. Adopt the combined simulation method of finite element and multi-body dynamics. Establish a vehicle-road coupling calculation model according to the physical entity information, and calculate the dynamic characteristics of the track structure under the disease condition by this model. Preprocess and integrate the dynamic characteristics and the real-time data of the sensor to form a disease database.

[0011] S4. Based on neural network and support vector machine technologies, use the dynamic characteristics in the disease database to train and obtain a disease damage identification model.

[0012] S5. Build the front-end of the digital twin intelligent operation and maintenance of rail transit based on the three-dimensional digital model of rail transit infrastructure. Send a disease prediction request from this operation and maintenance front-end to the disease damage identification model.

[0013] S6. After receiving the request, the disease damage identification model reads the real-time data of the sensor in the disease database, uses this data to conduct damage identification to generate a real-time prediction result, and returns the real-time prediction result to the operation and maintenance front-end.

[0014] As a preferred technical solution, the physical entity information in S1 includes: the information of operation and maintenance personnel within the operation section of rail transit infrastructure, the mechanical material performance information of track structure infrastructure, and the physical parameters of operating vehicles. Among them, the information of operation and maintenance personnel includes the number of operation and maintenance personnel and the real-time position data; the mechanical material performance information of track structure infrastructure includes the construction production date, density, elastic modulus, yield strength, and geometric parameters of the infrastructure; the physical parameters of operating vehicles include the vehicle self-weight, driving speed, and wheel-rail force parameters. The infrastructure includes rails, track slabs, subgrades, and bridges.

[0015] As a preferred technical solution, the process of digital twin modeling for rail transit infrastructure in S2 includes:

[0016] S21. Plan the acquisition path according to the physical entity information.

[0017] S22. Select the control point positions according to the physical entity information, measure and record the coordinate information of the selected control points, and perform spatial coordinate system conversion on them to obtain the control point information.

[0018] S23. Conduct data acquisition along the planned acquisition path, preprocess and integrate the acquired data to obtain the spatial data of the track structure infrastructure.

[0019] S24. Conduct digital twin geometric modeling with reference to the control point information from the spatial data of the track structure infrastructure.

[0020] As a preferred technical solution, the process of data collection in S23 is as follows: A mobile backpack scanner is used to conduct a full-line push scan, and for preset key areas, a mounted scanner is used to collect point cloud data. During the data collection process, the data quality is monitored in real time, and the scanning strategy is adjusted according to the data quality, so as to obtain the laser point cloud data of the corresponding infrastructure components.

[0021] As a preferred technical solution, the three-dimensional digital model of the rail transit infrastructure in S2 and the vehicle-road coupling calculation model in S3 are divided into grid cells of the same size, and each grid cell corresponds to a component ID constructed by a coding rule. The coding rule is: Each group of type area numbers, pile number area numbers, and component number areas constitutes a minimum unit.

[0022] As a preferred technical solution, the disease database in S3 includes a static information table and a dynamic information table. The static information table selects the component ID as the primary key; the dynamic information table uses the component ID and the timestamp to form a composite primary key; a foreign key is also set in the dynamic information table, and this foreign key points to the primary key of the static information table, that is, it points to the component ID.

[0023] As a preferred technical solution, the specific process of calculating the dynamic characteristics in the case of diseases by the vehicle-road coupling calculation model in S3 is as follows: Input the physical entity information into the vehicle-road coupling calculation model; and consider factors such as vehicle-rail interaction, vibration propagation, and load distribution to simulate the operating state of the basic vehicle-rail facilities, and then simulate the dynamic responses under various operating conditions; finally, analyze and obtain the dynamic characteristics; the dynamic characteristics are time series data, including vibration acceleration and vibration displacement.

[0024] As a preferred technical solution, the specific process of training a disease damage recognition model by using the dynamic characteristics in the disease database in S4 is as follows:

[0025] S41. Clean, denoise, and normalize the dynamic characteristics in the disease database, and fill in the missing data;

[0026] S42. Set the model training objective to learn the characteristics of typical disease types and distinguish between components with diseases and components without diseases, and set the loss function to binary cross-entropy;

[0027] S43. Iteratively train and evaluate the initial neural model according to the settings in S42, so as to obtain a disease damage recognition model.

[0028] According to another aspect of the present invention, there is provided an intelligent operation and maintenance system for rail transit infrastructure based on digital twin technology. This system operates using an intelligent operation and maintenance method for rail transit infrastructure based on digital twin technology as described above. The system includes a physical entity layer, a modeling and mapping layer, a data service layer, an analysis service layer, and an application service layer;

[0029] Among them, the physical entity layer is used to collect physical entity information and real-time sensor data; the modeling and mapping layer is used to model a three-dimensional digital model of rail transit infrastructure to establish a mapping relationship between the digital twin and the physical twin; the data service layer is a disease database, which is established based on the real-time sensor data in the physical entity layer and the dynamic characteristic data in the analysis service layer; the analysis service layer includes a vehicle-road coupling calculation model and a disease damage identification model. Among them, the vehicle-road coupling calculation model is established from the physical entity information in the physical entity layer. In this model, through the combined simulation of finite element and multi-body dynamics, the dynamic characteristic data under disease conditions is calculated; the disease damage identification model reads the real-time sensor data from the data service layer, receives the disease prediction request from the application service layer, and returns the real-time prediction result to the application service layer after identification; the application service layer is the front end of intelligent operation and maintenance of rail transit digital twin, which is constructed based on the three-dimensional digital model of rail transit infrastructure in the modeling and mapping layer.

[0030] As a preferred technical solution, the application service layer includes a disease visualization display module, a fault warning module, and an auxiliary decision-making support module;

[0031] The disease visualization display module is used to present the operation status and disease information of rail transit infrastructure to the operation and maintenance personnel through a graphical interface, and display the health status of facilities such as tracks, subgrades, bridges, and tunnels through the key mechanical performance indicators of the track structure collected in real time by the sensors of physical entities;

[0032] The fault warning module is used to give an early warning of potential faults in rail transit infrastructure based on the real-time prediction results of the algorithm model training module in the analysis service layer;

[0033] The auxiliary decision-making support module is used to intelligently analyze the historical operation and maintenance records and equipment operation data based on the prediction and analysis results of the algorithm model training module in the analysis service layer, provide corresponding operation and maintenance solutions for each fault or disease type, and push the optimized operation and maintenance solutions to the operation and maintenance personnel.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] 1. In the present invention, first, physical entity information and real-time sensor data are collected; then, a vehicle-road coupling calculation model is established by using the combined finite element and multi-body dynamics simulation method, and dynamic characteristics are calculated. The dynamic characteristics are integrated with the real-time sensor data to form a disease database; then, based on neural network and support vector machine technologies, a disease damage identification model is trained by using the dynamic characteristics in the disease database; a digital twin model of rail transit infrastructure is established, and a digital twin intelligent operation and maintenance front end for rail transit is constructed based on this model. During the operation and maintenance process, a disease prediction request is sent from this operation and maintenance front end to the disease damage identification model; when the disease damage identification model receives the request, it reads the real-time sensor data in the disease database, uses this data for damage identification to generate a real-time prediction result, and returns the real-time prediction result to the operation and maintenance front end. In the present invention, by designing an algorithm interface for artificial intelligence-assisted disease identification, the operation and maintenance front end is connected to the disease damage model, and thus potential fault modes and disease development laws of the track structure can be explored, and the disease identification model can be called in real time for prediction in a complex dynamic environment, improving the adaptability and prediction ability of the system.

[0036] 2. The three-dimensional digital model and the vehicle-road coupling calculation model of the rail transit infrastructure in the present invention are both divided into grid units of the same size, and each grid unit corresponds to a component ID constructed by a coding rule. The coding rule is: each group of type area number, pile number area number, and component number area constitutes a minimum unit. By constructing a set of coding rules for the three-dimensional model of rail transit infrastructure, an information interaction mechanism between the physical entity and the digital twin is established, realizing data mapping between the finite element simulation model and the three-dimensional model, and improving the operation and maintenance stability of rail transit infrastructure.

[0037] 3. In the present invention, by preprocessing and integrating the dynamic characteristics and real-time sensor data, a disease database is constructed. The disease database includes a static information table and a dynamic information table. The static information table selects the component ID as the primary key; the dynamic information table uses the component ID and the timestamp to form a composite primary key; a foreign key is also set in the dynamic information table, and this foreign key points to the primary key of the static information table, that is, to the component ID. By constructing a disease database based on multi-source data and vehicle-road coupling dynamic simulation, data support is provided for the development of advanced machine learning algorithms, and thus this method is applicable to various infrastructure operation and maintenance scenarios.

[0038] 4. In the intelligent operation and maintenance system of rail transit infrastructure based on digital twin technology of the present invention, the application service layer is the front end of the digital twin intelligent operation and maintenance of rail transit. It is constructed based on the three-dimensional digital model of the rail transit infrastructure in the modeling and mapping layer, including a disease visualization display module, a fault warning module, and an auxiliary decision-making support module, providing an integrated intelligent operation and maintenance platform with a user-friendly operation interface and real-time interaction capabilities. Operation and maintenance personnel can intuitively understand the health status and operation and maintenance suggestions of the rail transit infrastructure through this platform, greatly simplifying the operation process and improving work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the steps of the intelligent operation and maintenance method for rail transit infrastructure based on digital twin technology in the present invention;

[0040] Figure 2 It is a schematic diagram of the steps of the construction method of the digital twin system for intelligent operation and maintenance of rail transit in the present invention;

[0041] Figure 3 It is a schematic diagram of the three-dimensional modeling process in the present invention;

[0042] Figure 4 It is a schematic diagram of the algorithm model training steps in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0044] Embodiment 1

[0045] In this embodiment, for the prediction of typical diseases of the fastener missing in the rail structure of rail transit infrastructure and the intelligent operation and maintenance of rail transit infrastructure, an intelligent operation and maintenance method for rail transit infrastructure based on digital twin technology is adopted, as Figure 1 shown, including the following steps:

[0046] S1. Collect physical entity information and deploy sensors within the operation section of the rail transit infrastructure to collect real-time sensor data;

[0047] S2. Perform digital twin modeling on the rail transit infrastructure to obtain a three-dimensional digital model of the rail transit infrastructure;

[0048] S3. Adopt the combined simulation method of finite element and multibody dynamics. Establish a vehicle-road coupling calculation model according to the physical entity information, and calculate the dynamic characteristics of the track structure under the disease condition by this model. Preprocess and integrate the dynamic characteristics and the real-time data of the sensors to form a disease database.

[0049] S4. Based on neural network and support vector machine technologies, use the dynamic characteristics in the disease database to train a disease damage identification model.

[0050] S5. Build a digital twin intelligent operation and maintenance front-end for rail transit based on the three-dimensional digital model of rail transit infrastructure. Send a disease prediction request from this operation and maintenance front-end to the disease damage identification model.

[0051] S6. After receiving the request, the disease damage identification model reads the real-time data of the sensors in the disease database, uses this data to perform damage identification to generate a real-time prediction result, and returns the real-time prediction result to the operation and maintenance front-end.

[0052] In this solution, the physical entity information in S1 includes: the information of maintenance personnel within the operation section of rail transit infrastructure, the mechanical material properties information of track structure infrastructure, and the physical parameter information of operating vehicles. Among them, the information of maintenance personnel includes the number of maintenance personnel and real-time position data; the mechanical material properties information of track structure infrastructure includes the construction production date, density, elastic modulus, yield strength, and geometric parameters of the infrastructure; the physical parameters of operating vehicles include vehicle self-weight, driving speed, and wheel-rail force parameters. The infrastructure includes rails, track slabs, subgrades, and bridges.

[0053] In this solution, the process of digital twin modeling of rail transit infrastructure in S2 includes:

[0054] S21. Plan the acquisition path according to the physical entity information.

[0055] S22. Select the control point positions according to the physical entity information, measure and record the coordinate information of the selected control points, and perform spatial coordinate system conversion on them to obtain the control point information.

[0056] S23. Conduct data acquisition along the planned acquisition path, preprocess and integrate the acquired data to obtain the spatial data of the track structure infrastructure.

[0057] S24. Carry out digital twin geometric modeling with reference to the control point information from the spatial data of the track structure infrastructure.

[0058] In this solution, the process of data acquisition in S23 is as follows: A mobile backpack scanner is used to conduct a full-line push scan, and for preset key areas, a mounted scanner is used for point cloud acquisition. During the data acquisition process, the data quality is monitored in real time, and the scanning strategy is adjusted according to the data quality, so as to obtain the laser point cloud data of the corresponding infrastructure components.

[0059] In this solution, the three-dimensional digital model of the rail transit infrastructure in S2 and the vehicle-road coupling calculation model in S3 are divided into grid cells of the same size, and each grid cell corresponds to a component ID constructed by a coding rule. The coding rule is: Each group of type area number, stake number area number, and component number area constitutes a minimum unit.

[0060] In this solution, the disease database in S3 includes a static information table and a dynamic information table. The static information table selects the component ID as the primary key; the dynamic information table uses the component ID and the timestamp to form a composite primary key; a foreign key is also set in the dynamic information table, and this foreign key points to the primary key of the static information table, that is, it points to the component ID.

[0061] In this solution, the specific process of calculating the dynamic characteristics in the case of diseases by the vehicle-road coupling calculation model in S3 is as follows: Input the physical entity information into the vehicle-road coupling calculation model; and consider factors such as vehicle-rail interaction, vibration propagation, and load distribution to simulate the operating state of the basic vehicle-rail facilities, and then simulate the dynamic responses under various operating conditions; finally, analyze and obtain the dynamic characteristics; the dynamic characteristics are time series data, including vibration acceleration and vibration displacement.

[0062] In this embodiment, as Figure 3 shown, the specific steps of the modeling method are as follows:

[0063] First, conduct route planning. According to the geometric characteristics and importance of the track structure, plan a reasonable acquisition path, considering the installation position of the sensor, the acquisition accuracy requirements, and the convenience of subsequent data processing, and reduce dead corners and blind spots.

[0064] Then establish control points. Select appropriate control point positions according to the key structures of the rail transit facilities, accurately measure the selected control points, and record their coordinate information, and perform spatial coordinate system conversion on the acquired data to achieve consistent docking of data in different devices and different measurement processes.

[0065] Then deploy sensors along the predetermined acquisition route to conduct on-site data acquisition.

[0066] Finally, according to the spatial data of the track structure infrastructure, use computer-aided design tools for geometric modeling.

[0067] In this embodiment, as Figure 4As shown in the figure, the specific steps of model training and applying the model are as follows:

[0068] First, extract the original data of typical diseases related to rail transit infrastructure from the database. In the data preprocessing link, clean, denoise, normalize the original data, and fill in the missing data to provide input data in a unified format for model training. The goal of model training is to learn the characteristics of typical disease types and accurately classify whether there are diseases in components. The target loss function of the NNs model is set to Binary Cross-Entropy to establish the intelligent diagnosis ability of rail transit infrastructure diseases. Train and evaluate the disease recognition AI model and deploy it in the intelligent operation and maintenance backend system as the core part of the disease recognition module, which is responsible for analyzing and diagnosing the real-time state of rail transit facilities, identifying possible disease types and their severity. The system front-end interacts with users through a visual interface, displays the identified disease types and distribution, and at the same time receives the instructions input by users and dynamically updates the disease information.

[0069] In summary, in this solution, by designing an algorithm interface for artificial intelligence-assisted disease recognition, connecting the operation and maintenance front-end with the disease damage model, potential fault modes and disease development laws of the track structure can be mined, and in a complex dynamic environment, the disease recognition model can be called in real time for prediction, improving the adaptability and prediction ability of the system.

[0070] Embodiment 2

[0071] This application provides a method for constructing an intelligent operation and maintenance digital twin platform for rail transit infrastructure, including a physical entity layer, a modeling and mapping layer, a data service layer, an analysis service layer, and an application service layer. Among them, the physical entity layer is used to collect physical entity information and sensor real-time data; the modeling and mapping layer is used to model a three-dimensional digital model of rail transit infrastructure to establish a mapping relationship between the digital twin and the physical twin; the data service layer is a disease database, which is established based on the sensor real-time data in the physical entity layer and the dynamic characteristic data in the analysis service layer; the analysis service layer includes a vehicle-road coupling calculation model and a disease damage recognition model. Among them, the vehicle-road coupling calculation model is established from the physical entity information in the physical entity layer. In this model, through the combined simulation of finite element and multi-body dynamics, the dynamic characteristics of the track structure under disease conditions are calculated; the disease damage recognition model reads the sensor real-time data from the data service layer, receives the disease prediction request from the application service layer, and returns the real-time prediction result to the application service layer after recognition; the application service layer is the intelligent operation and maintenance front-end of the rail transit digital twin, which is constructed based on the three-dimensional digital model of rail transit infrastructure in the modeling and mapping layer.

[0072] In this embodiment, the construction process of the system is asFigure 2 As shown below, the specific steps are as follows:

[0073] First, construct the physical entity layer to obtain information such as the information of maintenance personnel, the mechanical material properties of track structure infrastructure, and the physical parameters of operating vehicles within the operation interval of rail transit infrastructure. The information of maintenance personnel includes data such as the number of maintenance personnel and their real-time locations; the mechanical material properties of track structure infrastructure include the construction production dates, densities, elastic moduli, yield strengths, geometric parameters, etc. of infrastructure such as rails, track slabs, subgrades, and bridges; the physical parameters of operating vehicles include parameters such as vehicle self-weight, driving speed, and wheel-rail forces. Deploy vibration acceleration, vibration displacement, and temperature and humidity sensors within the operation interval of rail transit infrastructure to achieve real-time perception of the physical entity state.

[0074] Next, construct the modeling and mapping layer to perform digital twin modeling on rail transit infrastructure. After defining the operation interval of rail transit infrastructure, formulate a data acquisition plan for the point cloud data of the high-precision 3D model of the infrastructure, and use a mobile backpack scanner to conduct a full-line push scan. For key areas with poor track structure conditions, use a mounted scanner to perform high-precision point cloud acquisition. During the data acquisition process, monitor the data quality in real time and adjust the scanning strategy in a timely manner to obtain the high-precision laser point cloud data of the corresponding infrastructure components. Perform preprocessing on the point cloud data using operations such as denoising and filtering. Use the Simultaneous Localization and Mapping (SLAM) technology to integrate the processed point cloud data, and construct a high-precision 3D model of rail transit infrastructure in 3D modeling software. In the 3D modeling software, write a batch processing script to divide the high-precision 3D model of the component into mesh units with the same size and shape as the finite element and multibody dynamics models, and construct a unique ID for each smallest 3D digital model unit according to the unified coding rule, that is, the type area code, the mileage area code, and the component number area.

[0075] Then, construct the data service layer. Adopt the combined finite element and multibody dynamics simulation method to establish a vehicle-road coupling simulation calculation model based on the mechanical material properties of the track structure infrastructure and the physical parameters of the operating vehicle obtained in A1, and calculate the dynamic characteristics of each component of the rail transit infrastructure under various typical disease conditions, including time series data such as vibration acceleration and vibration displacement. Based on the data obtained from the above simulation calculations and the real-time sensor data obtained from the physical entity layer in A1, write a program script to convert the wide-format data into long-format data, and create a typical disease database of rail transit infrastructure in a relational commercial database to complete sub-table storage. In the backend application of the intelligent maintenance system, use the SQLAlchemy library to achieve database interaction, and create a database access layer through the Object Relational Mapping (ORM) technology.

[0076] Rebuild the analysis service layer. By introducing a disease recognition AI model and finite element and multibody dynamics simulation analysis technologies, conduct intelligent diagnosis and prediction of the diseases of rail transit infrastructure. Adopt the finite element and multibody dynamics methods to perform noise reduction and interception, time domain, frequency domain, and time-frequency domain analysis on the data obtained from the simulation calculation in A3, construct typical disease dynamics discrimination indexes, and identify the diseases of components at different positions and in different degrees. Based on the typical disease database of rail transit infrastructure in A3, construct a disease damage recognition AI model based on Neural Networks (NNs) and Support Vector Machine (SVM). Read the real-time physical sensor data from the disease database of rail transit infrastructure in A3, accept the disease prediction request from the front end of the intelligent operation and maintenance system, and return the unique ID of the component that may have diseases.

[0077] Finally, build the application service layer and develop a rail transit digital twin intelligent operation and maintenance front-end system integrating multiple functions. The front-end user interface is designed and constructed based on the use of the open-source Cesium.js library. Upload the high-precision 3D model obtained in A2 to the Cesium ion cloud platform to complete the cloud storage of the model data to ensure the efficiency during network transmission and Web rendering. The 3D model includes the track bridge part, track structure, station, and adjacent road models, and is imported into the system according to the real geographical longitude and latitude coordinates of the above infrastructure and completed rendering.

[0078] In this embodiment, basic functions of the user operation interface are added to the application service layer:

[0079] (1) Click and drag with the left mouse button to move and rotate the view window, and slide the scroll wheel to zoom the window. Add a function for quick switching of the model perspective;

[0080] (2) Add a function for real-time output of geographical longitude and latitude coordinates, and display the geographical longitude and latitude coordinates and elevation information of the position pointed by the current mouse cursor in real time;

[0081] (3) Through the weather and temperature / humidity data API interface, add dynamic data information such as real-time weather, temperature, humidity, wind direction, etc. that are updated in real time;

[0082] (4) Add an animation effect. Read the track structure data information in real time or within a certain time period, and perform real-time rendering on the smallest model unit of the track component according to the magnitude of the physical quantity. Use different colors to represent the magnitude of the physical quantity, and restore the changes in vibration acceleration and vibration displacement at each unit of the track component when the train passes by;

[0083] (5) Establish an orbital structure data management and analysis system to query the vibration acceleration and vibration displacement data of orbital component units at any time or with any known name. Click on any unit on the model with the mouse to display the vibration acceleration, vibration displacement data of the measuring points on the orbital components within a certain period of time, and the corresponding time.

[0084] (6) By plotting the time-domain information charts of these physical quantities, compare them with the data collected by on-site sensors.

[0085] In this embodiment, the server is responsible for processing the backend logic and data storage. A lightweight Web framework is used to write the backend service to receive frontend requests, query, manage, and retrieve the physical property and dynamic characteristic data of any measuring point stored in the local database within any time period.

[0086] In this embodiment, when the disease prediction module identifies that an orbital component fails or has a disease, the system automatically issues an early warning, prompts the longitude and latitude information of the possibly failing component, and gives the navigation route for the maintenance personnel carrying the equipment of this system to conduct on-site inspection and fault troubleshooting. The results of the fault troubleshooting can be reported in real time by the maintenance personnel through this system for verification. The system records the information and is used to further improve the accuracy of the fastener damage classification of the NNs model and the SVM model.

[0087] In this embodiment, the technical processes of each part of the system are as follows:

[0088] The physical entity layer is the source of data for the digital twin service platform and the implementation object of operation and maintenance decision-making; the modeling and mapping layer is used for the modeling, mapping, and encoding of physical entities, and establishes an efficient information interaction mechanism between the digital twin and the physical twin; the data service layer establishes a long linear traffic infrastructure disease database based on real sensor data and simulation model calculation data of finite element and multibody dynamics; the analysis service layer constructs a disease recognition artificial intelligence model based on the above simulation experiment data, and conducts hidden feature learning training on the phenotypic characteristics of typical disease types and degrees; the application service layer is used to develop a geographic information system for multi-source data fusion and dynamic real-time interaction, and integrate key functions such as visualization, disease recognition, and decision support.

[0089] The physical entity layer includes information such as the information of maintenance personnel within the operation area of rail transit infrastructure, the mechanical material properties information of track structure infrastructure, and the physical parameters of operating vehicles. The information of maintenance personnel includes data such as the number of maintenance personnel and real-time positions; the mechanical material properties information of track structure infrastructure includes the production and construction dates, material densities, elastic moduli, yield strengths, geometric parameters, etc. of infrastructure such as rails, track slabs, subgrades, and bridges; the physical parameters of operating vehicles include parameters such as vehicle self-weight, driving speed, and wheel-rail forces. By deploying various sensors in the rail transit system, such as accelerometers, strain gauges, and temperature and humidity sensors, the real-time perception of the physical entity state is realized, providing basic data support for subsequent digital twin modeling and analysis.

[0090] The modeling mapping layer establishes the mapping relationship between the digital twin and the physical twin through high-precision three-dimensional modeling of the physical entity. In this application, the modeling mapping layer includes two modules: modeling method and model matching.

[0091] In this embodiment, the modeling method module is to obtain the dimensions and attributes of the physical entity, using three technical means: oblique photography modeling, terrestrial lidar, and infrastructure design drawings. The first two methods are mainly applied to obtain data during the operation stage of infrastructure, while infrastructure design drawings such as computer-aided design (CAD) drawings help to quickly and low-cost obtain the three-dimensional geometric data of infrastructure during the construction period. As Figure 3 shown, the modeling process includes key technical steps such as the planning of data acquisition routes, the setting of control points, data acquisition and processing, and three-dimensional modeling. For various complex indoor, underground, and signal-free environments of transportation infrastructure, without relying on the traditional Global Navigation Satellite System (GNSS) for measurement, it automatically identifies its own position and the relative spatial relationship with the exploration object during movement, and uses Simultaneous Localization and Mapping (SLAM) to construct an incremental map for accurate mapping of the model.

[0092] In this embodiment, the model matching module is used to realize the transmission and interaction of the node calculation results of the finite element and multi-body dynamics calculation models in the analysis service layer with the corresponding high-precision 3D model in the 3D geographic information system. In the modeling software, the high-precision 3D model restored according to the above modeling method is divided into 3D model units of a fixed size, and the grid of the calculation model units of the finite element and dynamics simulation experiment is kept consistent in size. A unique ID is constructed for each smallest 3D model unit to facilitate parsing by humans and computers. For the naming in the scenario of long linear rail transit infrastructure, three areas are designed for compilation: the type area, the mileage area, and the component number area. The identifier in the type area is used to specify the category or type of infrastructure elements. The mileage area is used to represent the specific location of the traffic infrastructure along the line. The identifier in the component number area is assigned to a single component or element in the traffic infrastructure, allowing each part to be accurately identified.

[0093] The data service layer integrates the sensor data collected in real time and the calculation data based on the finite element and multi-body dynamics simulation of vehicle-road coupling to form a rail transit infrastructure disease database.

[0094] In this embodiment, the database design and construction strategy is selected to create different tables for static information and dynamic information. A static information table is constructed to store attributes that do not change frequently, such as the construction year and design life of the track structure. The primary key of the static information table selects the unique ID of the component as the primary key to ensure the uniqueness of each record. The dynamic information table is used to store information that is updated frequently, such as time series data of strain, acceleration, temperature, etc. detected by sensors. A composite primary key is formed using the unique ID of the component and a timestamp such as the detection time. At the same time, a foreign key is set in the dynamic information table, pointing to the primary key of the static information table, that is, the unique ID of the component.

[0095] The analysis service layer introduces a disease identification AI model and finite element and multi-body dynamics simulation technology to perform intelligent diagnosis and prediction of diseases in rail transit infrastructure. It includes two modules: finite element and multi-body dynamics simulation and algorithm model training.

[0096] The dynamics simulation experiment module establishes a simulation calculation model of the rail transit infrastructure and its surrounding environment, sets material properties, boundary conditions, and load conditions according to the actual working conditions. Then, the grid is divided to generate a discrete finite element model to ensure sufficient accuracy. The dynamic response under different operating conditions is simulated. This module combines factors such as vehicle-rail interaction, vibration propagation, and load distribution to accurately simulate the operating states of facilities such as tracks, subgrades, bridges, and tunnels, and then analyzes possible disease types such as fatigue, wear, and cracks. Finally, the dynamic characteristics are output.

[0097] In this embodiment, the training process of the algorithm model training module is as follows: based on the above-mentioned calculation data of vehicle-road coupling finite element and multi-body dynamics, outliers or missing values are removed, and then the data is normalized. Hidden feature learning training is carried out on the phenotypic characteristics of typical disease types and degrees, such as interlayer contact damage and structural cracking damage. After the model training is completed, indicators such as the prediction accuracy, precision, and mean square error of the model are evaluated to confirm the actual application effect of the model.

[0098] The application service layer develops a rail transit digital twin intelligent operation and maintenance system integrating multiple functions. Through a dynamic real-time interactive geographic information system, functions such as disease visualization display, fault warning, and auxiliary decision-making support are integrated.

[0099] The disease visualization display module presents the operation status and disease information of rail transit infrastructure to the operation and maintenance personnel through a graphical interface. Based on the key mechanical performance indicators of the track structure collected in real time by the sensors of the physical entity, this module can display the health status of facilities such as tracks, subgrades, bridges, and tunnels.

[0100] The fault warning module gives an early warning of potential faults in rail transit infrastructure based on the real-time prediction results of the algorithm model training module in the analysis service layer. This module uses real-time data and historical data to predict faults and pushes warning information according to the disease diagnosis results.

[0101] The auxiliary decision-making support module shows the specific locations that need to be inspected on-site, the degree of disease development, and the specific inspection paths to the operation and maintenance personnel based on the disease analysis results of the algorithm model training module in the analysis service layer. This module provides an operable disease inspection plan for the operation and maintenance personnel by calculating and analyzing the real-time mechanical performance indicators of the track structure.

[0102] In this embodiment, this solution is designed based on the browser / server mode to avoid compatibility problems caused by differences in operating systems and devices during the subsequent use of the system. Through the modeling, mapping, and encoding of physical entities, an efficient information interaction mechanism is established between the digital twin and the physical twin. Based on real sensor data and the calculation data of vehicle-road coupling finite element and multi-body dynamics simulation, a disease database of long linear transportation infrastructure is established to provide reliable data support for the construction of the disease recognition AI model. Finally, the development of the front end of the interactive system and the integration of the AI model are completed to establish a digital twin of rail transit infrastructure with application service functions such as visualization, disease recognition, and decision-making support.

[0103] In summary, an integrated intelligent operation and maintenance platform is provided in this solution, which has a user-friendly operation interface and real-time interaction capabilities. Operation and maintenance personnel can intuitively understand the health status and operation and maintenance suggestions of rail transit infrastructure through this platform, greatly simplifying the operation process and improving work efficiency.

[0104] As described above, the above is only a specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An intelligent operation and maintenance method for rail transit infrastructure based on digital twin technology, characterized in that: The method steps include: S1. Collect physical entity information and deploy sensors within the operating area of ​​rail transit infrastructure to collect real-time sensor data; S2. Conduct digital twin modeling of rail transit infrastructure to obtain a three-dimensional digital model of the rail transit infrastructure; S3. Using the finite element and multi-body dynamics joint simulation method, a vehicle-road coupling calculation model is established based on the physical entity information, and the dynamic characteristics of the track structure under the condition of damage are calculated by the model. The dynamic characteristics are pre-processed and integrated with the real-time data of the sensor to form a damage database; S4. Based on neural network and support vector machine technology, the dynamic characteristics in the disease database are used to train the disease damage recognition model; S5. Construct a rail transit digital twin intelligent operation and maintenance front end based on the three-dimensional digital model of the rail transit infrastructure, and send a disease prediction request to the disease and damage identification model by the operation and maintenance front end; S6. After receiving the request, the disease damage identification model reads the real-time data of the sensor in the disease database, uses the data to perform damage identification to generate real-time prediction results, and returns the real-time prediction results to the operation and maintenance front end.

2. According to claim 1, a method for intelligent operation and maintenance of rail transit infrastructure based on digital twin technology is characterized in that: The physical entity information in S1 includes: operation and maintenance personnel information within the operating range of the rail transit infrastructure, mechanical material performance information of the rail structure infrastructure, and physical parameter information of the running vehicles, wherein the operation and maintenance personnel information includes the number of operation and maintenance personnel and real-time location data; the mechanical material performance information of the rail structure infrastructure includes the construction and production date, density, elastic modulus, yield strength, and geometric parameters of the infrastructure; the physical parameters of the running vehicles include the vehicle's own weight, driving speed, and wheel-rail force parameters, and the infrastructure includes rails, track plates, roadbeds, and bridges.

3. According to claim 1, a method for intelligent operation and maintenance of rail transit infrastructure based on digital twin technology is characterized in that: The process of digital twin modeling of rail transit infrastructure in S2 includes: S21. Plan a collection path based on physical entity information; S22, selecting the control point position according to the physical entity information, measuring and recording the coordinate information of the selected control point, and performing a spatial coordinate system conversion on the control point to obtain the control point information; S23, collecting data along the planned collection path, preprocessing and integrating the collected data, and obtaining spatial data of the track structure infrastructure; S24. Digital twin geometry modeling is performed based on the spatial data reference control point information of the rail structure infrastructure.

4. According to claim 3, a method for intelligent operation and maintenance of rail transit infrastructure based on digital twin technology is characterized in that: The data collection process in S23 is as follows: a mobile backpack scanner is used to conduct a forward scan along the entire line, and a mounted scanner is used to collect point cloud data for preset key areas, data quality is monitored in real time during the data collection process, and the scanning strategy is adjusted according to the data quality, thereby obtaining laser point cloud data of the corresponding infrastructure components.

5. According to claim 1, a method for intelligent operation and maintenance of rail transit infrastructure based on digital twin technology is characterized in that: The three-dimensional digital model of the rail transit infrastructure in S2 and the vehicle-road coupling calculation model in S3 are divided into grid units of the same size, and each grid unit corresponds to a component ID constructed according to a coding rule. The coding rule is: each group of type area code, pile number area code and component number area constitutes a minimum unit.

6. The intelligent operation and maintenance method of rail transit infrastructure based on digital twin technology according to claim 5 is characterized in that: The disease database in S3 includes a static information table and a dynamic information table. The static information table selects the component ID as the primary key; the dynamic information table uses the component ID and timestamp to form a composite primary key; a foreign key is also set in the dynamic information table, which points to the primary key of the static information table, that is, the component ID.

7. The intelligent operation and maintenance method of rail transit infrastructure based on digital twin technology according to claim 1 is characterized in that: The specific process of calculating the dynamic characteristics under the disease condition by the vehicle-road coupling calculation model in S3 is: inputting the physical entity information into the vehicle-road coupling calculation model; and considering the vehicle-track interaction, vibration propagation and load distribution factors to simulate the operating state of the basic vehicle-track facilities, and then simulate the dynamic response under various operating conditions; finally, analyzing and obtaining the dynamic characteristics of the track structure; the dynamic characteristics are time series data, including vibration acceleration and vibration displacement.

8. The intelligent operation and maintenance method of rail transit infrastructure based on digital twin technology according to claim 1 is characterized in that: The specific process of training the disease damage recognition model by using the dynamic characteristics in the disease database in S4 is as follows: S41. Clean, denoise, and normalize the dynamic characteristics in the disease database, and fill in missing data; S42, setting the model training goal to learn the characteristics of typical disease types and distinguishing between diseased components and non-defective components, and setting the loss function to binary cross entropy; S43. Iteratively train and evaluate the initial neural model according to the settings in S42 to obtain a disease and damage recognition model.

9. An intelligent operation and maintenance system for rail transit infrastructure based on digital twin technology, characterized in that: The system applies a rail transit infrastructure intelligent operation and maintenance method based on digital twin technology as described in any one of claims 1 to 8, and the system includes a physical entity layer, a modeling mapping layer, a data service layer, an analysis service layer and an application service layer; The physical entity layer is used to collect physical entity information and real-time sensor data; The modeling and mapping layer is used to model and obtain a three-dimensional digital model of the rail transit infrastructure to establish a mapping relationship between the digital twin and the physical twin; The data service layer is a disease database, which is established based on the real-time data of sensors in the physical entity layer and the dynamic characteristic data in the analysis service layer; The analysis service layer includes a vehicle-road coupling calculation model and a disease damage identification model, wherein the vehicle-road coupling calculation model is established by the physical entity information in the physical entity layer, in which the dynamic characteristic data under the disease condition is calculated through the joint simulation of finite element and multi-body dynamics; the disease damage identification model reads the real-time data of the sensor from the data service layer, receives the disease prediction request from the application service layer, and returns the real-time prediction result to the application service layer after identification; The application service layer is the intelligent operation and maintenance front end of the rail transit digital twin, which is constructed based on the three-dimensional digital model of the rail transit infrastructure in the modeling and mapping layer.

10. The intelligent operation and maintenance system for rail transit infrastructure based on digital twin technology according to claim 9 is characterized in that: The application service layer includes a disease visualization display module, a fault warning module and an auxiliary decision support module; The disease visualization display module is used to present the operating status and disease information of rail transit infrastructure to operation and maintenance personnel through a graphical interface, and to display the health status of facilities such as tracks, roadbeds, bridges, and tunnels through key mechanical performance indicators of track structures collected in real time by sensors of physical entities; The fault warning module is used to provide early warning of potential faults of rail transit infrastructure based on the real-time prediction results of the analysis service layer algorithm model training module; The auxiliary decision support module is used to predict the analysis results based on the analysis service layer algorithm model training module, intelligently analyze historical operation and maintenance records and equipment operation data, provide corresponding operation and maintenance plans for each fault or disease type, and push optimized operation and maintenance plans to operation and maintenance personnel.

Images