Urban rail track full life cycle intelligent management and control method based on digital twinning

Abstract

This invention discloses an intelligent management and control method for the entire life cycle of urban rail transit based on digital twins, relating to the field of intelligent operation and maintenance technology for rail transit facilities. The method includes initializing a digital twin rail model aligned with the physical rails and marking life-sensitive areas. Historical dynamic load spectrum sequences are extracted based on the unique codes of the target rail sections, and load transfer paths are simulated in the model accordingly, identifying the core load action area and edge influence area within the sensitive region. Boundary corrections are performed on these areas by fusing on-site detection maps, generating a preliminary set of focus areas for life analysis. Areas outside this set are defined as environmental coupling action areas and merged with adjacent parts of the focus areas to form the final rail life management domain. This method achieves precise location and proactive management of potential rail damage areas through refined simulation of load action and fusion of multi-source data.

Description

Technical Field

[0001] This invention belongs to the field of intelligent operation and maintenance technology for rail transit facilities, specifically a method for intelligent management and control of the entire life cycle of urban rail tracks based on digital twins. Background Technology

[0002] Current life assessments of urban rail transit primarily rely on periodic manual inspections and fixed monitoring data, combined with static mechanical models based on homogenization assumptions. Existing methods fail to adequately consider the dynamic load sequences generated by differences in train type, axle load, and speed during actual operation, and neglect the spatial non-uniformity of load transmission within the rail structure. This results in insufficient accuracy in identifying the true locations of localized stress concentrations and fatigue accumulation, making it difficult to accurately pinpoint the initiation micro-regions of potential damage.

[0003] In existing digital twin applications, rail condition modeling primarily focuses on geometric visualization and monitoring data integration. On-site non-destructive testing data and simulation-based mechanical analysis results are often used independently, lacking a deep fusion and cross-verification mechanism under a unified spatial benchmark. The high-risk areas predicted by the model often deviate from the actual detected defect locations, making it difficult for the digital model to accurately reflect the damage evolution process of the physical rail and limiting its role in accurate lifespan prediction and maintenance decision-making.

[0004] Current technologies cannot achieve refined simulation of the impact of local stress on rails based on real dynamic load spectra, nor can they effectively integrate measured data to calibrate the simulation prediction area. A method is needed that can integrate dynamic load transfer simulation and field inspection information in a digital twin space to accurately identify and continuously correct key areas of rail lifespan. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; To this end, the present invention proposes an intelligent management and control method for the entire life cycle of urban rail transit based on digital twins, including: Initialize a digital twin rail model that is spatially aligned with the physical rail, and mark multiple life-sensitive regions in the digital twin rail model based on mechanical properties; Obtain the unique code of the target rail section, and extract the dynamic load spectrum sequence related to train operation from the historical database based on the unique code; Based on the dynamic load spectrum sequence, the load transfer path is simulated on the digital twin rail model, thereby identifying the load core action area and the load edge influence area within the life-sensitive region. By integrating the field inspection maps of the target rail section, boundary corrections are performed on the load core action area and the load edge influence area to generate a preliminary set of life analysis focus areas; In the digital twin rail model, the area outside the preliminary life analysis focus area set is defined as the environmental coupling effect zone, and the part of the environmental coupling effect zone that is adjacent to the preliminary life analysis focus area set is merged with the preliminary life analysis focus area set to form the final rail life management domain.

[0006] Preferably, the method further includes: Each partition in the final rail life management domain is configured with an independent life evolution calculation task, and the execution priority order of the life evolution calculation task is set according to the geometric adjacency relationship and mechanical correlation strength between each partition. According to the execution priority order, the life evolution calculation task is executed sequentially for each partition in the final rail life management domain to generate the prediction result of the remaining service time of the partition; The remaining service life prediction results of all partitions are aggregated to drive the digital twin rail model to perform state evolution simulation and output maintenance decision instructions for the target rail section.

[0007] Preferably, in the digital twin rail model, multiple life-sensitive regions are marked according to mechanical properties, including: Multiple stress calculation sampling points are preset on the geometry of the digital twin rail model; A standard load condition is applied to the digital twin rail model, and the stress response value of each stress calculation sampling point is calculated. The calculated stress response value is compared with a preset stress threshold. The continuous spatial range formed by stress calculation sampling points where the stress response value continuously exceeds the preset stress threshold is marked as an independent life-sensitive region.

[0008] Preferably, extracting the dynamic load spectrum sequence related to train operation from the historical database based on the unique code includes: Parse the unique code to obtain the operating line number, laying location mileage information, and rail material type of the line where the target rail section is located; Using the operating line number and the laying location mileage information as indexes, query the historical database for all train numbers passing through the target rail section; For each train number record retrieved, the axle load data, running speed data, and passing timestamp of the corresponding train are retrieved. The axle load data and running speed data are combined in chronological order according to the passing timestamps to form the dynamic load spectrum sequence sorted by time.

[0009] Preferably, based on the dynamic load spectrum sequence, simulating the load transfer path on the digital twin rail model includes: Select a representative load condition from the dynamic load spectrum sequence. The representative load condition includes axle load value and running speed value. In the digital twin rail model, the theoretical contact point between the wheel and the rail is used as the origin of load application, and force and velocity boundary conditions corresponding to the representative load conditions are applied. Run the transient dynamics solver of the digital twin rail model to calculate the transmission process and spatial attenuation law of stress waves inside the rail model under the representative load conditions. Based on the transmission process and spatial attenuation law of the stress wave, the load energy transfer trajectory radiating outward from the origin of the load application is plotted. The section with the highest energy density on the load energy transfer trajectory constitutes the core action area of ​​the load, and the transition section with significantly decreased energy density constitutes the edge influence area of ​​the load.

[0010] Preferably, the boundary correction of the load core action area and the load edge influence area by integrating the field inspection map of the target rail section includes: The surface damage optical image and internal defect ultrasonic image of the target rail section are acquired by the detection equipment and together constitute the on-site detection map. The optical image of the surface damage is registered with the surface geometry of the digital twin rail model to identify the actual location of the damaged area in the image on the digital twin rail model. The ultrasonic imaging image of the internal defect is registered with the internal structure of the digital twin rail model to locate the three-dimensional coordinates of the defect in the digital twin rail model. The identified actual location and the located three-dimensional coordinates are superimposed and compared with the spatial range of the previously identified load core action area and load edge influence area; If the actual location or the three-dimensional coordinates are within a preset range of the load core action area or the load edge influence area, the original boundary remains unchanged; if they are outside the preset range but in their vicinity, the boundary of the load core action area or the load edge influence area is expanded accordingly to include the actual location or the three-dimensional coordinates, thereby generating the preliminary set of life analysis focus areas.

[0011] Preferably, merging the portion of the environmental coupling zone adjacent to the preliminary lifetime analysis focus area set with the preliminary lifetime analysis focus area set includes: Calculate the coordinates of the geometric center point of each region in the preliminary lifetime analysis focus region set; Using the coordinates of each geometric center point as the center and the preset radius of influence as the length, a sphere is constructed in the space of the digital twin rail model. The radius of influence is obtained by looking up a table based on the rail material and the environmental corrosion rate. Within the environmental coupling zone, identify all sub-regions that intersect with the spherical space; The overlapping sub-regions are extracted from the environmental coupling zone and added to the preliminary life analysis focus region set. The merged overall region is defined as the final rail life management domain.

[0012] Preferably, an independent life evolution calculation task is configured for each partition in the final rail life management domain, and the execution priority order of the life evolution calculation task is set according to the geometric adjacency relationship and mechanical correlation strength between each partition, including: The final rail life management domain is further subdivided into grids, such that each subdivided grid is an independent partition; For each partition, a corresponding material fatigue damage calculation model and creep evolution calculation model are defined. The material fatigue damage calculation model and the creep evolution calculation model together constitute the lifetime evolution calculation task of the partition. Analyze the spatial distance between any two partitions and the mechanical correlation strength between the two partitions, which is obtained by comparing the stress covariance of the two partitions under historical loads; Two partitions whose spatial distance is less than a set threshold and whose mechanical correlation strength is higher than a set correlation value are defined as a strongly correlated partition pair. When setting the execution priority order, it is mandatory that the lifetime evolution calculation tasks of two partitions belonging to the same strongly correlated partition pair must be executed consecutively, and the lifetime evolution calculation task corresponding to the partition with higher stress level in the strongly correlated partition pair is executed first.

[0013] Preferably, performing the life evolution calculation task sequentially on each partition in the final rail life management domain includes: Select the partition to be calculated according to the execution priority order; Load the material fatigue damage calculation model configured for the partition, and input the historical load spectrum segment experienced by the partition to calculate the cumulative fatigue damage increment of the partition under the current load cycle; Simultaneously, the creep evolution calculation model configured for the partition is loaded, and the historical data of the average stress and ambient temperature of the partition are input to calculate the creep strain increment of the partition at the current time step. Based on the calculated cumulative fatigue damage increment and creep strain increment, the material state parameters of the partition in the digital twin rail model are updated; Based on the updated material state parameters and the fracture toughness curve of the partition material, the theoretical number of failure cycles or time of the partition under the current operating load conditions is deduced, and the theoretical number of failure cycles or time is used as the prediction result of the remaining service life of the partition.

[0014] Preferably, the remaining service life prediction results of all partitions are aggregated to drive the digital twin rail model to perform state evolution simulation, and the maintenance decision instructions for the target rail section are output, including: Collect the remaining service life prediction results of all partitions in the final rail life management domain to form a remaining life distribution list; In the list of remaining service life distributions, find the remaining service life prediction result with the smallest value and mark it as the life prediction value of the weakest zone; In the digital twin rail model, the spatial location corresponding to the weakest section is located, and the type of rail structure component to which the spatial location belongs is retrieved; Based on the predicted lifespan of the weakest section and the type of rail structure component, a preset maintenance strategy mapping table is queried. The maintenance strategy mapping table defines the specific maintenance actions and maintenance timings corresponding to different lifespan ranges and component types. The maintenance decision instruction is generated by matching the corresponding specific maintenance action and maintenance timing from the maintenance strategy mapping table.

[0015] Compared with the prior art, the beneficial effects of the present invention are: Based on dynamic load spectrum sequences extracted from historical train operations, dynamic simulations of load transfer paths are performed in a digital twin rail model, surpassing the assumptions of static or uniformly distributed loads. This technology can further distinguish the core load area from the peripheral load influence area within a pre-defined life-sensitive region based on the mechanical response gradient. This refined regional division based on actual load history directly identifies the microscopic locations subjected to the most severe cyclic stresses, transforming the location of the core zone for rail fatigue damage initiation and propagation from empirical judgment to precise identification driven by data and mechanisms.

[0016] The field inspection maps reflecting the actual internal or surface condition of the rails are spatially fused with the load core action area and edge influence area identified through mechanical simulation. Using physical evidence such as actual defects and material changes provided by the inspection maps, the regional boundaries derived from purely theoretical simulations are calibrated and corrected. This process achieves a closed-loop feedback between simulation predictions and measured information in spatial location. The resulting set of life analysis focus areas possesses both mechanical inevitability and experimental conformity, improving the reliability and accuracy of risk area identification.

[0017] The set of focal areas, fused and calibrated from multi-source data, is intelligently merged with adjacent areas affected by environmental factors to form the final lifespan governance domain. This operation breaks the limitation of strictly relying on a single mechanism to define management boundaries, constructing a complete control space covering direct damage risks and indirect inducing factors. This enables maintenance strategies to target not only clearly identified core fatigue points but also the surrounding environment that may be affected or contribute to their development, achieving a leap from "point-based" treatment to "domain-based" collaborative prevention and control. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the steps of the intelligent management and control method for the entire life cycle of urban rail transit based on digital twins as described in this invention. Figure 2 To maintain the flowchart generated by the decision-making instructions; Figure 3 This is a flowchart for extracting dynamic load spectrum sequences; Figure 4 A comparison chart of the influence radii of steel rails made of different materials; Figure 5 A thermal diagram of the mechanical correlation strength of the rail life management zone. Detailed Implementation

[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1A digital twin rail 3D model, precisely aligned with the physical rail's spatial location, was established. Based on the material properties and structural form of this digital twin model, a pre-defined mechanical analysis process identified and marked multiple life-sensitive areas prone to damage due to design or construction reasons. A unique identification code was assigned to each target rail section. By parsing this code and retrieving its historical operational data, a dynamic load spectrum sequence reflecting the wheel force sequence and corresponding speed generated when a train passes through that section was extracted. Using this dynamic load spectrum sequence, dynamic simulations were performed in the digital twin rail model, simulating the path of load transmission from the wheel-rail contact point to the interior of the rail. Based on the simulation results, the core load-bearing area directly subjected to high loads and the load-edge influence area affected by these loads were further distinguished within the pre-marked life-sensitive areas. Integrating on-site rail surface and internal inspection data to form an on-site inspection map, this map information is spatially compared and fused with the simulated load core action area and load edge influence area. The boundaries of the simulated area are calibrated and corrected based on the actual defect locations, thus generating a preliminary set of life analysis focus areas that more closely resembles the actual damage condition. In the digital twin rail model, the rail area outside the aforementioned preliminary life analysis focus area set is defined as the environmental coupling action area subject to long-term effects from environmental factors such as temperature, humidity, and corrosive media. The proximity relationship between this area and the preliminary focus area set is analyzed, and local environmental coupling action areas adjacent to the preliminary set and potentially affected by its damage evolution are included, merging to form a comprehensive final rail life management domain, serving as the complete target range for subsequent refined life assessment and control.

[0021] In one embodiment of the present invention, see [reference] Figure 2 Each partition within the final rail life management domain is configured with an independent life evolution calculation task, which includes a damage evolution model tailored to the material properties of that partition. The spatial adjacency relationships between partitions and the degree of mutual influence of their stress responses under historical loads (i.e., mechanical correlation strength) are analyzed to determine the execution priority of the life evolution calculation tasks for each partition. Based on the set execution priority, the configured life evolution calculation tasks are executed sequentially for each partition within the final rail life management domain, outputting a predicted remaining service life for that partition under its current state and historical load history. The predicted remaining service life results for all partitions are summarized and aggregated. This data drives a digital twin rail model to perform an overall state evolution simulation, simulating the evolution trend of the rail state over a future period. Based on the simulation results, maintenance decision instructions for the target rail section are output.

[0022] In practical implementation, each partition in the final rail life management domain is configured with an independent life evolution calculation task. This task includes a combination of a material fatigue damage calculation model and a creep evolution calculation model. In some embodiments, the final rail life management domain is divided into three-dimensional meshes, with each mesh cell serving as an independent partition. A corresponding material parameter database is loaded for each partition to initialize the life evolution calculation task. The execution priority of the life evolution calculation task is set based on the geometric adjacency relationship and mechanical correlation strength between partitions. The geometric adjacency relationship is obtained by calculating the spatial distance matrix between partitions, and the mechanical correlation strength is obtained by analyzing the covariance matrix of the stress response sequences of the partitions under historical loads. For example, in a practical implementation, the covariance value of the stress time series of two partitions is calculated; a higher covariance value indicates a stronger mechanical correlation strength. The execution priority can be set according to the following formula: Wherein: the symbol "exp()" represents the natural exponential function, that is, an exponential function with the natural constant e (approximately 2.71828) as its base. This function operates on the parameter enclosed in parentheses, and The calculation result is e Power; This represents the association priority coefficient between partition i and partition j; This represents the normalized spatial distance between partition i and partition j; This represents the normalized value of the mechanical correlation strength between partition i and partition j; and It is a preset weighting factor used to balance the influence of geometric adjacency and mechanical correlation strength. By calculating the correlation priority coefficient of all partition pairs, the partitions are arranged in descending order of the coefficient to determine the execution order of the lifetime evolution calculation task.

[0023] In specific implementation, life evolution calculation tasks are performed sequentially on each partition within the final rail life management domain according to execution priority, generating predicted results for the remaining service time of each partition. For example, for a selected partition, the historical load spectrum fragments and ambient temperature data experienced by the partition are input into the material fatigue damage calculation model and the creep evolution calculation model. The material fatigue damage calculation model outputs the cumulative fatigue damage increment, and the creep evolution calculation model outputs the creep strain increment. Based on these increments, the material state parameters of the partition in the digital twin rail model are updated, thereby deducing the theoretical number of failure cycles or time for the partition under the current operating load conditions as the predicted result for the remaining service time. In some embodiments, the execution process adopts an iterative calculation framework, synchronously updating the global state of the digital twin rail model after each partition calculation to ensure that the calculation of subsequent partitions is based on the latest model state. Optionally, for high-priority partitions, the calculation task can allocate more computing resources to reduce simulation time.

[0024] The system aggregates the predicted remaining service life of all rail sections, drives the digital twin rail model to perform state evolution simulation, and outputs maintenance decision instructions for the target rail section. This aggregation process involves collecting the predicted remaining service life of all rail sections to form a list, identifying the section with the smallest predicted remaining service life in the list as the weakest section, and retrieving the rail structural component type to which the weakest section belongs in the digital twin rail model. Based on the predicted lifespan of the weakest section and the component type, a pre-defined maintenance strategy mapping table is queried. This table defines specific maintenance actions and timings for different lifespan intervals and component types, thereby matching and generating maintenance decision instructions. For example, in practical implementation, the maintenance decision instructions include suggested maintenance time, maintenance operation type, and impact range. Optionally, the state evolution simulation can simulate the rail lifespan evolution under different maintenance strategies to optimize the generation of maintenance decision instructions.

[0025] In one embodiment of the present invention, the process of marking multiple life-sensitive regions in a digital twin rail model based on mechanical properties is as follows: A series of stress calculation sampling points for mechanical calculations are preset on the geometric surface and key internal locations of the digital twin rail model. A standardized load condition is applied to the digital twin rail model, and the stress response value of each stress calculation sampling point under this condition is obtained through finite element analysis. The calculated stress response value of each stress calculation sampling point is compared with a preset stress threshold determined based on the material yield strength and safety factor. All stress calculation sampling points whose stress response values ​​continuously exceed the preset stress threshold are identified, and the continuous spatial range formed by these sampling points in three-dimensional space is marked as an independent life-sensitive region.

[0026] In practical implementation, multiple stress calculation sampling points are preset on the geometry of the digital twin rail model. These sampling points are arranged according to the geometric characteristics of the rail. For example, higher lattice density is used in areas with abrupt changes in geometry or design weaknesses, such as the rail head arc region, rail web connection, and rail bottom edge. In some embodiments, the coordinate information of the stress calculation sampling points is stored in a separate configuration file and mapped to the mesh nodes or cell centers of the digital twin rail model. A standard load condition is applied to the digital twin rail model. This standard load condition is simulated by a static or quasi-static wheel load with representative axle load and speed. This load is mainly vertical, and its effective area and distribution are determined according to wheel-rail contact theory. It can be understood that the standard load condition serves as a unified reference condition to trigger the mechanical response of the digital twin rail model for initial evaluation.

[0027] In practical implementation, the stress response value at each stress calculation sampling point is calculated. This calculation is performed by calling the finite element solver embedded in the digital twin rail model. The solver uses elasticity theory to solve for the stress field distribution of the rail model under standard load conditions and extracts the equivalent stress value at each stress calculation sampling point as the stress response value from the solution results. Optionally, the stress response value can be calculated based on the Mises stress criterion or the maximum principal stress criterion. The calculated stress response value is compared with a preset stress threshold. The preset stress threshold is determined comprehensively based on the yield strength, fatigue limit, and safety specifications of the rail material. The numerical relationship can be expressed as follows: in: This represents the preset stress threshold. Represents the yield strength of the rail material; This represents the fatigue limit of the rail material. It is the yield strength reduction factor; It is a safety factor, calculated by measuring the stress response value at each stress sampling point. and Compare them to determine their risk level.

[0028] In practical implementation, the continuous spatial range formed by stress calculation sampling points where the stress response value continuously exceeds a preset stress threshold is marked as an independent life-sensitive region. The determination process traverses all stress calculation sampling points to identify all those that meet the criteria. Stress calculation sampling points are selected based on the conditions, and cluster analysis is performed on these points based on their proximity in three-dimensional space. Stress calculation sampling points that are less than a set cluster radius are grouped into the same set. The spatial volume covered by each set is defined as an independent life-sensitive region. In some embodiments, the cluster analysis can employ a density-based spatial clustering algorithm. It is understood that the labeling results are stored in the metadata of the digital twin rail model in the form of a list of spatial coordinate boundaries and are associated with the model's geometric visualization components for highlighting. Optionally, each independent life-sensitive region is assigned a unique region identifier for subsequent management.

[0029] In one embodiment of the present invention, see [reference] Figure 3The method for extracting dynamic load spectrum sequences related to train operation from a historical database based on unique codes is as follows: The unique code of the target rail section is parsed to extract the operating line number, specific laying location mileage information, and rail material type. Using the parsed operating line number and laying location mileage information as a joint index key, a query is performed in the historical database storing train operation records to retrieve the historical records of all trains that have passed through the target rail section. For each train record in the query results, the axle load data, running speed data, and precise passing timestamp are retrieved from the database's associated table. The axle load data and corresponding running speed data for each train are combined in chronological order according to the passing timestamps to form a dynamic load spectrum sequence arranged in chronological order.

[0030] In practice, the unique code of the target rail section is parsed. This unique code is a structured string whose coding rules follow established industry or enterprise standards. The code sequentially includes fields such as the operating line number, laying location mileage information, and rail material type. The parsing process uses delimiters or fixed-length truncation to break down the unique code string into independent field data units, such as the operating line number "Line_03A", the laying location mileage information "K25+187.5", and the rail material type "U75V". Based on the unique code, the dynamic load spectrum sequence related to train operation is extracted from the historical database. In practice, the parsed operating line number and laying location mileage information are used as joint query conditions to generate a structured query language statement. The retrieval operation is then performed in a specific data table in the historical database storing train operation records. This historical database records information such as train number, passing time, location, and operating status.

[0031] In practical implementation, using the operating line number and laying location mileage information as indexes, the system queries the historical database for all train numbers passing through the target rail section. The query logic is to filter out all data rows where the location field matches the target laying location mileage information and the line identifier field matches the target operating line number. Each data row that meets the conditions represents a train passing event and includes the train number identifier corresponding to that event. In some embodiments, to improve query efficiency, the database will create a composite index on the operating line number and laying location mileage information fields. For each train number record retrieved, the corresponding train's axle load data, running speed data, and passing timestamp are retrieved. The axle load data is obtained from the train's basic attribute table through the train number identifier, representing the static vertical force applied to the rail by each wheel of the train. The running speed data is obtained from the train running status time series table through the train number identifier and passing timestamp, representing the instantaneous speed of the train when passing through the target rail section. The timestamp is directly extracted from the query result record, representing the precise time point of the train passing event.

[0032] In practice, axle load data and running speed data are combined in chronological order according to timestamps to form a time-sorted dynamic load spectrum sequence. The combination process creates a data entry for each timestamp, containing the timestamp, corresponding axle load data, and running speed data. All entries are then sorted in ascending order of timestamps to form a time series, which is the dynamic load spectrum sequence. This sequence describes the relationship between load and speed acting on the target rail section as it changes over time. In some embodiments, the data structure of the dynamic load spectrum sequence can be represented as an ordered list. ,in Represents the dynamic load spectrum sequence. This represents the timestamp indicating the passage of the k-th train. This represents the axle load data for the k-th train passing through. This represents the speed data of the k-th train passing by, where n represents the total number of train passing events found. Optional, axle load data. It can be a scalar value representing the average axle load, or a vector representing the specific axle load of each wheel. Optionally, for records with missing operating speed data, speed data from adjacent timestamps can be interpolated to fill the gaps, or typical operating speed data of the train in adjacent sections can be used to fill the gaps, to ensure the continuity of the dynamic load spectrum sequence.

[0033] In one embodiment of the present invention, the load transfer path is simulated on a digital twin rail model based on a dynamic load spectrum sequence. A load condition with statistical representativeness or extreme value characteristics is selected from the dynamic load spectrum sequence as a representative load condition, which includes specific axle load values ​​and operating speed values. In the digital twin rail model, the theoretical wheel-rail contact point is used as the load application origin, and force and velocity boundary conditions corresponding to the representative load condition are applied. The transient dynamics solver integrated into the digital twin rail model is run to calculate the dynamic process of stress wave transmission from the contact point to the surrounding area within the rail model under the representative load condition and its spatial attenuation law. Based on the calculated stress wave transmission process and spatial attenuation law, the load energy transfer trajectory radiating outward from the load application origin is plotted in the model. The continuous segment with the highest energy density on this trajectory is marked as the load core action area, and the transition segment where the energy density decreases significantly is marked as the load edge influence area. By fusing the on-site inspection images of the target rail section, boundary corrections are performed on the load core action area and load edge influence area. Surface damage optical images and internal defect ultrasonic images of the target rail section, acquired through inspection equipment, are obtained to form the on-site inspection atlas. The surface damage optical images are then coordinate-registered with the surface 3D geometry of the digital twin rail model to identify the actual spatial location of all damaged areas in the images on the digital twin rail model. The internal defect ultrasonic images are then registered with the internal structure of the digital twin rail model to locate the 3D coordinates of each defect within the digital twin rail model. The identified actual damage locations and located internal defect 3D coordinates are then overlaid and compared with the spatial ranges of the load core action area and load edge influence area previously identified through simulation. If the actual location or 3D coordinate point is within a preset proximity range of the load core action zone or load edge influence zone, the original simulation boundary remains unchanged. If the actual location or 3D coordinate point is outside the preset range but within its spatial proximity, the boundary of the load core action zone or load edge influence zone is expanded accordingly to include the actual location or 3D coordinate point. The resulting set of regions after this correction is the preliminary life analysis focus region set. The regions adjacent to the preliminary life analysis focus region set in the environmental coupling action zone are merged, and the geometric center coordinates of each region in the preliminary life analysis focus region set are calculated. Using each geometric center coordinate as the center and a fixed length derived from a table based on the rail material and the average annual corrosion rate of the environment as the influence radius, a sphere is constructed in the 3D space of the digital twin rail model. Within the environmental coupling action zone, all sub-regions that intersect with these spheres are identified. These intersecting sub-regions are extracted from the environmental coupling action zone and added to the preliminary life analysis focus region set. The merged overall region is defined as the final rail life management domain.

[0034] In practical implementation, load transfer paths are simulated on a digital twin rail model based on a dynamic load spectrum sequence. A representative load condition is selected from the dynamic load spectrum sequence. The selection criteria can be the combination of the mode axle load and the average speed appearing in the sequence, or the combination of an extreme axle load with a specific exceedance probability and its corresponding speed selected based on statistical distribution. The representative load condition includes a defined axle load value and a running speed value. In the digital twin rail model, the theoretical contact point between the wheel and the rail is used as the load application origin. The contact patch shape and pressure distribution are calculated based on Hertzian contact theory. Force boundary conditions corresponding to the representative load condition are applied, and velocity boundary conditions for load movement are set based on the running speed value to simulate the process of the wheel rolling over the rail surface. The transient dynamics solver of the digital twin rail model is run. The solver, based on an explicit or implicit time integration algorithm, calculates the dynamic process of stress wave propagation from the load application origin to the interior and ends of the rail model under the representative load condition, and records the amplitude decay of the stress field at various points in space over time. Based on the transmission process and spatial attenuation law of stress waves, the load energy transfer trajectory radiating outward from the load application origin is plotted. The load energy transfer trajectory is obtained by extracting stress cloud maps or vibration energy density cloud maps at equal time steps. The continuous section with the highest energy density is marked as the load core action area, while the transition section where the energy density decreases significantly is marked as the load edge influence area. The two are visualized and distinguished in the digital twin rail model with different colors or transparency.

[0035] The on-site inspection atlas of the target rail section is integrated to correct the boundaries of the load core action area and the load edge influence area. Surface damage optical images and internal defect ultrasonic images of the target rail section, acquired by inspection equipment, are obtained to jointly constitute the on-site inspection atlas. For example, surface damage optical images are acquired by vehicle-mounted or hand-held linear array cameras, while internal defect ultrasonic images are acquired by an ultrasonic flaw detection vehicle. The surface damage optical images are registered with the surface geometry of the digital twin rail model. The registration process is based on pre-deployed positioning markers or achieved through feature point matching algorithms to identify the actual three-dimensional coordinates of damage areas such as spalling, abrasions, and cracks in the images on the digital twin rail model. The internal defect ultrasonic images are registered with the internal structure of the digital twin rail model. The registration process uses the ultrasonic wave propagation time and probe position information, and an inversion algorithm to locate the three-dimensional coordinates of defects such as core damage and horizontal cracks within the digital twin rail model. The identified actual location and the located three-dimensional coordinates are superimposed and compared with the previously identified spatial range of the load core action area and the load edge influence area. The comparison process calculates the shortest Euclidean distance from each damage or defect point to the boundary of the load core action area and the load edge influence area.

[0036] If the actual location or three-dimensional coordinates are within a preset range of the load core action area or load edge influence area, the original boundary remains unchanged; if they are outside the preset range but in a nearby area, the boundary of the load core action area or load edge influence area is expanded accordingly to include the actual location or three-dimensional coordinates, thereby generating a preliminary set of focus areas for lifetime analysis. In some embodiments, the preset range can be determined by a distance threshold. Defined as the shortest distance between the damage point and the region boundary. satisfy If the damage point is within a preset range, it is determined to be within a nearby area, and a boundary expansion operation is triggered. The direction of boundary expansion is from the damage point to the geometric center of the original area. It can be understood that after boundary correction, the preliminary set of life analysis focus areas not only includes the theoretical high-stress area, but also encompasses the adjacent space of damage defects found in actual detection.

[0037] The regions adjacent to the initial lifetime analysis focus area set within the environmental coupling zone are merged. The geometric center coordinates of each region in the initial lifetime analysis focus area set are calculated. These coordinates are obtained by calculating the arithmetic mean of the coordinates of all voxels in the region. A sphere is constructed in the space of the digital twin rail model, centered on each geometric center coordinate and with a preset influence radius as its length. The influence radius is determined by referring to a table based on the rail material and environmental corrosion rate. Table 1 shows reference values ​​for the influence radius under various materials and typical environmental corrosion rates.

[0038] Table 1: Reference Values ​​of the Radius of Influence for Different Rail Materials and Environmental Corrosion Rates Within the environmental coupling zone, all sub-regions that intersect with these spherical spaces are identified. Intersection is determined by calculating the spatial relationship between the geometric envelope of the sub-regions and the spheres. The intersecting sub-regions are extracted from the environmental coupling zone and added to the initial life analysis focus region set. The merged overall region is defined as the final rail life management domain. In some embodiments, the merging operation may cause the originally separate initial life analysis focus regions to connect due to expansion; in this case, the connected region is considered a single entity. Optionally, the influence radius... Dynamic calculations can be performed using formulas: in: This represents the volume of the corresponding region in the set of focus regions for the preliminary lifetime analysis. Represents the rate of environmental corrosion; and This formula, based on coefficients determined according to material properties, provides a personalized method for determining the radius of influence that considers regional volume and corrosion rate. It can be understood that the final rail life management domain is a complete three-dimensional space integrating the theoretical high-stress zone, the actual damage point, and the surrounding area potentially affected by environmental corrosion.

[0039] See Figure 4 This is a comparison chart of the influence radii of rails made of different materials, showing the environmental coupling influence radii of three common rail materials (U71Mn, U75V, and PG4) under two different corrosion rates. For all materials, the higher the corrosion rate, the larger the influence radius. This indicates that environmental corrosion significantly expands the scope of the rail life management domain. At the same corrosion rate, U71Mn has the largest influence radius, while PG4 has the smallest, reflecting the differences in corrosion resistance and environmental sensitivity among different materials. This chart provides a direct basis for the division of the environmental coupling effect zone in digital twin rail life management. In areas with severe corrosion (0.05 mm / year), a larger area of ​​environmental coupling needs to be included in the life management domain to avoid missing potential corrosion damage. For materials like U71Mn, which are more sensitive to corrosion, a larger safety margin needs to be reserved when designing maintenance strategies.

[0040] In one embodiment of the present invention, an independent life evolution calculation task is configured for each partition in the final rail life management domain. The execution priority is set based on the geometric adjacency and mechanical correlation strength between the partitions. The final rail life management domain is subdivided into a three-dimensional mesh, making each subdivided mesh an independent partition. A corresponding material fatigue damage calculation model and creep evolution calculation model are defined for each partition; these two models together constitute the life evolution calculation task for that partition. The spatial Euclidean distance between any two partitions and the stress response correlation between the two partitions under historical load sequences are analyzed. This mechanical correlation strength is quantified by calculating the covariance of the stress time series of the two partitions. Two partitions with a spatial distance less than a set threshold and a calculated mechanical correlation strength higher than a set correlation value are defined as a strongly correlated partition pair. When setting the execution priority, the life evolution calculation tasks of two partitions belonging to the same strongly correlated partition pair must be scheduled for consecutive execution. Furthermore, within this strongly correlated partition pair, the life evolution calculation task corresponding to the partition with the higher average stress level derived from historical data analysis is prioritized for execution.

[0041] In practical implementation, the final rail life management domain is subdivided into meshes. Mesh subdivision uses tetrahedral or hexahedral elements to discretize the three-dimensional space of the final rail life management domain. Each subdivided mesh serves as an independent partition and is assigned a unique partition identifier; for example, a hexahedral element with dimensions of 10mm x 10mm x 5mm constitutes an independent partition. For each partition, a corresponding material fatigue damage calculation model and creep evolution calculation model are defined. The material fatigue damage calculation model is constructed based on the Miner linear cumulative damage criterion or the Paris formula based on fracture mechanics. The creep evolution calculation model is constructed based on the Norton power law or time hardening model. The material fatigue damage calculation model and the creep evolution calculation model together constitute the life evolution calculation task for the partition. In some embodiments, the specific parameters of the model are retrieved from the material property library of the digital twin rail model according to the rail material type corresponding to the partition location.

[0042] In practical implementation, the spatial distance between any two zones and the mechanical correlation strength between the two zones are analyzed. The spatial distance is obtained by calculating the Euclidean distance between the geometric center points of the two zones. The mechanical correlation strength is obtained by comparing the covariance of the stress response time series of the two zones under historical load spectrum sequences. It can be understood that the stress response time series is a data sequence of equivalent stress over time extracted from the historical transient dynamic simulation results of the digital twin rail model. The quantitative calculation of the mechanical correlation strength can follow the formula: in: This indicates the strength of the mechanical connection between partition a and partition b; and These represent the stress response time series vectors for partition a and partition b, respectively. This represents the covariance of two stress response time series vectors; and These represent the variances of the two stress response time series vectors, respectively. Two partitions with a spatial distance less than a set threshold and a mechanical correlation strength higher than a set correlation value are defined as strongly correlated partition pairs. The set threshold is determined proportionally based on the overall size of the final rail life management domain, and the set correlation value is selected based on the mechanical correlation strength distribution obtained from historical data analysis.

[0043] Each partition in the final rail life management domain is configured with an independent life evolution calculation task. The execution priority of these tasks is set based on the geometric adjacency and mechanical correlation strength between partitions. In specific implementations, the execution priority setting mandates that life evolution calculation tasks for two partitions belonging to the same strongly correlated partition pair must be executed consecutively. Furthermore, the life evolution calculation task corresponding to the partition with the higher stress level in the strongly correlated partition pair is executed first. The stress level is determined by comparing the peak or average equivalent stress of the two partitions under historical loads. In some embodiments, the setting process first identifies all strongly correlated partition pairs and generates task pairs. These task pairs are then inserted into a global task execution queue, with the insertion position sorted according to the stress level of the partition that is executed first in the task pair. For independent partitions that do not belong to any strongly correlated partition pair, their life evolution calculation tasks are sorted according to the average distance from their geometric center to all queued task partitions before being inserted into the queue. Optionally, the sorting of independent partitions follows the principle of executing closer partitions earlier. This setting method ensures that mechanically closely correlated and high-stress areas can complete life evolution calculations preferentially and continuously.

[0044] See Figure 5 This is a heatmap showing the mechanical correlation strength between 20 sub-regions in the rail life management domain. It visually displays the degree of mechanical correlation between these 20 sub-regions and serves as a key basis for prioritizing computational tasks in digital twin rail life management. The darker the color, the stronger the mechanical correlation between the two regions (closer to 1.0), indicating that their stress changes under load are highly synchronized and their mechanical behaviors significantly influence each other. The darker the color, the weaker the correlation (closer to 0.0), indicating that the mechanical behaviors of the two regions are relatively independent. This graph provides a direct basis for optimizing computational resources in the digital twin system. For the strongly correlated dark red areas, higher computational priority needs to be allocated, and the continuity of computational tasks should be ensured to guarantee the accuracy of life evolution simulation. For the weakly correlated blue areas, parallel computing strategies can be adopted to significantly improve overall simulation efficiency.

[0045] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A digital-twin-based intelligent management and control method for the whole life cycle of urban rail steel rails, characterized in that, include: Initialize a digital twin rail model that is spatially aligned with the physical rail, and mark multiple life-sensitive regions in the digital twin rail model based on mechanical properties; Obtain the unique code of the target rail section, and extract the dynamic load spectrum sequence related to train operation from the historical database based on the unique code; Based on the dynamic load spectrum sequence, the load transfer path is simulated on the digital twin rail model, thereby identifying the load core action area and the load edge influence area within the life-sensitive region. By integrating the field inspection maps of the target rail section, boundary corrections are performed on the load core action area and the load edge influence area to generate a preliminary set of life analysis focus areas; In the digital twin rail model, the area outside the preliminary life analysis focus area set is defined as the environmental coupling effect area, and the part of the environmental coupling effect area that is adjacent to the preliminary life analysis focus area set is merged with the preliminary life analysis focus area set to form the final rail life management domain. Based on the unique code, the dynamic load spectrum sequence related to train operation extracted from the historical database includes: Parse the unique code to obtain the operating line number, laying location mileage information, and rail material type of the line where the target rail section is located; Using the operating line number and the laying location mileage information as indexes, query the historical database for all train numbers passing through the target rail section; For each train number record retrieved, the axle load data, running speed data, and passing timestamp of the corresponding train are retrieved. The axle load data and running speed data are combined in the order of the passing timestamps to form the dynamic load spectrum sequence sorted by time. Based on the dynamic load spectrum sequence, the load transfer path simulated on the digital twin rail model includes: Select a representative load condition from the dynamic load spectrum sequence. The representative load condition includes axle load value and running speed value. In the digital twin rail model, the theoretical contact point between the wheel and the rail is used as the origin of load application, and force and velocity boundary conditions corresponding to the representative load conditions are applied. Run the transient dynamics solver of the digital twin rail model to calculate the transmission process and spatial attenuation law of stress waves inside the rail model under the representative load conditions. Based on the transmission process and spatial attenuation law of the stress wave, the load energy transmission trajectory radiating outward from the origin of the load application is plotted. The section with the highest energy density on the load energy transmission trajectory constitutes the core action area of ​​the load, and the transition section with significantly decreased energy density constitutes the edge influence area of ​​the load. By integrating the field inspection images of the target rail section, boundary correction is performed on the load core action area and the load edge influence area, including: The surface damage optical image and internal defect ultrasonic image of the target rail section are acquired by the detection equipment and together constitute the on-site detection map. The optical image of the surface damage is registered with the surface geometry of the digital twin rail model to identify the actual location of the damaged area in the image on the digital twin rail model. The ultrasonic imaging image of the internal defect is registered with the internal structure of the digital twin rail model to locate the three-dimensional coordinates of the defect in the digital twin rail model. The identified actual location and the located three-dimensional coordinates are superimposed and compared with the spatial range of the previously identified load core action area and load edge influence area; If the actual location or the three-dimensional coordinates are within a preset range of the load core action area or the load edge influence area, the original boundary remains unchanged; if they are outside the preset range but in their vicinity, the boundary of the load core action area or the load edge influence area is expanded accordingly to include the actual location or the three-dimensional coordinates, thereby generating the preliminary set of life analysis focus areas.

2. The digital-twin-based urban rail steel rail full-life-cycle intelligent management and control method according to claim 1, characterized in that, The method further includes: Each partition in the final rail life management domain is configured with an independent life evolution calculation task, and the execution priority order of the life evolution calculation task is set according to the geometric adjacency relationship and mechanical correlation strength between each partition. According to the execution priority order, the life evolution calculation task is executed sequentially for each partition in the final rail life management domain to generate the prediction result of the remaining service time of the partition; The remaining service life prediction results of all partitions are aggregated to drive the digital twin rail model to perform state evolution simulation and output maintenance decision instructions for the target rail section.

3. The digital-twin-based urban rail steel rail full-life-cycle intelligent management and control method according to claim 1, characterized in that, In the digital twin rail model, several life-sensitive regions are marked based on mechanical properties, including: Multiple stress calculation sampling points are preset on the geometry of the digital twin rail model; A standard load condition is applied to the digital twin rail model, and the stress response value of each stress calculation sampling point is calculated. The calculated stress response value is compared with a preset stress threshold. The continuous spatial range formed by stress calculation sampling points where the stress response value continuously exceeds the preset stress threshold is marked as an independent life-sensitive region.

4. The digital-twin-based urban rail steel rail full-life-cycle intelligent management and control method according to claim 1, characterized in that, The process of merging the portion of the environmental coupling zone adjacent to the preliminary lifetime analysis focus area set with the preliminary lifetime analysis focus area set includes: Calculate the coordinates of the geometric center point of each region in the preliminary lifetime analysis focus region set; Using the coordinates of each geometric center point as the center and the preset radius of influence as the length, a sphere is constructed in the space of the digital twin rail model. The radius of influence is obtained by looking up a table based on the rail material and the environmental corrosion rate. Within the environmental coupling zone, identify all sub-regions that intersect with the spherical space; The overlapping sub-regions are extracted from the environmental coupling zone and added to the preliminary life analysis focus region set. The merged overall region is defined as the final rail life management domain.

5. The intelligent management and control method for the entire life cycle of urban rail transit based on digital twins according to claim 2, characterized in that, Each partition in the final rail life management domain is configured with an independent life evolution calculation task, and the execution priority order of the life evolution calculation tasks is set according to the geometric adjacency relationship and mechanical correlation strength between the partitions, including: The final rail life management domain is further subdivided into grids, such that each subdivided grid is an independent partition; For each partition, a corresponding material fatigue damage calculation model and creep evolution calculation model are defined. The material fatigue damage calculation model and the creep evolution calculation model together constitute the lifetime evolution calculation task of the partition. Analyze the spatial distance between any two partitions and the mechanical correlation strength between the two partitions, which is obtained by comparing the stress covariance of the two partitions under historical loads; Two partitions whose spatial distance is less than a set threshold and whose mechanical correlation strength is higher than a set correlation value are defined as a strongly correlated partition pair. When setting the execution priority order, it is mandatory that the lifetime evolution calculation tasks of two partitions belonging to the same strongly correlated partition pair must be executed consecutively, and the lifetime evolution calculation task corresponding to the partition with higher stress level in the strongly correlated partition pair is executed first.

6. The intelligent management and control method for the entire life cycle of urban rail transit based on digital twins according to claim 5, characterized in that, The life evolution calculation task is performed sequentially on each partition in the final rail life management domain, including: Select the partition to be calculated according to the execution priority order; Load the material fatigue damage calculation model configured for the partition, and input the historical load spectrum segment experienced by the partition to calculate the cumulative fatigue damage increment of the partition under the current load cycle; Simultaneously, the creep evolution calculation model configured for the partition is loaded, and the historical data of the average stress and ambient temperature of the partition are input to calculate the creep strain increment of the partition at the current time step. Based on the calculated cumulative fatigue damage increment and creep strain increment, the material state parameters of the partition in the digital twin rail model are updated; Based on the updated material state parameters and the fracture toughness curve of the partition material, the theoretical number of failure cycles or time of the partition under the current operating load conditions is deduced, and the theoretical number of failure cycles or time is used as the prediction result of the remaining service life of the partition.

7. The intelligent management and control method for the entire life cycle of urban rail transit based on digital twins according to claim 2, characterized in that, The remaining service life prediction results of all zones are aggregated to drive the digital twin rail model to perform state evolution simulation, and the maintenance decision instructions for the target rail section are output, including: Collect the remaining service life prediction results of all partitions in the final rail life management domain to form a remaining life distribution list; In the list of remaining service life distributions, find the remaining service life prediction result with the smallest value and mark it as the life prediction value of the weakest zone; In the digital twin rail model, the spatial location corresponding to the weakest section is located, and the type of rail structure component to which the spatial location belongs is retrieved; Based on the predicted lifespan of the weakest section and the type of rail structure component, a preset maintenance strategy mapping table is queried. The maintenance strategy mapping table defines the specific maintenance actions and maintenance timings corresponding to different lifespan ranges and component types. The maintenance decision instruction is generated by matching the corresponding specific maintenance action and maintenance timing from the maintenance strategy mapping table.

Images