Track line infrastructure degradation trend analysis method based on digital twinning

Abstract

The invention relates to the technical field of digital twinning, in particular to a track line infrastructure degradation trend analysis method based on digital twinning. The method comprises the following steps: establishing a digital twin model of track multi-modal state features based on track multi-source heterogeneous data, and generating a track multi-modal state feature model; performing track structure dynamic response feature analysis based on the track multi-modal state feature model to generate track structure dynamic response feature data; performing track degradation behavior characteristic analysis on the track structure dynamic response characteristic data to generate track degradation behavior characteristic data; establishing an orbit degradation trend prediction model based on the orbit degradation behavior characteristic data; and transmitting the track structure dynamic response characteristic data to a track degradation trend prediction model for track facility degradation trend prediction processing, and generating track facility degradation trend prediction data. According to the invention, the intelligent analysis of the degradation trend of the track line infrastructure is realized.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a method for analyzing the deterioration trend of railway infrastructure based on digital twins. Background Technology

[0002] With the rapid development of the rail transit industry, the mileage of rail lines continues to increase, and the operational load is constantly rising. The safe and stable operation of rail line infrastructure has become a core prerequisite for ensuring the operational efficiency and passenger safety of rail transit. Rail facilities in different scenarios are affected by multiple deterioration factors, and are prone to deterioration problems such as rail wear, track bed compaction, and sleeper cracking. Deterioration trend analysis of rail line infrastructure is a key technology for achieving intelligent operation and maintenance and proactively avoiding risks. Its accuracy and timeliness directly determine the scientific nature of operation and maintenance decisions and the service life of rail facilities. However, existing technologies for analyzing the deterioration trends of rail transit infrastructure still rely on manual inspections, periodic testing, or data from single-type sensors. These technologies can only obtain local static state information, ignoring the inherent correlation between multi-source heterogeneous data and the dynamic evolution characteristics of the deterioration process. Furthermore, they lack effective multi-scale fusion and spatial mapping methods, making it difficult to accurately extract the deterioration correlation features contained in heterogeneous data. This results in the inability to construct facility state models that fit actual working conditions, and even more so, the difficulty in dynamically modeling and mapping the multimodal states of rail transit facilities in real time. Moreover, existing prediction models often fail to consider the causal relationship between external driving factors and the deterioration state, relying only on simple fitting of historical data. This fails to effectively capture hidden deterioration trends and the coupling effects of multiple factors, ultimately leading to insufficient accuracy in predicting the deterioration trends of rail transit infrastructure. Summary of the Invention

[0003] Based on this, the present invention provides a method for analyzing the deterioration trend of railway infrastructure based on digital twins, in order to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method for analyzing the deterioration trend of rail transit infrastructure based on digital twins includes the following steps: Step S1: Collect multi-source heterogeneous data of the track using multi-source sensors deployed on the track infrastructure to generate multi-source heterogeneous track data; establish a digital twin model of the track multi-modal state characteristics based on the multi-source heterogeneous track data to generate a multi-modal track state characteristic model. Step S2: Perform dynamic response feature analysis of the track structure based on the track multimodal state feature model to generate dynamic response feature data of the track structure; Step S3: Perform track degradation behavior characteristic analysis on the dynamic response characteristic data of the track structure to generate track degradation behavior characteristic data; Step S4: Establish a track deterioration trend prediction model based on track deterioration behavior characteristic data; transmit the dynamic response characteristic data of the track structure to the track deterioration trend prediction model for track facility deterioration trend prediction processing, and generate track facility deterioration trend prediction data.

[0005] Furthermore, the multi-source heterogeneous track data mentioned in step S1 includes track geometry data, sleeper-track bed structural stress data, rail vibration acceleration data, track environment data, and train operation load data.

[0006] Furthermore, step S1 includes the following steps: Step S11: Collect multi-source heterogeneous data of the track by using multi-source sensors deployed on the track infrastructure to generate multi-source heterogeneous data of the track. Step S12: Perform multi-scale fusion state feature analysis on the multi-source heterogeneous data of the orbit to generate heterogeneous time-series state feature data of the orbit; Step S13: Perform spatial mapping processing on the heterogeneous time-series state feature data of the orbit to generate mapped heterogeneous orbit state feature data; Step S14: Establish a digital twin model of the orbital multimodal state features by mapping the heterogeneous state feature data of the orbit, and generate the orbital multimodal state feature model.

[0007] Furthermore, step S12 includes the following steps: Step S121: Design a multi-scale heterogeneous coding strategy and a multi-scale feature analysis strategy for orbits based on multi-source heterogeneous orbital data; Step S122: Use the orbital multi-scale heterogeneous coding strategy to perform multi-scale coding preprocessing on the orbital heterogeneous data to generate orbital multi-scale heterogeneous data; Step S123: Perform multi-scale heterogeneous time series data analysis on the multi-scale heterogeneous orbital data to generate multi-scale heterogeneous orbital time series data; Step S124: Use the orbital multi-scale feature analysis strategy to perform multi-scale heterogeneous time series feature analysis on the orbital multi-scale time series data to generate orbital multi-scale heterogeneous time series feature data; Step S125: Perform multi-scale fusion state feature analysis on the multi-scale heterogeneous time series feature data of the orbit to generate heterogeneous time series state feature data of the orbit.

[0008] Furthermore, step S2 includes the following steps: Step S21: Perform finite element mesh generation on the track multimodal state characteristic model to generate finite element mesh track multimodal state characteristic data; Step S22: Based on the multimodal state characteristic data of the track using finite element mesh, perform stress and strain distribution characteristic analysis on the track structure to generate stress-strain distribution characteristic data of the track structure; Step S23: Analyze the hidden strain trend inside the track based on the stress-strain distribution characteristic data of the track structure, and generate hidden strain trend data inside the track. Step S24: Analyze the dynamic response characteristics of the track structure based on the stress-strain distribution characteristics data of the track structure and the implicit strain trend data inside the track, and generate dynamic response characteristic data of the track structure.

[0009] Furthermore, step S22 includes the following steps: Step S221: Perform boundary condition analysis on the finite element mesh track based on the multimodal state characteristic data of the track, and generate boundary condition data for the track; Step S222: Perform dynamic load characteristic analysis on the finite element mesh track based on the multimodal state characteristic data of the track, and generate dynamic load characteristic data of the track. Step S223: Analyze the distribution characteristics of stress and strain of the track structure based on the grid track boundary condition data and grid track dynamic load characteristic data, and generate stress-strain distribution characteristic data of the track structure.

[0010] Furthermore, step S3 includes the following steps: Step S31: Perform track structure response residual analysis based on the track structure dynamic response characteristic data to generate track structure response residual data; Step S32: Perform residual sequence analysis on the track structure response residual data to generate track structure response residual sequence data; Step S33: Perform abnormal feature analysis of track structure response residuals based on track structure response residual sequence data, and generate abnormal feature data of track structure response residuals; Step S34: Analyze the track deterioration behavior characteristics based on the abnormal characteristic data of track structure response residuals, and generate track deterioration behavior characteristic data.

[0011] Furthermore, step S33 includes the following steps: Step S331: Perform random fluctuation analysis of track structure response residuals based on the track structure response residual sequence data to generate random fluctuation data of track structure response residuals; Step S332: Extract abnormal fluctuations of track structure response residuals from the random fluctuation data of track structure response residuals to obtain abnormal fluctuation data of track structure response residuals; Step S333: Perform abnormal characteristic analysis of track structure response residuals based on the abnormal fluctuation data of track structure response residuals, and generate abnormal characteristic data of track structure response residuals.

[0012] Furthermore, step S4 includes the following steps: Step S41: Perform external driving factor analysis on the track based on the dynamic response characteristic data of the track structure, and generate external driving factor data of the track; Step S42: Perform causal correlation processing on external driving factor degradation based on orbital external driving factor data and orbital degradation behavior characteristic data to generate causal correlation data on external driving factor degradation. Step S43: Perform track degradation state specificity analysis based on track degradation behavior characteristic data to generate track degradation state specificity data; Step S44: Based on the orbital degradation state-specific data and the causal correlation data of external driving factor degradation, perform degradation impact characteristic analysis of external driving factor coupling, and generate factor coupling degradation impact characteristic data; Step S45: Analyze the spatial characteristics of the orbital degradation state evolution based on the orbital degradation state specific data to generate spatial characteristic data of the orbital degradation state evolution; Step S46: Establish the relationship between the external driving factors and the orbital degradation trend prediction by using the factor coupling degradation influence feature data and the orbital degradation state evolution spatial feature data, and generate an orbital degradation trend prediction model. Step S47: Transmit the dynamic response characteristic data of the track structure to the track deterioration trend prediction model for track facility deterioration trend prediction processing, and generate track facility deterioration trend prediction data.

[0013] Furthermore, step S44 includes the following steps: Step S441: Based on the orbital degradation state specific data and the external driving factor degradation causal correlation data, perform adaptive correlation feature analysis on external driving factors and degradation state to generate external driving factor-degradation state adaptive correlation feature data. Step S442: Perform degradation impact feature analysis on the external driving factor-degradation state adaptive correlation feature data to generate degradation impact feature data of external driving factor; Step S443: Perform degradation impact characteristic analysis of factor coupling based on the degradation impact characteristic data of external driving factors, and generate degradation impact characteristic data of factor coupling.

[0014] The beneficial effects of this application are as follows: This invention comprehensively collects multi-source heterogeneous data, including track geometry, sleeper-ballast structure stress, rail vibration acceleration, track environment, and train operation load, through multi-source sensors, overcoming the limitations of single data acquisition and providing full-dimensional, high-fidelity data support for subsequent analysis. Simultaneously, through multi-scale fusion, spatial mapping, and digital twin modeling processes, especially the progressive processing of multi-scale heterogeneous coding, temporal analysis, and feature fusion, it achieves in-depth mining of multi-source heterogeneous data from coding preprocessing to correlation feature extraction. This solves the problem of existing technologies lacking effective data fusion methods and struggling to capture the correlation features of heterogeneous data. The constructed track multimodal state feature model can accurately map the real working conditions and multi-dimensional states of track facilities, laying a precise model foundation for subsequent dynamic response analysis. By visualizing the model through finite element mesh generation and combining boundary conditions with dynamic load characteristic analysis, the stress-strain distribution data of the track structure is accurately obtained. Simultaneously, the implicit strain trend within the track is deeply explored, achieving a full-dimensional analysis from the surface response to the internal implicit characteristics. Compared to existing technologies that can only obtain local static state information, this layered decomposition analysis not only accurately captures the dynamic response law of the track structure under load and environmental influences but also provides targeted and highly accurate core data for subsequent degradation behavior characteristic analysis. Through a progressive residual analysis process, the pain point of existing technologies being unable to accurately identify implicit degradation characteristics is effectively solved, demonstrating a significant advantage in analytical depth. Based on dynamic response characteristic data, through layered processing of residual analysis, residual sequence analysis, and anomaly feature extraction, especially the refined analysis of residual random fluctuations, it is clarified that conventional residual fluctuations are an inevitable result of the track facility's normal operation. Accurate extraction and feature analysis are performed only on residuals deviating from this conventional fluctuation range, enabling precise capture of minute anomalies caused by degradation in the track structure response, achieving early identification and accurate characterization of degradation behavior characteristics. This design overcomes the limitations of traditional methods, which can only identify explicit degradation and struggle to distinguish between normal operational fluctuations and abnormal degradation fluctuations, effectively avoiding the problem of misjudging normal operational fluctuations as degradation characteristics. Simultaneously, its standardized analysis process ensures the reliability and relevance of degradation behavior characteristic data, providing high-quality core input for trend prediction model construction and achieving seamless integration of dynamic response analysis and degradation feature mining, further improving the entire process analysis system. Through refined model construction and multi-factor coupling analysis, the accuracy and foresight of degradation trend prediction are significantly improved, effectively compensating for the shortcomings of existing technology prediction models in terms of detail and neglect of factor correlation. Not only is the influence mechanism of multiple factors on track degradation clarified through external driving factor analysis and causal correlation construction of degradation states, but also the coupling effect law of multiple driving factors is accurately captured through refined processes such as adaptive correlation analysis and factor coupling effect analysis.Meanwhile, the prediction model constructed by combining the spatial characteristics of the degradation state evolution can fully integrate dynamic response data, degradation behavior characteristics and factor coupling effects to achieve accurate prediction of the degradation trend of track facilities, providing forward-looking and scientific technical support for rail transit operation and maintenance decisions, and helping to achieve early prevention and control of degradation risks.

[0015] Therefore, the digital twin-based method for analyzing the degradation trend of railway infrastructure in this invention precisely addresses the shortcomings of existing technologies, possessing significant technical advantages and application value. This method collects multi-source heterogeneous data from multiple sensors, combining multi-scale fusion and spatial mapping techniques to effectively overcome the limitations of single data acquisition. It achieves accurate extraction of the inherent correlation characteristics of multi-source heterogeneous data, thereby constructing a digital twin model of the multi-modal state characteristics of the track that closely reflects actual operating conditions. This model can dynamically map the full-dimensional state of the track facilities in real time, compensating for the inaccuracies of traditional static analysis and modeling. Simultaneously, by deeply analyzing the dynamic response characteristics and degradation behavior characteristics of the track structure, a trend prediction model integrating external driving factors and the causal relationship of degradation states is established. This abandons the crude prediction method that relies solely on historical data fitting, accurately capturing hidden degradation trends and the coupling effects of multiple factors, significantly improving the prediction accuracy and foresight of track facility degradation trends. Overall, this method constructs a closed-loop analysis system covering the entire process from data collection, dynamic modeling, feature analysis to trend prediction, providing scientific and reliable technical support for intelligent operation and maintenance of rail transit. It helps to optimize operation and maintenance decisions, extend the service life of rail facilities, and effectively ensure the safety and efficiency of rail line operation. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the steps in the method for analyzing the deterioration trend of railway infrastructure based on digital twins according to the present invention. Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S4. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. Functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods. The term "and / or" as used herein includes any and all combinations of one or more of the associated items listed.

[0019] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a method for analyzing the deterioration trend of rail transit infrastructure based on digital twins. In the embodiments of this invention, please refer to... Figure 1 The diagram shown is a flowchart illustrating the steps of a digital twin-based method for analyzing the deterioration trend of rail transit infrastructure according to the present invention. The digital twin-based method includes the following steps: Step S1: Collect multi-source heterogeneous data of the track using multi-source sensors deployed on the track infrastructure to generate multi-source heterogeneous track data; establish a digital twin model of the track multi-modal state characteristics based on the multi-source heterogeneous track data to generate a multi-modal track state characteristic model. In this embodiment of the invention, multi-source sensor arrays are first deployed in key sections of the track infrastructure to comprehensively cover core areas such as rails, sleepers, track bed, and the surrounding environment, forming a data acquisition network without blind spots. Corresponding types of sensors are deployed at key locations on the rails, sleepers, and track bed to collect data on track geometry, structural stress, vibration response, etc. Dedicated monitoring modules are deployed around the track and on the train bogies to simultaneously collect environmental parameters and train operating load data. All sensors are time-synchronized to ensure consistency in the spatiotemporal dimensions of various data, generating multi-source heterogeneous track data encompassing multiple dimensions. Based on this data, a digital twin model of the track's multimodal state characteristics is established. First, a three-dimensional spatial coordinate system for the track is constructed, and spatial units are divided according to the actual segmentation of the track. Multi-source heterogeneous data is then accurately mapped to corresponding units according to their spatial location, realizing the transformation of data from a temporal dimension to a spatial dimension. Subsequently, three core modules were built: geometric mode, physical mode, and state mode. The geometric mode replicates the actual shape and spatial position relationship of each structure of the track. The physical mode integrates the material properties and mechanical characteristics of each structure. The state mode associates the mapped data with the geometric mode in real time, realizing the collaborative mapping of multiple modes of track geometry, structural stress, vibration response, etc. Finally, a multimodal state feature model of the track is generated. The model is updated in real time according to the sensor acquisition frequency, accurately reproducing the real-time operating status of the track facilities.

[0020] Step S2: Perform dynamic response feature analysis of the track structure based on the track multimodal state feature model to generate dynamic response feature data of the track structure; In this embodiment of the invention, a multimodal state characteristic model of the track is used as the core to systematically analyze the dynamic response characteristics of the track structure, comprehensively capturing the mechanical response and state change laws of the track under different operating conditions. First, a finite element mesh is generated for the digital twin model, covering the core structures such as rails, sleepers, track bed, and contact interfaces. The contact interface mesh is densified to ensure accurate mapping of force transmission at the structural contact points. After meshing, the material properties and geometric information of each structure are associated with the corresponding mesh elements to generate a finite element mesh model with attribute identifiers. Boundary constraints are set based on the actual track installation and fixing method, and corresponding constraint types are set for different structural stress characteristics to limit non-permissible displacements while preserving reasonable deformation space. Damping constraints are also set at the track ends to reduce the interference of boundary reflections on the analysis results. Combining previously collected train operation load data, the load is accurately applied to the corresponding mesh elements according to the actual wheel-rail contact position. Simultaneously, the load intensity is adjusted based on environmental factors to match the impact of environmental changes on the track's mechanical response. By analyzing the stress and strain characteristics of each grid element using professional mechanical calculation methods, the mechanical response of easily deteriorated areas is focused on capturing subtle structural deformations and stress concentration phenomena. Implicit strain trends are analyzed simultaneously, and stress and strain distribution data and implicit strain trend data are integrated. The characteristics are sorted out from multiple dimensions such as response amplitude, frequency and correlation characteristics, and finally dynamic response characteristic data of the track structure are generated, providing comprehensive data support for subsequent deterioration analysis.

[0021] Step S3: Perform track degradation behavior characteristic analysis on the dynamic response characteristic data of the track structure to generate track degradation behavior characteristic data; In this embodiment of the invention, based on the dynamic response characteristic data of the track structure, and combined with the track deterioration mechanism and typical forms, a systematic analysis of track deterioration behavior characteristics is conducted to accurately characterize the type, degree, and distribution pattern of deterioration. First, track structure response residual analysis is performed to extract core indicators from the dynamic response characteristic data. This is compared with the theoretical response values ​​under the corresponding operating conditions of the digital twin model. The deviation between the two is calculated, and the corresponding spatiotemporal information and operating condition parameters are correlated to generate track structure response residual data. The residual data undergoes both temporal and spatial sequence analysis. The residual data is concatenated at a fixed period to form a temporal sequence. Spatial residual information is integrated by dividing the track into segments, calculating sequence characteristic parameters, and eliminating outliers caused by sensor failures to ensure sequence continuity and effectiveness. Based on the residual sequence data, abnormal fluctuation analysis is conducted. First, the normal fluctuation range of each indicator residual is defined. Abnormal fluctuations exceeding the range are screened, and the spatiotemporal location, duration, and trend of abnormal fluctuations are recorded. Simultaneously, operating condition and environmental parameters are correlated to analyze the correlation characteristics between abnormal fluctuations and external factors. By establishing a correspondence between abnormal fluctuation characteristics and known degradation types, quantifying the degree of degradation according to the abnormal exceedance, tracing the operating condition data during abnormal periods to identify the dominant degradation factors, estimating the degradation initiation time and development rate, and statistically analyzing the spatial distribution range and core area of ​​each degradation type, the final result is generated track degradation behavior characteristic data, which fully covers the core information of degradation.

[0022] Step S4: Establish a track deterioration trend prediction model based on track deterioration behavior characteristic data; transmit the dynamic response characteristic data of the track structure to the track deterioration trend prediction model for track facility deterioration trend prediction processing, and generate track facility deterioration trend prediction data.

[0023] In this embodiment of the invention, a track degradation trend prediction model is constructed based on track degradation behavior characteristic data to ensure that the model can accurately reflect the degradation evolution law and the influence mechanism of external factors. First, the external driving factors affecting track degradation are analyzed, identifying core factors of load and environment, quantifying the degree, scope, and temporal characteristics of each factor's influence on different degradation types, establishing the causal relationship between factors and degradation behavior, and clarifying the influence law of single factors and the coupled superposition effect of multiple factors. Combining the specificity of track degradation state and spatial evolution characteristics, a spatiotemporal dual prediction framework is constructed. Prediction nodes are set at fixed periods in the time dimension, and accurate predictions are carried out in the spatial dimension according to the line segment grid units. Key parameters such as factor influence coefficients and degradation diffusion rates are integrated into the prediction relationship, and specific prediction logic is established for different degradation types. Model parameters are calibrated using historical degradation data to correct prediction errors, and a real-time update mechanism is set to ensure that the model can incorporate the latest operating data to dynamically optimize the prediction relationship, ultimately generating a track degradation trend prediction model. Based on this model, we carry out the prediction and processing of track facility deterioration trends, extract real-time indicators from the dynamic response feature data of track structure, associate the historical deterioration status of the corresponding area with the effects of external factors, substitute them into the model to perform multi-period prediction, output the deterioration type, degree and distribution range of each spatial unit in different time periods, and simultaneously label the confidence level of the prediction results, and finally generate track facility deterioration trend prediction data to provide forward-looking support for track operation and maintenance decisions.

[0024] Furthermore, the multi-source heterogeneous track data mentioned in step S1 includes track geometry data, sleeper-track bed structural stress data, rail vibration acceleration data, track environment data, and train operation load data.

[0025] Furthermore, step S1 includes the following steps: Step S11: Collect multi-source heterogeneous data of the track by using multi-source sensors deployed on the track infrastructure to generate multi-source heterogeneous data of the track. In this embodiment of the invention, a multi-source sensor array is deployed in key sections of the track infrastructure, covering core areas such as rails, sleepers, track bed, and the surrounding environment. A laser displacement sensor is installed every 20 meters along the longitudinal direction of the rail to collect track geometry data, focusing on capturing data on rail elevation, gauge, and horizontal deviation. Fiber optic stress sensors are embedded in the contact surface between the sleeper and the track bed, with two sensors installed on each sleeper, located at the two ends of the sleeper at approximately 1 / 4 of their length, to collect stress data on the sleeper-track bed structure, with a measurement range of 0-20 MPa and a data update cycle of 50 ms. Piezoelectric accelerometers are installed on the rail web to collect rail vibration acceleration data, with a measurement range of ±50g and a frequency response range of 0. 0.1Hz-10kHz; Environmental monitoring modules are deployed every 50 meters on both sides of the line to synchronously collect temperature, humidity and rainfall data. The temperature measurement range is -40℃ to 60℃, and the rainfall acquisition resolution is 0.1mm; Pressure sensors are installed on the train bogies to collect train running load data, focusing on recording axle load and wheel-rail interaction force. The axle load measurement range is 0-30t, and the wheel-rail interaction force acquisition frequency is consistent with that of the acceleration sensor. All sensor data acquisition is time-synchronized to generate multi-source heterogeneous track data containing five types of data.

[0026] Step S12: Perform multi-scale fusion state feature analysis on the multi-source heterogeneous data of the orbit to generate heterogeneous time-series state feature data of the orbit; In this embodiment of the invention, a multi-scale heterogeneous coding strategy and a multi-scale feature analysis strategy for track data are designed to address the characteristics of multi-source heterogeneous track data. The coding strategy divides data into three scale levels based on data type: geometric data is encoded at the millimeter scale, stress data at the micrometer scale, vibration acceleration data at the microsecond scale, and environmental and load data at the second scale. During the coding process, data of different scales are normalized: geometric data is normalized to the 0-10 range, stress data to the 0-2 range, vibration acceleration data to the 0-5 range, and environmental and load data to the 0-1 range, generating multi-scale heterogeneous track data. Subsequently, a sliding window time series analysis is performed on the multi-scale heterogeneous track data with a 10-second time window, extracting the maximum, minimum, and mean values ​​of various data types within each window to form multi-scale heterogeneous track time series data. A multi-scale feature analysis strategy is implemented using wavelet decomposition algorithm to perform three-level wavelet decomposition on heterogeneous time-series data of orbits. Low-frequency components and high-frequency components in the first two levels are extracted. The low-frequency components reflect the long-term trend of data changes, while the high-frequency components reflect instantaneous fluctuation characteristics. Then, through feature dimension alignment technology, the feature vectors of different types of data are unified to the same dimension, realizing multi-scale fusion state feature analysis and processing of heterogeneous data to generate heterogeneous time-series state feature data of orbits.

[0027] Step S13: Perform spatial mapping processing on the heterogeneous time-series state feature data of the orbit to generate mapped heterogeneous orbit state feature data; In this embodiment of the invention, a three-dimensional spatial coordinate system for the track is established, with the starting point of the track as the origin, the X-axis along the track extension direction, the Y-axis perpendicular to the track direction, and the Z-axis perpendicular to the ground direction. The coordinate accuracy is controlled within ±0.001m. The track is divided into several spatial units in 20-meter increments, with each spatial unit corresponding to a unique coordinate interval. Based on this coordinate system, a heterogeneous temporal state characteristic spatial mapping rule is designed. Track geometric data is mapped to spatial nodes corresponding to the XYZ three-dimensional coordinates, and each geometric feature index corresponds to the attribute value of a specific coordinate node. The sleeper-track bed structural stress data is bound to the spatial coordinates according to the sleeper number, and the stress data of each sleeper is mapped to the center coordinate point of its spatial unit. Rail vibration acceleration data is mapped according to the sensor installation location coordinates, synchronously associated with the vibration characteristics of the corresponding spatial unit. Environmental data and train operation load data are mapped to the corresponding spatial units according to the acquisition location coordinates, and the load data is additionally associated with the X-axis coordinate interval corresponding to the train operation trajectory. The mapping process employs a spatial interpolation algorithm to supplement the features of spatial units with missing data, ensuring that each spatial unit has complete heterogeneous temporal state feature data. Ultimately, it generates heterogeneous state feature data of the mapped track, realizing the transformation of data from the temporal dimension to the spatial dimension, and providing spatial data support for the construction of digital twin models.

[0028] Step S14: Establish a digital twin model of the orbital multimodal state features by mapping the heterogeneous state feature data of the orbit, and generate the orbital multimodal state feature model.

[0029] In this embodiment of the invention, a digital twin model of track multimodal state characteristics is constructed based on mapped heterogeneous track state characteristic data. The model includes three core modules: geometric mode, physical mode, and state mode. The geometric mode, based on mapped spatial coordinate data, replicates the three-dimensional geometry of the track line, accurately restoring the structural dimensions and spatial relationships of the rails, sleepers, and track bed. The rail cross-sectional dimensions are constructed according to the parameters of a standard 60kg / m rail, the sleepers are replicated according to the dimensions of concrete sleepers, and the track bed thickness is controlled at 30cm. The error between the geometric mode and the actual track line does not exceed 0.01m. The physical mode integrates the track material properties and structural mechanical characteristics. For example, the rails use Q235 steel parameters, with an elastic modulus set at 206GPa and a Poisson's ratio of 0.3; the sleepers use C50 concrete parameters, with an elastic modulus of 34.5GPa; and the track bed uses graded crushed stone parameters with a density of 1800kg / m³. The physical mode is calibrated using stress and vibration characteristics from the mapped data to ensure consistency with the actual track physical characteristics. The state mode maps the heterogeneous state characteristic data of the track to the corresponding spatial unit of the geometric mode in real time. The state index of each spatial unit changes dynamically with the data update, realizing the collaborative mapping of multi-modal states under track geometry, structural stress, vibration response, environmental conditions and load. Finally, a multi-modal state characteristic model of the track is generated. The model update frequency is synchronized with the sensor data acquisition frequency, accurately reproducing the real-time operating status of the track infrastructure.

[0030] Furthermore, step S12 includes the following steps: Step S121: Design a multi-scale heterogeneous coding strategy and a multi-scale feature analysis strategy for orbits based on multi-source heterogeneous orbital data; In this embodiment of the invention, the types, magnitudes, and variation characteristics of multi-source heterogeneous track data are analyzed, and targeted multi-scale heterogeneous coding strategies and multi-scale feature analysis strategies for track are designed. The coding strategy divides the data into three scale levels based on their physical attributes. Geometric morphology data, due to its millimeter-level measurement accuracy, uses millimeter-level coding with a 16-bit code length. The high 8 bits represent the spatial partition number, and the low 8 bits represent the specific measured value. Stress data, with micrometer-level measurement accuracy, uses micrometer-level coding with a 24-bit code length. The high 10 bits represent the sleeper number, the middle 8 bits represent the stress range, and the low 6 bits represent the precise value. Vibration acceleration data, with its high frequency of change, uses microsecond-level coding with a 32-bit code length. The high 12 bits represent the timestamp, the middle 8 bits represent the acceleration level, and the low 12 bits represent the instantaneous fluctuation value. Environmental and load data, with relatively long change cycles, use second-level coding with a 20-bit code length. The high 10 bits represent the acquisition location number, and the low 10 bits represent the measured value. The multi-scale feature analysis strategy explicitly adopts the wavelet decomposition algorithm, sets a three-level decomposition hierarchy, sets a low-frequency component retention threshold of 0.05Hz and a high-frequency component screening threshold of 1kHz, and determines the feature dimension alignment standard, uniformly standardizing the feature vector dimensions of various data to 256 dimensions to ensure consistency in subsequent feature analysis and fusion.

[0031] Step S122: Use the orbital multi-scale heterogeneous coding strategy to perform multi-scale coding preprocessing on the orbital heterogeneous data to generate orbital multi-scale heterogeneous data; In this embodiment of the invention, based on the designed multi-scale heterogeneous coding strategy for tracks, multi-scale coding preprocessing is performed on multi-source heterogeneous track data. Geometric data is converted according to millimeter-level coding rules, mapping rail height, gauge, and horizontal deviation values ​​to 16-bit codes. Spatial partition numbers are assigned sequentially for every 20 meters of track, and specific measured values ​​are converted to binary codes with an accuracy of ±0.01mm. Stress data is converted according to 24-bit micrometer-level coding, sleeper numbers are assigned continuously from the starting point according to the track laying sequence, stress ranges are divided into 128 intervals from 0-20MPa, and precise values ​​are supplemented according to measurement accuracy. Vibration acceleration data is converted according to 32-bit microsecond-level coding, timestamps are accurate to milliseconds and extended to microsecond units, acceleration measures are divided into 256 levels with ±50g, and instantaneous fluctuation values ​​are calculated and converted according to frequency response range. Environmental and load data are converted according to 20-bit second-level coding, with acquisition location numbers corresponding to sensor deployment locations, and measured values ​​converted to coding forms according to their respective accuracy standards. Normalization is performed synchronously during the encoding process, compressing geometric data to the 0-10 range, stress data to the 0-2 range, vibration acceleration data to the 0-5 range, and environmental and load data to the 0-1 range respectively, eliminating interference from data of different magnitudes and generating heterogeneous multi-scale orbital data.

[0032] Step S123: Perform multi-scale heterogeneous time series data analysis on the multi-scale heterogeneous orbital data to generate multi-scale heterogeneous orbital time series data; In this embodiment of the invention, a sliding window time series analysis is performed on multi-scale heterogeneous track data with a fixed time window of 10 seconds. The sliding window step size is set to 2 seconds to ensure the continuity and coverage of the time series data. Within each time window, for geometric data, the maximum value of rail height deviation, the minimum value of track gauge, and the average value of horizontal deviation are extracted, and the timestamp and acquisition location code of the corresponding data are recorded simultaneously. For sleeper-track bed structural stress data, the maximum value, minimum value, and interval fluctuation difference of stress at the sensors at both ends of each sleeper are extracted and associated with the sleeper number and spatial partition information. For rail vibration acceleration data, the peak acceleration, root mean square acceleration, and peak vibration frequency within each window are extracted and associated with the sensor installation location coordinates. For environmental data, the average temperature, average humidity, and cumulative rainfall are extracted and bound to the acquisition points on both sides of the track. For train operation load data, the average axle load, peak wheel-rail force, and load duration are extracted and associated with the X-axis coordinate interval corresponding to the train operation trajectory. After the various data features within each window are extracted, they are integrated in order of encoding bit length to form multi-scale heterogeneous time-series data of orbits containing time, spatial and feature dimensions. Each window corresponds to a complete set of time-series data to ensure the temporal correlation of the data.

[0033] Step S124: Use the orbital multi-scale feature analysis strategy to perform multi-scale heterogeneous time series feature analysis on the orbital multi-scale time series data to generate orbital multi-scale heterogeneous time series feature data; In this embodiment of the invention, the wavelet decomposition algorithm determined in step S121 is used to perform multi-scale heterogeneous time series feature analysis on the track multi-scale time series data according to a three-level decomposition hierarchy. The first level of decomposition decomposes the time series data into high-frequency components of 1Hz-10kHz and low-frequency components of 0.1Hz-1Hz, and selects and retains vibration features with amplitude greater than 0.5g in the high-frequency components and geometric features with change rate greater than 0.1mm / h in the low-frequency components; the second level of decomposition further decomposes the high-frequency components of the first level to obtain two frequency band components of 5Hz-10kHz and 1Hz-5Hz, and extracts the instantaneous impact features of rail vibration in the 5Hz-10kHz frequency band and the sleeper stress fluctuation features in the 1Hz-5Hz frequency band; the third level of decomposition decomposes the high-frequency components of the second level to obtain frequency band components of 8Hz-10kHz and 5Hz-8kHz, and extracts the wheel-rail contact vibration features in the 8Hz-10kHz frequency band. After decomposition, the low-frequency components and the first two high-frequency components are retained, while the noise interference in the third high-frequency components is discarded. The retained components are then quantized to convert indicators such as vibration frequency, stress fluctuation amplitude, and geometric change rate into specific values, generating multi-scale heterogeneous time-series characteristic data of the orbit. Each characteristic indicator corresponds to a unique scale level and data type identifier.

[0034] Step S125: Perform multi-scale fusion state feature analysis on the multi-scale heterogeneous time series feature data of the orbit to generate heterogeneous time series state feature data of the orbit.

[0035] In this embodiment of the invention, multi-scale fusion state feature analysis is performed on the heterogeneous time-series feature data of the track at multiple scales. Feature dimension alignment technology is used to unify feature vectors of different types and scales to a standard 256-dimensional dimension. Geometric morphological features occupy the first 64 dimensions, filled sequentially with spatial partitions and measurement values ​​encoded at the millimeter scale, with each dimension corresponding to a specific spatial location's geometric feature index. Sleeper-track bed structural stress features occupy dimensions 65-128, filled with sleeper numbers and stress values ​​encoded at the micrometer scale, and associated with spatial unit coordinates. Rail vibration acceleration features occupy dimensions 129-192, filled with timestamps and vibration features encoded at the microsecond scale, and bound to sensor installation locations. Environmental and load features occupy dimensions 193-256, filled with acquisition locations and measurement features encoded at the second scale, distinguishing the dimensional ranges of environmental and load data. After dimensional alignment, a feature-weighted fusion method is adopted, with geometric morphology features, stress features, and vibration features each accounting for 25% of the weight, and environmental features and load features each accounting for 12.5%. By integrating the features of each dimension through weighted calculation, the redundancy of heterogeneous data is eliminated, and heterogeneous time-series state feature data of the track is generated. This ensures that the data retains multi-scale feature information and has a unified fusion format, providing standardized data support for subsequent spatial mapping processing.

[0036] Furthermore, step S2 includes the following steps: Step S21: Perform finite element mesh generation on the track multimodal state characteristic model to generate finite element mesh track multimodal state characteristic data; In this embodiment of the invention, based on the multimodal state characteristic model of the track, a structured mesh generation method is used to perform track finite element mesh generation on the model. The generation range covers the core structures such as rails, sleepers, track bed and contact interface, ensuring that the mesh is completely matched with the physical shape and material properties of the actual track structure. The rail section is divided according to the standard 60kg / m rail cross-section dimensions, with one grid unit set every 0.5 meters longitudinally and 8 layers of grids laterally based on the cross-sectional profile. The thickness of each layer of grid is evenly distributed. The unit type is quadrilateral solid unit, and the grid size is controlled at 0.05m × 0.05m. The sleepers are divided according to the concrete sleeper dimensions, with one grid unit set every 0.1 meters along the length direction, 4 layers in the width direction, and 6 layers in the height direction. The unit type is hexahedral solid unit, and the grid size is controlled at 0.1m × 0.1m × 0.1m. The track bed is divided according to a 30cm thickness, with one grid unit set every 1 meter longitudinally and 1.5 meters on each side of the track center laterally, divided into 15 layers of grids. The unit type is tetrahedral solid unit, and the grid size is controlled at 0.2m × 0.2m × 0.2m. The grid at the contact interface is densified, and the grid size is reduced to 1 / 2 of the surrounding grid to ensure accurate mapping of force transmission at the structural contact points. After the mesh is divided, the material parameters, geometric coordinates and state attributes of each structure are associated with the corresponding mesh element to generate finite element mesh track multimodal state characteristic data. Each mesh element corresponds to a unique identifier and complete structural attribute information.

[0037] Step S22: Based on the multimodal state characteristic data of the track using finite element mesh, perform stress and strain distribution characteristic analysis on the track structure to generate stress-strain distribution characteristic data of the track structure; In this embodiment of the invention, relying on the multimodal state characteristic data of the track using finite element mesh, and combining the physical characteristics of the track structure carried by the digital twin model with the actual operating conditions, the system conducts a distribution characteristic analysis of the stress and strain of the track structure. This accurately captures the variation patterns and distribution trends of stress and strain in different structural parts and under different operating conditions, generating comprehensive quantitative stress-strain distribution characteristic data of the track structure. In the early stages of analysis, deep binding between data and the model is completed, fully associating the finite element mesh elements with the material properties, geometric shape, and spatial location information of the corresponding track structure. This ensures that each mesh element carries complete basic characteristic data, and the contact interface is further densified with additional associated structural contact properties, providing accurate model support for subsequent mechanical analysis. Subsequently, boundary constraints are set based on the actual installation and fixing method of the track. For the different stress characteristics of the rails, sleepers, and track bed structures, corresponding constraint types are set to limit non-permissible displacements while preserving reasonable stress deformation space. Simultaneously, damping constraints are set at the track ends to weaken the interference of boundary reflections on the analysis results, ensuring that the constraint conditions highly match the actual operating conditions. During load application, based on previously collected train operation load data, the load is precisely mapped to the corresponding rail grid cells according to the actual wheel-rail contact position, achieving a correspondence between load and grid cells. Simultaneously, the load intensity is corrected based on environmental factors to match the impact of environmental changes on the track structure's mechanical response. The loading frequency is synchronized with the sensor data acquisition frequency to ensure the real-time performance and accuracy of load application. Stress-strain analysis is conducted using professional mechanical calculation methods. Numerical calculations are performed by constructing mechanical equilibrium relationships within the grid cells, focusing on core indicators such as tensile stress, compressive stress, shear stress, linear strain, and shear strain. The analysis focuses on the mechanical response of easily deteriorated areas such as the rail head, sleeper ends, and ballast surface, accurately capturing subtle structural deformations and stress concentration phenomena. During the analysis, stress and strain data of the entire grid cell are recorded at fixed time intervals, and corresponding working parameters and environmental conditions are synchronously associated to form analysis results with both spatial distribution and temporal variation. This clearly presents the distribution differences of stress and strain in different structural parts, the fluctuation law with working conditions, and the distribution characteristics of stress concentration areas. The final generated stress-strain distribution characteristic data of the track structure fully covers the mechanical response information, distribution law, and easy-to-deteriorate area markings of each grid cell in the spatiotemporal dimensions, providing solid data support for subsequent implicit strain trend analysis. At the same time, it meets the core requirement of digital twin models to accurately replicate the mechanical state of track structures.

[0038] Step S23: Analyze the hidden strain trend inside the track based on the stress-strain distribution characteristic data of the track structure, and generate hidden strain trend data inside the track. In this embodiment of the invention, relying on the stress-strain distribution characteristic data of the track structure, a combination of difference analysis and trend extrapolation is used to conduct implicit strain trend analysis within the track, focusing on capturing strain changes and cumulative trends in areas of the structure that are not directly measured. First, 12 consecutive hours of stress-strain data are selected and divided into 72 10-minute time intervals. The strain values ​​of each grid cell within each time interval are extracted, and the strain difference between adjacent time intervals for the same grid cell is calculated. The difference threshold is set to 0.0005ε, and grid cells with strain differences exceeding the threshold are selected and marked as potential implicit strain areas. Then, for these potential implicit strain areas, the load conditions and environmental data for the corresponding time intervals are traced, and the correlation between strain changes and load intensity and temperature fluctuations is analyzed. A linear regression algorithm is used to extrapolate the strain change trend for the next 24 hours, setting each hour as a prediction time interval, and calculating the cumulative strain increment for each time interval. Simultaneously, for the core area within the track structure where no sensors are deployed, interpolation calculations are performed using strain data from surrounding grid cells to supplement the strain values ​​of the internal area, ensuring that the implicit strain analysis covers the entire track structure. The final result is data on the latent strain trend inside the track, including the coordinates of the latent strain area, the strain values ​​at each time period, the cumulative increment, and the prediction results of future trends, which accurately reflects the emergence and development of latent degradation inside the track.

[0039] Step S24: Analyze the dynamic response characteristics of the track structure based on the stress-strain distribution characteristics data of the track structure and the implicit strain trend data inside the track, and generate dynamic response characteristic data of the track structure.

[0040] In this embodiment of the invention, the stress-strain distribution characteristic data of the track structure and the implicit strain trend data within the track are integrated to conduct dynamic response characteristic analysis of the track structure from three dimensions: dynamic response amplitude, response frequency, and correlation characteristics. First, dynamic change indicators are extracted from the stress-strain distribution data, including the stress fluctuation amplitude corresponding to rail vibration and the sleeper strain response period. The fluctuation amplitude is calculated based on the difference between the maximum and minimum values ​​at each time period, and the response period is statistically calculated based on the time interval of repeated stress-strain changes, accurately capturing the real-time dynamic response of the track structure to train loads. Then, combined with the implicit strain trend data, the correlation between implicit strain accumulation and dynamic response is analyzed. The correlation value between the implicit strain increment and the stress fluctuation amplitude is calculated, and regions with an absolute correlation value higher than 0.7 are marked as strongly correlated regions, clarifying the degree of influence of implicit strain on the track's dynamic response. Simultaneously, different operating conditions are divided according to train operating speed, including three levels: 60km / h, 120km / h, and 180km / h. The differences in the dynamic response of the track structure under different operating conditions are compared, and characteristic indicators for each operating condition are extracted, including the maximum stress response value and strain response lag time, with the lag time controlled within 50ms. After the analysis is completed, indicators such as dynamic response amplitude, frequency, correlation characteristics and operating condition differences are integrated to generate dynamic response characteristic data of the track structure. This provides comprehensive and accurate dynamic data support for subsequent analysis of deterioration behavior characteristics, enabling a deep characterization of the track structure's operating status.

[0041] Furthermore, step S22 includes the following steps: Step S221: Perform boundary condition analysis on the finite element mesh track based on the multimodal state characteristic data of the track, and generate boundary condition data for the track; In this embodiment of the invention, the multimodal state characteristic data of finite element mesh track is used as the core, combined with the actual stress conditions and installation and fixing methods of the track structure, to conduct boundary condition analysis of the mesh track. First, the mesh units are classified and organized according to their structural types. The rail mesh units are divided into sections according to the track, with each 200-meter section constrained. The mesh units at both ends of each section are set as elastic support constraints, allowing only ±0.001m displacement in the vertical direction, while completely restricting horizontal and longitudinal displacement. The elastic support stiffness is set to 200N / mm to match the actual support characteristics of the rail. The mesh units in the contact area between the sleeper and the track bed are set as friction constraints, with a fixed friction coefficient of 0.3, restricting the horizontal sliding of the sleeper relative to the track bed while allowing compressive displacement within 0.002m in the vertical direction, consistent with the contact stress logic between the sleeper and the track bed. The mesh units at the bottom and sides of the track bed are set as fixed constraints, completely restricting displacement in all directions, simulating the fixed connection between the track bed and the roadbed. Additional damping constraints are added to the mesh elements at both ends of the track, with a damping coefficient set to 50 N·s / m, to reduce the interference of boundary reflections on the analysis results. After the analysis is completed, the constraint type, constraint direction, displacement limit, and stiffness parameters are recorded according to the mesh element identifier to generate mesh track boundary condition data, ensuring that the boundary constraints of each mesh element are completely consistent with the actual working conditions.

[0042] Step S222: Perform dynamic load characteristic analysis on the finite element mesh track based on the multimodal state characteristic data of the track, and generate dynamic load characteristic data of the track. In this embodiment of the invention, based on the multimodal state characteristic data of the finite element mesh track and combined with the collected train operation load data, a dynamic load characteristic analysis of the mesh track is carried out. First, the axle load, wheel-rail force, and load action time series in the train operation load are extracted. The axle load is determined to be 25t based on the actual measured value. The wheel-rail force is decomposed into instantaneous impact load and steady-state load according to the time series. The peak value of the instantaneous impact load is set to 120kN, and the sustained value of the steady-state load is set to 80kN. Then, the wheel-rail contact position is precisely mapped to the rail mesh unit. Each wheelset corresponds to 4 consecutive rail mesh units, and the load is evenly distributed among these 4 mesh units, with the contact area controlled at 0.02m × 0.05m. The loading frequency is kept consistent with the sensor data acquisition frequency at 100Hz. The load action duration is calculated based on a train speed of 120km / h. The duration of the load on each mesh unit is 0.015s, and the load action interval is calculated to be 0.075s based on a train wheelbase of 2.5m. Simultaneously, by incorporating temperature parameters from environmental data, the load intensity is corrected. For every ±10℃ deviation of the temperature from the baseline value of 25℃, the load intensity is adjusted by ±2%, matching the impact of temperature on the mechanical properties of the track structure. After the analysis is completed, the grid cell identifier, load type, load value, duration of action, and adjustment coefficient are integrated to generate grid track dynamic load characteristic data, achieving precise binding between the load and the grid cells.

[0043] Step S223: Analyze the distribution characteristics of stress and strain of the track structure based on the grid track boundary condition data and grid track dynamic load characteristic data, and generate stress-strain distribution characteristic data of the track structure.

[0044] In this embodiment of the invention, grid track boundary condition data and grid track dynamic load characteristic data are integrated to analyze the distribution characteristics of stress and strain in the track structure based on the principle of structural mechanics equilibrium. Boundary constraint parameters and load parameters are associated with corresponding grid cells to construct the mechanical calculation equations for each grid cell. The calculation process employs an explicit integration algorithm with a time step of 10 μs to ensure the accuracy of the dynamic response calculation. Stress calculation focuses on three core indicators: tensile stress, compressive stress, and shear stress. The calculation accuracy for tensile and compressive stress is controlled within 0.001 MPa, and the calculation accuracy for shear stress is controlled within 0.0005 MPa. Emphasis is placed on statistically analyzing stress values ​​in easily deteriorated areas such as the rail head, sleeper ends, and the surface of the track bed. Strain calculation simultaneously performs linear strain and shear strain analysis, with the calculation accuracy controlled within 0.001ε and the calculation accuracy for shear strain within 0.0005ε, capturing subtle deformations of the track structure under load. Stress and strain values ​​for the entire grid cell are recorded every 10 ms according to a time series, and the corresponding load conditions and boundary conditions are labeled, forming an analysis result with both spatial distribution and temporal variation dimensions. Finally, stress-strain distribution characteristic data of the track structure were generated, clearly showing the stress-strain law of different grid elements under load, providing accurate data support for subsequent implicit strain analysis.

[0045] Furthermore, step S3 includes the following steps: Step S31: Perform track structure response residual analysis based on the track structure dynamic response characteristic data to generate track structure response residual data; In this embodiment of the invention, track structure response residual analysis is conducted based on the dynamic response characteristic data of the track structure and the theoretical dynamic response values ​​corresponding to the track multimodal state characteristic model. First, core indicators are extracted from the dynamic response characteristic data, including rail stress fluctuation amplitude, sleeper strain response period, maximum stress response value, and strain response lag time. Actual measured values ​​of each indicator are analyzed at 10ms intervals. Simultaneously, theoretical dynamic response values ​​under corresponding time nodes and load conditions are retrieved from the track multimodal state characteristic model. The theoretical values ​​are derived from mechanical formulas based on material parameters, geometric structure, and boundary conditions in the model, ensuring consistency with the actual measurement scenario. The residual calculation adopts the absolute value of the difference between the actual measured value and the theoretical value, and sets a uniform calculation accuracy of 0.0001MPa (stress index), 0.0001ε (strain index), and 1ms (time index). For the rail stress fluctuation amplitude, the residual threshold is tentatively set at 0.005MPa, the sleeper strain response period residual threshold is tentatively set at 2ms, the maximum stress response value residual threshold is tentatively set at 0.01MPa, and the strain response lag time residual threshold is tentatively set at 1ms. During the calculation process, the residual is calculated one by one for each time node and each characteristic index, synchronously associated with the corresponding grid cell coordinates, load conditions, and environmental parameters, forming a complete dataset containing residual values, index types, spatiotemporal information, and operating conditions. This generates track structure response residual data, accurately reflecting the deviation between the actual response and the theoretical response, laying the foundation for subsequent anomaly identification.

[0046] Step S32: Perform residual sequence analysis on the track structure response residual data to generate track structure response residual sequence data; In this embodiment of the invention, residual sequence analysis is performed on the track structure response residual data. A dual analysis system is constructed according to the time and spatial dimensions to ensure a comprehensive understanding of the residual variation patterns. In the time dimension, a continuous 24-hour analysis period is used, and the residual data is concatenated into a time series at 10ms intervals. Each characteristic index corresponds to an independent time series. Simultaneously, time periods are divided according to train operating conditions, including empty-load periods, fully loaded periods, low-speed operating periods (60km / h and below), medium-speed operating periods (60km / h-120km / h), and high-speed operating periods (120km / h-180km / h), respectively, to analyze the variation patterns of the residual sequences within different time periods. In the spatial dimension, the track line is divided into 20-meter spatial units, and the residual data is associated with the corresponding spatial units. Within each spatial unit, residual sequences are categorized and integrated according to characteristic indicators, with a focus on the residual distribution in easily deteriorated areas such as the rail head, sleeper ends, and ballast surface. During the analysis, the mean, variance, range, and fluctuation frequency of each time series were calculated. The mean was calculated on an hourly rolling basis, the variance was controlled with an accuracy of 0.000001, the range was the difference between the maximum and minimum values ​​within the series, and the fluctuation frequency was the ratio of the number of times the residual exceeded a provisional threshold to the total number of times. Simultaneously, abnormal residual values ​​caused by temporary sensor malfunctions were removed. The malfunction judgment criterion was that the abrupt change in residual amplitude at three adjacent time points exceeded 10 times the mean. After removal, the residuals were supplemented by interpolation using adjacent valid residual values ​​to ensure the continuity of the series. Finally, track structure response residual sequence data was generated, achieving an ordered spatiotemporal dimension of the residual data.

[0047] Step S33: Perform abnormal feature analysis of track structure response residuals based on track structure response residual sequence data, and generate abnormal feature data of track structure response residuals; In this embodiment of the invention, based on track structure response residual sequence data and combined with the standard for defining the normal fluctuation range of residuals, an abnormal characteristic analysis of track structure response residuals is conducted to clarify the boundaries and corresponding characteristics of normal fluctuations and abnormal fluctuations. First, the normal fluctuation range of the residuals of each characteristic index is determined through statistical analysis. Based on the mean and variance of the residuals during the unloaded period and under normal operating conditions in the residual sequence data, the normal fluctuation range is defined according to the principle of 3 times the standard deviation. Residuals exceeding this range are judged as abnormal fluctuations. Specifically, the normal range for rail stress fluctuation amplitude residuals is controlled within ±0.015MPa, the normal range for sleeper strain response period residuals is controlled within ±6ms, the normal range for maximum stress response value residuals is controlled within ±0.03MPa, and the normal range for strain response lag time residuals is controlled within ±3ms. During the analysis, residual values ​​are judged one by one according to the time series, and abnormal residuals exceeding the normal range are marked. The time nodes, spatial unit coordinates, load conditions, and environmental parameters corresponding to the abnormal residuals are recorded simultaneously, including train speed, axle load, ambient temperature, and humidity. For continuously occurring abnormal residuals (three or more consecutive time points exceeding the normal range), their fluctuation trends are further analyzed, including increasing trends, decreasing trends, and random fluctuation trends. The rate of change of the abnormal residuals is calculated, with the rate accuracy controlled within 0.0001 MPa / ms (stress index) and 0.1 ms / ms (time index). Simultaneously, the physical properties of the corresponding space unit's orbital structure are correlated to investigate whether the abnormal residuals are concentrated in specific material regions or structurally weak parts. This results in a complete analysis of the spatiotemporal distribution, trend changes, and associated attributes of the abnormal residuals, generating abnormal characteristic data of the orbital structure response residuals.

[0048] Step S34: Analyze the track deterioration behavior characteristics based on the abnormal characteristic data of track structure response residuals, and generate track deterioration behavior characteristic data.

[0049] In this embodiment of the invention, based on the abnormal residual characteristics data of the track structure response, combined with the track deterioration mechanism and typical deterioration forms, a track deterioration behavior characteristic analysis is conducted to accurately characterize the type, degree, and distribution pattern of deterioration. First, a correspondence is established between the abnormal residual characteristics and known track deterioration types. A continuously increasing residual in the rail stress fluctuation amplitude, concentrated in a specific spatial unit at the rail head, corresponds to rail wear deterioration; an abnormal residual in the sleeper strain response period, accompanied by a sudden change in the maximum stress response value residual, corresponds to sleeper cracking deterioration; abnormal residuals in the ballast area, positively correlated with the load intensity, correspond to ballast compaction deterioration; and multiple abnormal residuals, concentrated in the contact interface area, correspond to poor sleeper-ballast contact deterioration. During the analysis, the degree of deterioration is quantified according to the magnitude of the abnormal residuals exceeding the normal range: mild deterioration is when the residual exceeds the normal range by 1-2 times, moderate deterioration by 2-3 times, and severe deterioration by more than 3 times. Simultaneously, the number and distribution range of spatial units corresponding to each degree of deterioration are statistically analyzed, and the coordinates and extension trend of the core deterioration area are marked. Simultaneously, by tracing the load conditions and environmental data during the abnormal residual periods, the correlation between deterioration behavior and train axle load, operating speed, and temperature fluctuations was analyzed to identify the dominant deterioration factors. For example, the frequency of abnormal rail stress residuals increased under high-temperature conditions, indicating that temperature was the dominant factor in rail deterioration in this area. Furthermore, by combining the duration of residual abnormalities, the onset time and development rate of deterioration behavior were estimated, providing a basis for subsequent deterioration trend prediction. Finally, track deterioration behavior characteristic data were generated, including core information such as deterioration type, degree, spatial distribution, dominant factors, and development rate, achieving a precise quantitative characterization of deterioration behavior.

[0050] Furthermore, step S33 includes the following steps: Step S331: Perform random fluctuation analysis of track structure response residuals based on the track structure response residual sequence data to generate random fluctuation data of track structure response residuals; In this embodiment of the invention, based on the residual sequence data of track structure response, a stochastic fluctuation analysis of the track structure response residuals is conducted by combining both time and spatial dimensions. The residual sequences are categorized and organized according to characteristic indicators, including rail stress fluctuation amplitude residuals, sleeper strain response period residuals, maximum stress response value residuals, and strain response lag time residuals. Each indicator corresponds to an independent spatiotemporal sequence dataset, with a time interval of 10ms and a spatial range divided into 20-meter units. The analysis process employs a combination of stationarity testing and fluctuation intensity quantification. Stationarity testing is achieved by calculating the mean deviation of the residuals over 100 consecutive time points. A mean deviation within 0.0001MPa (stress indicator) and 0.1ms (time indicator) is considered a stationary fluctuation; deviations exceeding this are considered non-stationary fluctuations. Fluctuation intensity quantification uses both variance and fluctuation frequency indicators. Variance calculation is performed in a rolling statistical window of 500ms, with the variance accuracy of time-related residuals controlled within 0.01ms². Fluctuation frequency is calculated by counting the number of times the residuals deviate from the time-period mean within each statistical window, with a deviation threshold set at 10% of the time-period mean. Simultaneously, the load conditions and environmental parameters of the corresponding spatial units are correlated to distinguish the fluctuation differences under no-load, full-load, and different speed levels, and the coordinates, temperature, and humidity values ​​of the grid units corresponding to each fluctuation state are recorded synchronously. After the analysis is completed, the stability judgment results, fluctuation variance, fluctuation frequency, and associated load condition parameters are integrated to generate random fluctuation data of the track structure response residuals, accurately characterizing the spatiotemporal characteristics and load condition correlation of the residual random fluctuations.

[0051] Step S332: Extract abnormal fluctuations of track structure response residuals from the random fluctuation data of track structure response residuals to obtain abnormal fluctuation data of track structure response residuals; In this embodiment of the invention, based on the random fluctuation data of the track structure response residuals, and combined with preset fluctuation thresholds and anomaly judgment rules, abnormal fluctuations in the track structure response residuals are extracted. First, based on the stationary fluctuation data statistically obtained by S331, the normal fluctuation thresholds for each characteristic index are determined. The normal fluctuation variance threshold for the rail stress fluctuation amplitude residual is set to 0.000005 MPa², and the fluctuation frequency threshold is set to no more than 5 times per 500 ms; the normal fluctuation variance threshold for the sleeper strain response period residual is set to 0.05 ms², and the fluctuation frequency threshold is set to no more than 3 times per 500 ms; the normal fluctuation variance threshold for the maximum stress response value residual is set to 0.00001 MPa², and the fluctuation frequency threshold is set to no more than 4 times per 500 ms; the normal fluctuation variance threshold for the strain response lag time residual is set to 0.02 ms², and the fluctuation frequency threshold is set to no more than 2 times per 500 ms. During the extraction process, the fluctuation data for each statistical window are compared against thresholds. Fluctuations with a single indicator exceeding the normal threshold in variance or frequency are marked as potential abnormal fluctuations. Fluctuations with both indicators exceeding the threshold simultaneously are directly identified as abnormal fluctuations. For marked abnormal fluctuations, the complete residual sequence at the corresponding time point is traced, and the start time, duration, peak deviation, and corresponding spatial unit coordinates of the abnormal fluctuations are extracted. Fluctuations with a duration of less than three statistical windows (1.5s) and no continuous diffusion trend are identified as transient interference fluctuations and removed. Only abnormal fluctuations with a duration of ≥3 statistical windows or a continuous diffusion trend are retained. Finally, the spatiotemporal information, fluctuation parameters, and associated operating condition data of the effective abnormal fluctuations are integrated to obtain abnormal fluctuation data of the track structure response residuals, ensuring the accuracy and effectiveness of abnormal fluctuation extraction.

[0052] Step S333: Perform abnormal characteristic analysis of track structure response residuals based on the abnormal fluctuation data of track structure response residuals, and generate abnormal characteristic data of track structure response residuals.

[0053] In this embodiment of the invention, based on the abnormal fluctuation data of track structure response residuals, combined with the physical characteristics and deterioration initiation mechanism of the track structure, an abnormal characteristic analysis of track structure response residuals is carried out. First, the abnormal fluctuation data is classified and integrated according to spatial units and characteristic indicators. The grid unit, track structure part (rail head, sleeper ends, ballast surface, etc.), and characteristic indicator type corresponding to each abnormal fluctuation are labeled. Load data and environmental data for the corresponding time period are simultaneously correlated, including changes in train axle load, running speed, ambient temperature, and humidity. The abnormal characteristic analysis focuses on three core dimensions: fluctuation trend, distribution pattern, and correlation characteristics. The fluctuation trend is classified into increasing, decreasing, and oscillating types based on the peak value change of abnormal fluctuations. For increasing fluctuations, the daily growth rate of the peak value is calculated, with accuracy controlled at 0.0001 MPa / day (stress indicator) and 0.1 ms / day (time indicator). The distribution pattern analysis determines whether abnormal fluctuations are concentrated in specific spatial sections or structural parts. The frequency of abnormal fluctuations and the number of covered grid units within the same section are statistically analyzed. Sections with a frequency ≥ 5 times and covered units ≥ 3 are marked as key abnormal areas. Correlation characteristic analysis clarifies the correlation between abnormal fluctuations and load intensity and temperature fluctuations by comparing the operating conditions during periods of abnormal fluctuations and periods of normal fluctuations. For example, if the frequency of abnormal fluctuations increases by more than two times in the same spatial unit for every 5°C increase in temperature, it is determined that there is a strong correlation between temperature and abnormal fluctuations in that area. Simultaneously, the duration, peak deviation amplitude, and diffusion range of abnormal fluctuations are recorded to quantify their severity, forming a complete analysis result that includes the trend, spatial distribution, related factors, and severity of abnormal fluctuations. This generates abnormal characteristic data of track structure response residuals, providing precise targeted data for subsequent analysis of degradation behavior characteristics.

[0054] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1 A detailed flowchart illustrating the implementation steps of step S4 is provided in this embodiment. Step S4 includes: Step S41: Perform external driving factor analysis on the track based on the dynamic response characteristic data of the track structure, and generate external driving factor data of the track; In this embodiment of the invention, the dynamic response characteristic data of the track structure is used as the core, combined with the original environmental and load data, to conduct an analysis of external driving factors of the track, accurately identify key external factors affecting track deterioration, and quantify their characteristics. First, the types of external factors are identified, clearly divided into load-related factors and environmental factors. Load-related factors include train axle load, operating speed, peak wheel-rail force, and load application frequency. Environmental factors include ambient temperature, humidity, rainfall, and diurnal temperature range. For load-related factors, data for each parameter are statistically analyzed on a 24-hour cycle. Axle load is divided into three levels based on a 25t baseline: light load (≤20t), standard load (20t-28t), and heavy load (>28t). The duration and application frequency of each level are statistically analyzed. Operating speed is divided into three operating conditions: 60km / h, 120km / h, and 180km / h. The average and fluctuation amplitude of the peak wheel-rail force under different operating conditions are calculated. Load application frequency is statistically analyzed based on the number of trains passing per hour, with an accuracy controlled to 1 train / hour. For environmental factors, temperature was calculated based on hourly averages, divided into three ranges: low temperature (≤0℃), normal temperature (0℃-30℃), and high temperature (>30℃). Humidity was calculated based on daily averages, divided into three ranges: dry (≤40%RH), humid (40%RH-70%RH), and damp (>70%RH). Rainfall was calculated based on daily cumulative values, and diurnal temperature range was calculated as the difference between the highest and lowest temperatures of the day. During the analysis, the changing patterns of each external factor and the dynamic response indicators of the orbital structure were correlated, and the correlation values ​​between the factor changes and the fluctuation amplitude of the response indicators were calculated. Factors with absolute correlation values ​​higher than 0.6 were selected as key driving factors. Finally, the types, quantitative indicators, spatiotemporal distribution, and correlation characteristics of the key driving factors were integrated to generate orbital external driving factor data, providing a precise basis for subsequent causal correlation analysis.

[0055] Step S42: Perform causal correlation processing on external driving factor degradation based on orbital external driving factor data and orbital degradation behavior characteristic data to generate causal correlation data on external driving factor degradation. In this embodiment of the invention, data on external driving factors of the track and data on track deterioration behavior characteristics are integrated. A combination of causal correlation quantification and attribution analysis is used to process the causal correlation of external driving factor deterioration. Data is categorized and sorted according to deterioration type, and four types of deterioration behaviors—rail wear, sleeper cracking, track bed compaction, and poor contact—are analyzed and mapped one-to-one with external driving factors. For rail wear deterioration, the correlation coefficients between train axle load, operating speed, high-temperature duration, and wear degree are calculated. The correlation coefficient calculation uses the Pearson correlation algorithm with an accuracy controlled within 0.01. It is determined that when the axle load is >28t and the high-temperature duration is >8 hours / day, the correlation coefficient reaches 0.85, which is the main driving combination. For sleeper cracking deterioration, the influence of peak wheel-rail force and diurnal temperature difference is focused. The cracking frequency when the peak force is >120kN and the temperature difference is >15℃ is statistically analyzed, accounting for 72%, which is identified as the core driving condition. To address ballast bed compaction and deterioration, the synergistic effect of rainfall and load frequency was analyzed. When daily rainfall > 5 mm and daily train traffic > 60 trains, the compaction and deterioration rate increased threefold. For poor contact deterioration, humidity and wheel-rail vibration amplitude were correlated. When humidity > 70% RH and vibration amplitude > 5g, the incidence of poor contact significantly increased. During the analysis, the influence weights of each driving factor on different deterioration types were quantified. Load-related factors accounted for 65% and 58% of rail wear and sleeper cracking, respectively, while environmental factors accounted for 70% and 62% of ballast bed compaction and poor contact, respectively. Finally, causal correlation data of external driving factors was generated, including the dominant driving factors, correlation coefficients, influence weights, and synergistic effects for each deterioration type, clearly presenting the causal relationship between external factors and deterioration behavior.

[0056] Step S43: Perform track degradation state specificity analysis based on track degradation behavior characteristic data to generate track degradation state specificity data; In this embodiment of the invention, based on track deterioration behavior characteristic data and combined with the physical properties of track structure and the mechanism of deterioration initiation, a specific analysis of track deterioration state is carried out to accurately capture the unique characteristics and evolution patterns of different deterioration types. First, the data is split according to deterioration type, and a specific analysis index system is constructed for each type. For rail wear deterioration, the focus is on the wear rate, the distribution of wear locations, and the correspondence with the wheel-rail contact position. The wear rate is statistically analyzed with a daily accuracy of 0.001 mm, distinguishing the wear differences between different parts of the rail head and side. For sleeper cracking deterioration, the focus is on analyzing the crack initiation location, expansion rate, and crack width. Crack locations are concentrated at the two ends of the sleeper (1 / 4 of the way down), the expansion rate is statistically analyzed with a daily accuracy of 0.01 mm, and the width is classified into levels of <0.1 mm, 0.1 mm-0.3 mm, and >0.3 mm. The analysis focuses on the density changes, permeability, and overlap with the load-bearing area in track bed compaction deterioration. Density changes are statistically analyzed based on a baseline value of 1800 kg / m³, and permeability is quantified by the depth of water penetration per hour. The analysis also focuses on the uneven distribution of stress at the contact interface, changes in the friction coefficient, and abnormal vibration response characteristics. The friction coefficient fluctuation range is statistically analyzed based on a baseline value of 0.3. During the analysis, the evolutionary differences of different deterioration types were compared. Rail wear showed a linear increasing trend, sleeper cracking exhibited a phased expansion characteristic, track bed compaction gradually worsened with accumulated rainfall, and poor contact fluctuated periodically due to humidity influences. Specific thresholds for each deterioration type were also identified; for example, a rail head wear rate > 0.003 mm / day and a sleeper crack width > 0.3 mm were considered accelerated deterioration states. Finally, specific data on track deterioration states were generated, including specific indicators, evolution patterns, state thresholds, and key characteristic parameters for each deterioration type, providing targeted data for subsequent evolutionary spatial analysis and prediction model construction.

[0057] Step S44: Based on the orbital degradation state-specific data and the causal correlation data of external driving factor degradation, perform degradation impact characteristic analysis of external driving factor coupling, and generate factor coupling degradation impact characteristic data; In this embodiment of the invention, based on track degradation state-specific data and causal correlation data of external driving factors, the degradation impact characteristics of external driving factor coupling are analyzed to quantify the influence of multi-factor synergy on degradation state. First, core coupling factor combinations are screened, and four typical coupling combinations are identified based on causal correlation data: axle load-high temperature coupling (heavy load + high temperature), wheel-rail force-day / night temperature difference coupling (high force + large temperature difference), rainfall-load frequency coupling (rainy + high-frequency load), and humidity-vibration amplitude coupling (high humidity + strong vibration). For each coupling combination, a single-factor control group and a coupled-factor experimental group are set up to compare and analyze the differences in degradation state. In the axle load-high temperature coupling group, the rail wear rate is increased by 40% and 55% compared to the single heavy load group and single high temperature group, respectively, clarifying the superimposed effect of coupling. The intensity of coupling effects was quantified using coupling effect coefficients. These coefficients were calculated as the ratio of the degradation rate of the coupled group to the average rate of the individual factor group. The coupling coefficient for axle load-high temperature was 1.95, for wheel-rail force-diurnal temperature difference it was 1.82, for rainfall-load frequency it was 1.78, and for humidity-vibration amplitude it was 1.65. During the analysis, degradation-specific indicators were simultaneously correlated to clarify the differences in the impact of coupling factors on different specific indicators. For example, the wheel-rail force-diurnal temperature difference coupling had an impact coefficient of 2.1 on the sleeper crack propagation rate and 1.5 on the crack width. Simultaneously, the effect thresholds for each coupling combination were determined; for example, the coupling effect was significantly enhanced when the axle load was >28t and the temperature was >35℃. Finally, characteristic data on the degradation effects of factor coupling were generated, including coupling factor combinations, coupling effect coefficients, superposition effects, differences in the effects of specific indicators, and effect thresholds, accurately characterizing the mechanism of multi-factor coupling on degradation.

[0058] Step S45: Analyze the spatial characteristics of the orbital degradation state evolution based on the orbital degradation state specific data to generate spatial characteristic data of the orbital degradation state evolution; In this embodiment of the invention, based on specific data of track deterioration status and combined with the geometric morphology of the track's three-dimensional spatial coordinate system and digital twin model, spatial characteristic analysis of track deterioration status evolution is conducted to accurately present the distribution, diffusion patterns, and correlation characteristics of deterioration in spatial dimensions. First, a track spatial evolution analysis grid is constructed based on 20-meter spatial units. Deterioration status data is associated with corresponding grid units, and the deterioration type, degree, initiation time, and current evolution stage of each unit are labeled. The spatial distribution pattern of deterioration is analyzed, and the proportion of deterioration types in each spatial segment is statistically analyzed. Rail wear is concentrated in curved sections (radius < 800 meters), accounting for 68%; sleeper cracking is concentrated in the transition sections at both ends of bridges, accounting for 75%; and track bed compaction is concentrated in the station entry and exit sections, accounting for 62%. The analysis of deterioration diffusion characteristics, using daily time units, statistically analyzed the expansion area and diffusion direction of deteriorated regions. The longitudinal diffusion rate of rail wear along the track was 0.5-1 meter per day; sleeper cracking spread laterally from a single sleeper to adjacent sleepers at a rate of 1-2 sleepers per day; and track bed compaction spread in sheet-like patterns, with the diffusion area increasing by 0.5-1 square meters per day. Simultaneously, the structural attributes of spatial units were correlated to analyze the relationship between deterioration and track geometry parameters and material distribution. In curved sections, the outer rail wear rate was twice as fast as the inner rail, and the cracking rate of C50 concrete sleepers was 30% lower than that of C40 sleepers. Furthermore, combining the spatial mapping relationship of the digital twin model, the locational correlation between deteriorated areas and key track structures (rail joints, sleeper support points) was marked, identifying a high-incidence area of ​​deterioration within 0.5 meters of the critical structures. Finally, spatial characteristic data of track deterioration evolution were generated, including the spatial distribution ratio of deterioration, diffusion rate, diffusion direction, structural correlation characteristics, and key area markings, providing spatial dimension support for trend prediction models.

[0059] Step S46: Establish the relationship between the external driving factors and the orbital degradation trend prediction by using the factor coupling degradation influence feature data and the orbital degradation state evolution spatial feature data, and generate an orbital degradation trend prediction model. In this embodiment of the invention, the characteristic data of the coupling effect of degradation factors and the spatial characteristic data of track degradation state evolution are integrated to construct the track degradation trend prediction relationship of the coupling effect of external driving factors, generating an accurate and controllable track degradation trend prediction model. Data preprocessing and dimensional alignment are performed, and the two types of data are double-matched according to degradation type and spatiotemporal dimension. The four types of degradation data—rail wear, sleeper cracking, track bed compaction, and poor contact—are respectively associated with specific coupling factor combinations and spatial evolution characteristics. The time dimension is uniformly aligned to a daily unit, and the spatial dimension is strictly matched to the coordinates of 20-meter grid units, ensuring that the coupling factor data and spatial evolution data within the same grid unit and the same time period accurately correspond. Simultaneously, redundant data items and outliers are removed, and spatial interpolation is used to supplement missing data, ensuring the integrity and consistency of the input data and building a solid data foundation for the model. The model adopts a four-layer architecture: "data input layer - feature fusion layer - classification prediction layer - calibration layer". Each layer is functionally independent yet works in concert. The data input layer receives two types of preprocessed data and transmits them to the feature fusion layer according to their degradation type. Each layer of data carries a unique spatiotemporal identifier and a degradation type label. The feature fusion layer uses a weighted fusion algorithm to deeply fuse the coupling coefficient, action threshold, and superposition effect in the factor coupling influence features with the diffusion rate, distribution pattern, and structural correlation characteristics in the spatial evolution features. During the fusion process, feature weights are allocated according to the degradation type: 60% for load-related coupling factor features, 20% for environmental coupling factor features, and 20% for spatial evolution features. This ensures that the fused features highlight both the coupling-driven core and the spatial evolution pattern. The classification prediction layer establishes dedicated prediction logic for each of the four types of deterioration, creating differentiated prediction equations and defining parameter configuration rules. The rail wear prediction equation focuses on incorporating the axle load-high temperature coupling coefficient, initial wear rate, and longitudinal diffusion rate, with parameter weights strictly set at 60% for the coupling coefficient, 25% for the initial rate, and 15% for the diffusion rate. Simultaneously, it correlates key spatial attributes such as curve sections and rail joints to correct the prediction results. The sleeper cracking prediction equation incorporates the wheel-rail force-day-night temperature difference coupling coefficient, initial crack width, and lateral diffusion rate, with parameter weights allocated at 55% for the coupling coefficient, 30% for the initial width, and 15% for the diffusion rate. It also simultaneously overlays sleeper material properties and installation location features to optimize accuracy. The track bed compaction and poor contact prediction equations are constructed similarly, incorporating corresponding coupling factor combinations and spatial evolution parameters. The track bed compaction equation additionally correlates rainfall lag effect parameters, while the poor contact equation supplements humidity fluctuation correlation parameters, ensuring that the prediction logic for each type of deterioration aligns with its unique evolution mechanism.The model parameter calibration process employs a two-way verification method using historical degradation data. Measured degradation data from the past 30 days across the entire line under different operating conditions and in different spatial units are selected and substituted into the prediction equation to generate prediction results. Errors are calculated by comparing the predicted values ​​with the measured values ​​point by point. The prediction error for rail wear is strictly controlled within ±0.001 mm / day, the prediction error for sleeper crack propagation is controlled within ±0.01 mm / day, and the prediction error for track bed compaction diffusion area is controlled within ±0.1 square meters / day. If the error for a single type of degradation exceeds the threshold, the parameter weights of the corresponding prediction equation are adjusted in reverse until the error meets the standard. To ensure the model adapts to real-time changes in operating conditions, a dynamic update mechanism is specifically designed. The update frequency is consistent with the sensor data acquisition frequency at 100 Hz. Every 100 milliseconds, the latest factor coupling data, spatial evolution data, and dynamic response data are automatically incorporated to correct the prediction equation parameters and coupling influence relationships in real time, and the prediction baseline values ​​of each grid unit are updated synchronously. Furthermore, the model incorporates a digital twin collaborative linkage function, which can feed the prediction results back to the orbital multimodal state characteristic model in real time, enabling dynamic comparison between predicted data and actual operational data, and further optimizing prediction accuracy. The final generated orbital degradation trend prediction model can accurately capture the short-term and medium-term evolution trends of different spatial units and different degradation types under the coupling effect of multiple factors, and has the core characteristics of clear type differentiation, accurate spatial positioning, traceable parameters, and dynamic optimization.

[0060] Step S47: Transmit the dynamic response characteristic data of the track structure to the track deterioration trend prediction model for track facility deterioration trend prediction processing, and generate track facility deterioration trend prediction data.

[0061] In this embodiment of the invention, a track deterioration trend prediction model is used as the core. Dynamic response characteristic data of the track structure is input, and track facility deterioration trend prediction processing is performed to generate accurate and quantified track facility deterioration trend prediction data. The prediction processing is carried out in three time periods: 7 days, 30 days, and 90 days. Prediction results are output daily for each period, covering all 20-meter spatial units along the entire line. Predictions are made separately for each type of deterioration. During the prediction process, real-time indicators are first extracted from the dynamic response characteristic data of the track structure, including rail stress fluctuation amplitude, peak wheel-rail force, ambient temperature, and humidity. These are then synchronously correlated with the historical deterioration state and coupling factor effects of the corresponding spatial units and substituted into the corresponding equations of the prediction model. Regarding rail wear, the total wear and distribution range for the next 90 days are predicted. It is determined that under heavy load and high temperature conditions, the total wear of the outer rail in curved sections will reach 0.27 mm in 90 days, exceeding the slight deterioration threshold of 0.07 mm. Regarding sleeper cracking, the number and width of cracks are predicted for the next 30 days. In the transition sections at both ends of the bridge, 12 sleepers will have cracks exceeding 0.3 mm in width, entering a state of severe deterioration. Regarding track bed compaction and poor contact, the diffusion area and degree of deterioration for the next 7 and 30 days are predicted respectively. It is determined that track bed compaction in station entry and exit sections will increase by 3.5 square meters after 7 days, and the incidence of poor contact in high-humidity sections will increase by 20% after 30 days. The confidence levels of each predicted data are simultaneously marked in the prediction results, determined based on historical verification errors. The confidence levels for the 7-day cycle are 95%, for the 30-day cycle 90%, and for the 90-day cycle 85%. The final result is a prediction of the deterioration trend of track facilities, which includes the deterioration type, prediction degree, distribution range and confidence level of each time period and each spatial unit, providing accurate and forward-looking support for track operation and maintenance decisions.

[0062] Furthermore, step S44 includes the following steps: Step S441: Based on the orbital degradation state specific data and the external driving factor degradation causal correlation data, perform adaptive correlation feature analysis on external driving factors and degradation state to generate external driving factor-degradation state adaptive correlation feature data. In this embodiment of the invention, specific data on track deterioration status and causal correlation data of external driving factors are integrated to conduct adaptive correlation feature analysis of external driving factors and deterioration status. By dynamically adjusting the analysis dimensions and thresholds, the correlation patterns between factors and deterioration status under different scenarios are accurately captured. First, the dataset is split according to deterioration type. For four types of deterioration—rail wear, sleeper cracking, ballast bed compaction, and poor contact—corresponding dominant driving factor data is established, creating a two-way correlation system of "factor-deterioration index." The adaptive analysis adopts a hierarchical iterative mode. The first layer calculates basic correlation coefficients according to load-related and environmental factors. The accuracy of the basic correlation coefficient calculation between load-related factors and rail wear and sleeper cracking is controlled at 0.01, and the accuracy of the basic correlation coefficient calculation between environmental factors and ballast bed compaction and poor contact is also controlled at 0.01. The second layer, based on real-time operating conditions fed back by the digital twin model, dynamically adjusts the correlation analysis thresholds. When the train speed is ≥180km / h, the correlation threshold between axle load and rail wear is lowered from 0.6 to 0.55; when the ambient temperature is ≥35℃, the correlation threshold between high temperature and rail wear is lowered from 0.65 to 0.6. During the analysis, low-correlation factors (correlation coefficient <0.5) are automatically removed, while core correlation factors are retained. Furthermore, for cases of abrupt changes in correlation patterns over three consecutive time periods, adaptive calibration is triggered, recalculating the correlation dimension weights to ensure the analysis results closely reflect real-time operating condition changes. Finally, adaptive correlation feature data of external driving factors and deterioration states is generated, including core correlation factors, dynamic correlation coefficients, adaptive thresholds, and weight adjustment records for each deterioration type, laying a precise foundation for subsequent influence feature analysis.

[0063] Step S442: Perform degradation impact feature analysis on the external driving factor-degradation state adaptive correlation feature data to generate degradation impact feature data of external driving factor; In this embodiment of the invention, relying on the adaptive correlation feature data of external driving factors and deterioration state, the core related factors are focused on to conduct a deterioration influence feature analysis of external driving factors, quantifying the degree, scope, and variation law of the influence of a single factor on the deterioration state. First, the core driving factors for each type of deterioration are extracted: rail wear corresponds to axle load and high temperature; sleeper cracking corresponds to peak wheel-rail force and diurnal temperature difference; track bed compaction corresponds to rainfall and load frequency; and poor contact corresponds to humidity and wheel-rail vibration amplitude. For each core factor, a gradient influence experimental group is set up: axle load is divided into gradients of 20t, 25t, 28t, and 30t; temperature is divided into gradients of 25℃, 30℃, 35℃, and 40℃; and wheel-rail force is divided into gradients of 100kN, 120kN, and 140kN. Each gradient corresponds to an independent analysis sample. During the analysis, other factors were kept at a baseline state, and only the gradient of the target factor was changed to quantify the impact of the target factor on the deterioration-specific indicators. When the axle load increased from 25t to 30t, the rail wear rate increased from 0.002mm / day to 0.0035mm / day, an increase of 75%; when the temperature rose from 25℃ to 40℃, the rail wear rate increased by 60%. Simultaneously, combining the three-dimensional spatial coordinates of the track, the spatial range of the factor's influence was analyzed. The impact of high temperature on rail wear was concentrated in the curve section (radius < 800 meters), covering the entire length of the corresponding section; the impact of peak wheel-rail force on sleeper cracking was concentrated in the two-quarter region at both ends of the sleeper, affecting a single sleeper and two adjacent sleepers. Furthermore, the time lag of the statistical factor's influence was analyzed: the lag time for the impact of rainfall on track bed compaction was 24 hours, while the impact of high temperature on rail wear had no significant lag. The final data on the deterioration impact of external driving factors includes the magnitude of the gradient impact of a single factor, the spatial range of its effect, the temporal lag, and the differences in the impact of specific indicators, clearly presenting the deterioration driving mechanism of a single factor.

[0064] Step S443: Perform degradation impact characteristic analysis of factor coupling based on the degradation impact characteristic data of external driving factors, and generate degradation impact characteristic data of factor coupling.

[0065] In this embodiment of the invention, based on the degradation influence characteristic data of external driving factors, combined with the degradation initiation mechanism and the synergistic effect of actual working conditions, a degradation influence characteristic analysis of factor coupling is carried out to quantify the superposition effect and coupling mechanism of multi-factor synergy on the degradation state. First, four typical coupling factor combinations are determined: axle load-high temperature (rail wear), wheel-rail force-diurnal temperature difference (sleeper cracking), rainfall-load frequency (ballast bed compaction), and humidity-vibration amplitude (poor contact). For each combination, a coupling experimental group and a single-factor control group are set up. The control group uses the baseline gradient of each factor, while the experimental group uses the combination of high-influence gradients of factors. The coupling influence strength is characterized by the coupling influence coefficient. The coefficient is calculated as the ratio of the degradation rate of the coupling group to the average rate of the single-factor group. For the axle load of 30t + temperature of 40℃ coupling group, the rail wear rate is 0.005mm / day; for the single axle load (30t), it is 0.0035mm / day; for the single high temperature group (40℃), it is 0.0032mm / day; and the coupling influence coefficient is 1.52. During the analysis, the threshold of the coupling effect was clearly defined. When the axle load > 28t and the temperature > 35℃, the coupling effect was significantly enhanced, with the coupling coefficient increasing from 1.3 to 1.52. When the wheel-rail force > 120kN and the diurnal temperature difference > 15℃, the sleeper crack propagation rate increased by 82% compared to the single factor group. Simultaneously, specific indicators of the deterioration state were correlated, and the differences in the influence of coupling factors on different indicators were analyzed. The influence coefficient of axle load-high temperature coupling on the rail wear rate was 1.52, and the influence coefficient on the wear distribution was 1.2, indicating that the coupling effect has a more significant impact on the wear rate. Furthermore, the dynamic simulation function of the digital twin model was used to verify the coupling influence law. The deterioration process under coupled conditions was reproduced through the model, ensuring that the analysis results are consistent with actual working conditions. Finally, characteristic data of the factor coupling deterioration influence were generated, including the combination of coupling factors, coupling influence coefficient, threshold, superposition effect, and differences in the influence of specific indicators, accurately depicting the core mechanism of multi-factor synergistic driving deterioration.

[0066] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0067] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for analyzing the deterioration trend of rail transit infrastructure based on digital twins, characterized in that, Includes the following steps: Step S1: Collect multi-source heterogeneous data of the track using multi-source sensors deployed on the track infrastructure to generate multi-source heterogeneous track data; establish a digital twin model of the track multi-modal state characteristics based on the multi-source heterogeneous track data to generate a multi-modal track state characteristic model. Step S2: Perform dynamic response feature analysis of the track structure based on the track multimodal state feature model to generate dynamic response feature data of the track structure; Step S3: Perform track degradation behavior characteristic analysis on the dynamic response characteristic data of the track structure to generate track degradation behavior characteristic data; Step S4: Establish a track deterioration trend prediction model based on track deterioration behavior characteristic data; transmit the dynamic response characteristic data of the track structure to the track deterioration trend prediction model for track facility deterioration trend prediction processing, and generate track facility deterioration trend prediction data.

2. The method for analyzing the deterioration trend of rail transit infrastructure based on digital twins according to claim 1, characterized in that, The multi-source heterogeneous track data mentioned in step S1 includes track geometry data, sleeper-track bed structural stress data, rail vibration acceleration data, track environment data, and train operation load data.

3. The method for analyzing the deterioration trend of rail transit infrastructure based on digital twins according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Collect multi-source heterogeneous data of the track by using multi-source sensors deployed on the track infrastructure to generate multi-source heterogeneous data of the track. Step S12: Perform multi-scale fusion state feature analysis on the multi-source heterogeneous data of the orbit to generate heterogeneous time-series state feature data of the orbit; Step S13: Perform spatial mapping processing on the heterogeneous time-series state feature data of the orbit to generate mapped heterogeneous orbit state feature data; Step S14: Establish a digital twin model of the orbital multimodal state features by mapping the heterogeneous state feature data of the orbit, and generate the orbital multimodal state feature model.

4. The method for analyzing the deterioration trend of rail transit infrastructure based on digital twins according to claim 3, characterized in that, Step S12 includes the following steps: Step S121: Design a multi-scale heterogeneous coding strategy and a multi-scale feature analysis strategy for orbits based on multi-source heterogeneous orbital data; Step S122: Use the orbital multi-scale heterogeneous coding strategy to perform multi-scale coding preprocessing on the orbital heterogeneous data to generate orbital multi-scale heterogeneous data; Step S123: Perform multi-scale heterogeneous time series data analysis on the multi-scale heterogeneous orbital data to generate multi-scale heterogeneous orbital time series data; Step S124: Use the orbital multi-scale feature analysis strategy to perform multi-scale heterogeneous time series feature analysis on the orbital multi-scale time series data to generate orbital multi-scale heterogeneous time series feature data; Step S125: Perform multi-scale fusion state feature analysis on the multi-scale heterogeneous time series feature data of the orbit to generate heterogeneous time series state feature data of the orbit.

5. The method for analyzing the deterioration trend of rail transit infrastructure based on digital twins according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Perform finite element mesh generation on the track multimodal state characteristic model to generate finite element mesh track multimodal state characteristic data; Step S22: Based on the multimodal state characteristic data of the track using finite element mesh, perform stress and strain distribution characteristic analysis on the track structure to generate stress-strain distribution characteristic data of the track structure; Step S23: Analyze the hidden strain trend inside the track based on the stress-strain distribution characteristic data of the track structure, and generate hidden strain trend data inside the track. Step S24: Analyze the dynamic response characteristics of the track structure based on the stress-strain distribution characteristics data of the track structure and the implicit strain trend data inside the track, and generate dynamic response characteristic data of the track structure.

6. The method for analyzing the deterioration trend of rail transit infrastructure based on digital twins according to claim 5, characterized in that, Step S22 includes the following steps: Step S221: Perform boundary condition analysis on the finite element mesh track based on the multimodal state characteristic data of the track, and generate boundary condition data for the track; Step S222: Perform dynamic load characteristic analysis on the finite element mesh track based on the multimodal state characteristic data of the track, and generate dynamic load characteristic data of the track. Step S223: Analyze the distribution characteristics of stress and strain of the track structure based on the grid track boundary condition data and grid track dynamic load characteristic data, and generate stress-strain distribution characteristic data of the track structure.

7. The method for analyzing the deterioration trend of rail transit infrastructure based on digital twins according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Perform track structure response residual analysis based on the track structure dynamic response characteristic data to generate track structure response residual data; Step S32: Perform residual sequence analysis on the track structure response residual data to generate track structure response residual sequence data; Step S33: Perform abnormal feature analysis of track structure response residuals based on track structure response residual sequence data, and generate abnormal feature data of track structure response residuals; Step S34: Analyze the track deterioration behavior characteristics based on the abnormal characteristic data of track structure response residuals, and generate track deterioration behavior characteristic data.

8. The method for analyzing the deterioration trend of rail transit infrastructure based on digital twins according to claim 7, characterized in that, Step S33 includes the following steps: Step S331: Perform random fluctuation analysis of track structure response residuals based on the track structure response residual sequence data to generate random fluctuation data of track structure response residuals; Step S332: Extract abnormal fluctuations of track structure response residuals from the random fluctuation data of track structure response residuals to obtain abnormal fluctuation data of track structure response residuals; Step S333: Perform abnormal characteristic analysis of track structure response residuals based on the abnormal fluctuation data of track structure response residuals, and generate abnormal characteristic data of track structure response residuals.

9. The method for analyzing the deterioration trend of rail transit infrastructure based on digital twins according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Perform external driving factor analysis on the track based on the dynamic response characteristic data of the track structure, and generate external driving factor data of the track; Step S42: Perform causal correlation processing on external driving factor degradation based on orbital external driving factor data and orbital degradation behavior characteristic data to generate causal correlation data on external driving factor degradation. Step S43: Perform track degradation state specificity analysis based on track degradation behavior characteristic data to generate track degradation state specificity data; Step S44: Based on the orbital degradation state-specific data and the causal correlation data of external driving factor degradation, perform degradation impact characteristic analysis of external driving factor coupling, and generate factor coupling degradation impact characteristic data; Step S45: Analyze the spatial characteristics of the orbital degradation state evolution based on the orbital degradation state specific data to generate spatial characteristic data of the orbital degradation state evolution; Step S46: Establish the relationship between the external driving factors and the orbital degradation trend prediction by using the factor coupling degradation influence feature data and the orbital degradation state evolution spatial feature data, and generate an orbital degradation trend prediction model. Step S47: Transmit the dynamic response characteristic data of the track structure to the track deterioration trend prediction model for track facility deterioration trend prediction processing, and generate track facility deterioration trend prediction data.

10. The method for analyzing the deterioration trend of rail transit infrastructure based on digital twins according to claim 9, characterized in that, Step S44 includes the following steps: Step S441: Based on the orbital degradation state specific data and the external driving factor degradation causal correlation data, perform adaptive correlation feature analysis on external driving factors and degradation state to generate external driving factor-degradation state adaptive correlation feature data. Step S442: Perform degradation impact feature analysis on the external driving factor-degradation state adaptive correlation feature data to generate degradation impact feature data of external driving factor; Step S443: Perform degradation impact characteristic analysis of factor coupling based on the degradation impact characteristic data of external driving factors, and generate degradation impact characteristic data of factor coupling.

Images