Rail infrastructure safety assurance method, system, device and storage medium

Abstract

The present invention discloses a rail infrastructure safety assurance method, system, device, and storage medium, relating to the field of intelligent safety monitoring and control technology for rail transit. The method comprises: collecting raw track data through a multi-source perception module; processing the raw track data using a bilateral filter and region growing method to obtain track data containing semantic tags; extracting features from the track data through a data fusion hub to construct a spatiotemporal correlation feature matrix; performing thermal-mechanical coupling simulation based on the track data through a digital twin decision platform to obtain simulation results; and generating optimized control instructions and alarm information through an adaptive control module based on the spatiotemporal correlation feature matrix and the simulation results. The present invention solves the problems of perception system fragmentation, insufficient intelligent diagnosis and decision-making, and passive risk management in related technologies.

Description

Technical Field

[0001] The present invention relates to the field of intelligent safety monitoring and control technology for rail transit, and in particular to a rail infrastructure safety assurance method, system, equipment and storage medium. Background Art

[0002] With the gradual shortening of train tracking intervals and the development of unmanned rail transit technology, the need for environmental awareness for train operations is becoming increasingly prominent. The construction of ultra-large-scale railway networks, in particular, presents complex geographical, geological, topographical, and meteorological conditions. These complex factors create diverse, multi-dimensional, wide-ranging, sudden, and difficult-to-identify safety risks for train operations. Common safety hazards include infrastructure failures, train breakdowns, foreign object intrusion, natural environmental changes, and disasters, all of which can directly threaten train safety.

[0003] Current rail infrastructure safety assurance technologies face the following key bottlenecks:

[0004] 1. Fragmentation of the perception system:

[0005] Existing monitoring systems mostly use a single sensor network (such as deploying only strain gauges or cameras), resulting in:

[0006] Lack of data dimension: Traditional fiber optic sensors can only collect axial strain and cannot simultaneously obtain track temperature gradients (affecting the calculation accuracy of rail expansion joints by 5-8%).

[0007] Difficulty in time and space alignment: There is a millisecond-level delay between the vibration signal (10kHz sampling) and the video data (30fps), resulting in a false alarm rate of up to 22% for turnout jams.

[0008] Poor environmental adaptability: The recognition accuracy of conventional cameras drops to 61% in rainy and foggy weather, which cannot meet all-weather monitoring needs.

[0009] (2) Insufficient intelligence in diagnostic decision-making:

[0010] Mainstream solutions rely on threshold alarms and manual judgment, which have the following problems:

[0011] Serious false alarms and missed alarms: The fixed threshold alarm system has a false trigger rate of over 35% when the wheel-rail adhesion coefficient changes dynamically.

[0012] Significant response delay: The average time from anomaly detection to dispatch instruction issuance is 8.7 minutes, which fails to meet the 5-minute train control response requirement for heavy-load trains.

[0013] Lack of root cause location: The existing system can only report "track geometry exceeds the limit", but cannot distinguish whether it is caused by loose fasteners (accounting for 62%) or roadbed settlement (accounting for 29%).

[0014] (3) Dilemma of passive risk management:

[0015] The traditional method adopts a passive mode of "fault occurrence - manual inspection - planned maintenance". Insufficient preventive maintenance has resulted in sudden failures accounting for 78%, and the loss of a single unplanned shutdown exceeds 2 million yuan per hour, resulting in high maintenance costs.

[0016] Therefore, there is an urgent need for a technical solution that can not only improve the holographic, intelligent and autonomous level of line perception, but also effectively enhance the protection capabilities for train operation safety, ultimately achieve more accurate and efficient fault prediction and risk response, and ensure the safe and stable operation of trains. Summary of the Invention

[0017] In order to solve the above technical problems, the present invention provides a rail infrastructure safety assurance method, system, device and storage medium.

[0018] To achieve the above object, the present invention provides the following technical solutions:

[0019] In a first aspect, an embodiment of the present invention provides a rail infrastructure safety assurance method, the method comprising:

[0020] Collect raw track data through multi-source perception modules;

[0021] Processing the original track data using a bilateral filter and a region growing method to obtain track data containing semantic labels;

[0022] Extracting features from the orbital data through a data fusion hub to construct a spatiotemporal correlation feature matrix;

[0023] Based on the track data, a thermal-mechanical coupling simulation is performed through a digital twin decision platform to obtain simulation results;

[0024] Based on the spatiotemporal correlation characteristic matrix and the simulation results, an optimized control instruction and an alarm message are generated through an adaptive control module.

[0025] In a second aspect, an embodiment of the present invention provides a rail infrastructure safety assurance system, comprising: a multi-source perception module, a data fusion hub, a digital twin platform, and an adaptive control module.

[0026] A multi-source perception module configured to collect raw track data;

[0027] A data fusion hub is configured to process the raw track data using a bilateral filter and a region growing method to obtain track data containing semantic labels; perform feature extraction on the track data to construct a spatiotemporal correlation feature matrix;

[0028] A digital twin decision platform is configured to perform a thermal-mechanical coupling simulation based on the track data to obtain simulation results, wherein the simulation results include track thermal expansion coefficient and current track stress field distribution data;

[0029] The adaptive control module is configured to generate optimized control instructions and warning information based on the spatiotemporal correlation characteristic matrix and the simulation results.

[0030] In a third aspect, an embodiment of the present invention further provides an electronic device comprising: a memory and a processor; the processor is configured to read and execute a computer program stored in the memory to implement the steps of the aforementioned rail infrastructure safety assurance method.

[0031] In a fourth aspect, an embodiment of the present invention further provides a computer storage medium, wherein the computer storage medium stores computer executable instructions, and when the computer executable instructions are executed, the steps of the aforementioned rail infrastructure safety assurance method are implemented.

[0032] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include:

[0033] The original track data is collected through a multi-source perception module; the original track data is processed using a bilateral filter and a region growing method to obtain track data containing semantic labels; the track data is feature extracted through a data fusion center to construct a spatiotemporal correlation feature matrix; based on the track data, a thermal coupling simulation is performed through a digital twin decision platform to obtain simulation results; based on the spatiotemporal correlation feature matrix and the simulation results, an adaptive control module is used to generate optimized control instructions and alarm information. Through the present invention, high-precision multi-source perception, intelligent decision analysis, and digital twin simulation are integrated. Through the deep integration of BIM modeling, 3D physical simulation, real-time data monitoring, and predictive maintenance technology, three-dimensional perception and dynamic risk control of all elements, all time periods, and all dimensions of the train operating environment are achieved, which significantly improves the reliability of railway infrastructure and the safety of train operation, and solves the problems of fragmentation of perception systems, insufficient intelligence of diagnostic decisions, and passive risk management in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 This is a flow chart of an embodiment of a method for ensuring the safety of railway infrastructure according to the present invention;

[0036] Figure 2 This is a schematic diagram of the functional modules of an embodiment of a rail infrastructure safety assurance system of the present invention;

[0037] Figure 3 The figure is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0039] To make the objectives, technical solutions and advantages of the present invention more clear, embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] In a first aspect, an embodiment of the present invention provides a method for ensuring the safety of rail infrastructure.

[0041] In one embodiment, referring to Figure 1 , Figure 1 This is a flow chart of an embodiment of the method for ensuring the safety of rail infrastructure according to the present invention. Figure 1 As shown, rail infrastructure safety assurance methods include:

[0042] Step S10, collecting original track data through a multi-source perception module;

[0043] In this embodiment, the multi-source perception module includes a fiber grating sensor, a lidar, an interferometric synthetic aperture radar, a multispectral camera, and an infrared thermal imager.

[0044] High-precision fiber Bragg grating sensors (FBGs) are deployed along the track to monitor the stress, temperature, and deformation data of the track in real time. The strain measurement accuracy of fiber Bragg grating sensors (FBGs) can reach ±0.5με (micro strain, ), the temperature measurement accuracy is ±0.5℃ and the sampling frequency is 10kHz.

[0045] LiDAR (laser radar) and InSAR (interferometric synthetic aperture radar) technologies are used to perform large-scale terrain scanning, combined with three-dimensional point cloud modeling to achieve millimeter-level precision monitoring of the environment around the track and obtain terrain data.

[0046] Deploy multispectral cameras and infrared thermal imagers, and use image recognition algorithms (such as YOLOv5) to detect anomalies such as track foreign objects intruding into the limit and equipment overheating in real time, and obtain environmental image data of the track.

[0047] That is, the multi-source perception module can collect the track's stress data, deformation data, temperature data, terrain data, and environmental image data.

[0048] This embodiment utilizes fiber Bragg grating sensors to synchronously acquire track temperature gradients, resolving the data dimension loss issue in existing technologies. Furthermore, multispectral cameras and infrared thermal imagers are unaffected by rain or fog, enabling track inspection. This addresses the issue of existing equipment's poor environmental adaptability and inability to meet all-weather monitoring requirements.

[0049] The multi-source sensing module uses a GPS clock synchronization unit (nanosecond-level accuracy) to achieve spatiotemporal alignment of vibration signals and video data. The timestamps of the vibration signal (10kHz sampling) and the video frame (30fps) are synchronized using the PTP precision time protocol. The data fusion hub uses a cubic spline interpolation algorithm to align non-uniformly sampled data, with a maximum delay error of ≤0.1ms. All sensors have integrated GPS modules (such as the u-blox ZED-F9T), outputting a 1PPS (pulse per second) signal with a synchronization error of ≤20ns. The video camera also has a built-in hardware encoder that writes frame timestamps into the video stream metadata. This solves the spatiotemporal alignment difficulties encountered in existing technologies: the millisecond-level delay between the vibration signal (10kHz sampling) and the video data (30fps), which results in a high false alarm rate for turnout jams.

[0050] in, .

[0051] Step S20, processing the original track data using a bilateral filter and a region growing method to obtain track data containing semantic labels;

[0052] In this embodiment, after obtaining the stress data, deformation data, temperature data, terrain data, and environmental image data of the track through the multi-source perception module, since the terrain data obtained when LiDAR (laser radar) and InSAR (interferometric synthetic aperture radar) are scanned, the terrain data obtained is point cloud data. Therefore, the terrain data in the original track data is denoised and segmented using a bilateral filter and region growing method to obtain track data containing semantic labels such as track, fasteners, and ballast.

[0053] Among them, the bilateral filter is a nonlinear filter that combines the similarity of the spatial domain and the pixel value domain. It performs filtering by considering the spatial distance and grayscale value difference between pixels. The bilateral filter can retain edge details and effectively suppress noise.

[0054] Region growing is the process of growing groups of pixels or regions into larger regions. Starting with a set of seed points, regions are grown from these points by merging neighboring pixels with similar properties to each seed point, such as intensity, grayscale, texture color, etc.

[0055] The code for processing the original trajectory data using bilateral filter and region growing method is shown below.

[0056] rawpoints=lidar.scan() # Raw point cloud collection

[0057] denoised =bilateral filter(raw points)#Bilateral filtering noise reduction

[0058] segmented=region_growing(denoised) #region growing method to segment track components

[0059] Step S30, extracting features from the track data through a data fusion hub to construct a spatiotemporal correlation feature matrix;

[0060] In some specific embodiments, step S30 includes:

[0061] Step S301, using wavelet transform and Kalman filter algorithm to perform noise reduction and data smoothing on the track data to obtain first track data;

[0062] In this embodiment, wavelet transform is used to eliminate noise from the track data containing semantic labels, and a Kalman filter algorithm is used to smooth the data to obtain the first track data.

[0063] Among them, the wavelet transform (WT) is a new transform analysis method that provides a "time-frequency" window that changes with frequency, making it an ideal tool for time-frequency signal analysis and processing. Its main features are that it can fully highlight certain aspects of the problem through transformation, perform localized analysis of time (space) frequency, and gradually refine the signal (function) at multiple scales through scaling and translation operations, ultimately achieving time subdivision at high frequencies and frequency subdivision at low frequencies. It can automatically adapt to the requirements of time-frequency signal analysis, thus focusing on any detail of the signal.

[0064] Kalman filtering is an algorithm that uses a linear system's state equation and observational data from the system's input and output to optimally estimate the system's state. Because the observational data includes the effects of noise and interference within the system, optimal estimation can also be viewed as a filtering process. Data filtering is a data processing technique that removes noise and restores true data.

[0065] Step S302, using principal component analysis and t-distributed random neighbor embedding algorithm to reduce the dimension of the first track data to obtain second track data;

[0066] In this embodiment, principal component analysis (PCA) is a commonly used data dimensionality reduction technique, which is used to convert a high-dimensional data set into a low-dimensional representation while retaining the main features of the data.

[0067] t-distributed Stochastic Neighbor Embedding (t-SNE) is a nonlinear technique for dimensionality reduction of high-dimensional data. The algorithm is particularly suitable for mapping high-dimensional data into two-dimensional or three-dimensional space for easy visualization.

[0068] The first track data is reduced in dimensionality using principal component analysis and a t-distributed random neighbor embedding algorithm to obtain second track data, so as to remove redundant features in the data and reduce the computational complexity of high-dimensional data.

[0069] Step S303: extract key features from the second track data using a spatiotemporal graph convolutional network to obtain feature track data;

[0070] In this embodiment, the network structure adopts a three-layer spatiotemporal graph convolutional network (ST-GCN), each layer of which contains:

[0071] Spatiotemporal convolution module: The spatial dimension uses a graph convolution kernel (kernel size K=3), and the temporal dimension uses a one-dimensional convolution kernel (kernel size T=5, stride 1);

[0072] Activation function: ReLU;

[0073] Normalization layer: Batch Normalization (BatchNorm);

[0074] Input data: Second track data, format is X∈RN×T×D, where R is a real number, N is the number of nodes, T is the time step, and D is the feature dimension;

[0075] Output data: characteristic orbital data, also known as the spatiotemporal correlation characteristic matrix M∈RN×T×128.

[0076] The execution code of the spatiotemporal graph convolutional network (ST-GCN) is as follows:

[0077] # Space-time alignment pseudocode

[0078] def time_align(

[0079] vibration signal,#vibration signal (10kHz sampling)

[0080] video frame#Video frame (30fps) ):

[0082] timestamp = get precise timestamp() #Synchronize to GPS clock (nanosecond accuracy)

[0083] return aligned_data#Feature track data after key feature extraction

[0084] Step S304: constructing a spatiotemporal graph based on the characteristic trajectory data through a graph neural network library;

[0085] In this embodiment, the graph structure update rule is: when the deformation difference between nodes is greater than 3 mm or the temperature gradient difference is greater than 2°C / m, edges are dynamically added / deleted.

[0086] Building tools and algorithms:

[0087] PyTorch Geometric: for efficiently building and processing spatiotemporal graphs, supporting GPU acceleration;

[0088] DTW (Dynamic Time Warping) algorithm: Calculates the similarity of node time series data and solves the problem of asynchronous sampling.

[0089] Based on the feature trajectory data, a spatiotemporal graph G = (V, E) is constructed using a graph neural network library, such as the PyTorch Geometric library.

[0090] 1. Node definition:

[0091] The track is divided into nodes v1, v2, ..., v with a spacing of 50m. N , the characteristic orbital data of each node includes:

[0092] Vibration spectrum: Extract the amplitude of the first five main frequencies (0-500Hz) through FFT;

[0093] Temperature gradient: (lateral temperature change rate), (time temperature change rate);

[0094] LiDAR (Light Laser Detection and Ranging) deformation vector: three-dimensional deformation (Δx, Δy, Δz), accuracy ±1mm;

[0095] 2. Edge connection rules:

[0096] If the distance between two nodes is ≤100m and the deformation correlation coefficient ρ is ≥0.7, then the edge e is established. ij ; Edge ij ∈E: Calculate the spatiotemporal correlation between nodes based on the dynamic time warping (DTW) algorithm, and the edge weight w ij =1+DTW(v i , v j ), the update cycle is 1 minute.

[0097] 3. Construction of space-time graph:

[0098] The Python library NetworkX is used to generate a spatiotemporal graph G = (V, E) with N ≤ 2000 nodes (covering 100 km of track); edge weights are updated every 5 minutes to respond to deformation propagation.

[0099] Step S305: generating a spatiotemporal correlation feature matrix based on the characteristic trajectory data and the spatiotemporal graph.

[0100] In this embodiment, a spatiotemporal graph convolution operation is first performed to obtain the time series features of each node; then, feature aggregation is performed on the time series features of each node to obtain a node-level feature vector; finally, a spatiotemporal correlation feature matrix is ​​constructed based on the node-level feature vector.

[0101] Spatiotemporal graph convolution operation: The feature trajectory data is processed through a three-layer spatiotemporal graph convolutional network (ST-GCN) to obtain the time series features of each node. Each layer of the spatiotemporal graph convolutional network (ST-GCN) performs the following steps: spatial convolution: graph convolution based on the adjacency matrix A, with a kernel size of 3 and a weight matrix W∈RD×D′ (R is a real number, D=64, D′=128); temporal convolution: a one-dimensional convolution kernel (size 5, stride 1) to extract time dimension features; activation and normalization: ReLU activation function and batch normalization (BatchNorm);

[0102] Feature aggregation: Adaptive average pooling (AdaptiveAvgPool1d) is performed on the time series features of each node, and the node-level feature vector h is output. i ∈R128;

[0103] Construct a spatiotemporal correlation feature matrix: sort all node-level feature vectors by spatial position, extract high-order features through the spatiotemporal graph convolutional network (ST-GCN), and construct a spatiotemporal correlation feature matrix M∈RN×T×128, where N is the number of nodes and T=6 represents 6 time steps (30-minute window).

[0104] Step S40: performing a thermal-mechanical coupling simulation based on the track data through a digital twin decision platform to obtain a simulation result;

[0105] In some specific embodiments, step S40 includes:

[0106] Step S401, using a semantic segmentation network to identify and classify the track data to determine the faulty component;

[0107] Step S402: adjusting the parameters of the faulty component through the building information model in the digital twin platform;

[0108] Step S403 : performing a thermomechanical coupling simulation based on the adjusted parameters to obtain the track thermal expansion coefficient and the current track stress field distribution data.

[0109] In this embodiment, the digital twin decision-making platform is built based on a building information model (BIM) and a physics engine (such as NVIDIA PhysX). The digital twin decision-making platform includes geometric, behavioral, and physical models. The geometric model uses Revit (a software suite) to implement parametric modeling of equipment such as signals and switches, with LOD400 accuracy. The behavioral model uses Petri nets (a mathematical representation of discrete parallel systems) to construct interlocking logic simulations, supporting dynamic deduction of route arrangements and signal state switching. The physical model uses ANSYS (a large-scale, general-purpose finite element analysis (FEA) software) to perform coupled simulations of track stress and thermodynamic fields, achieving a computational accuracy of 0.01 mm.

[0110] A semantic segmentation network (U-Net++) was used to identify and classify point cloud data from LiDAR (laser radar) scans to determine the faulty component. For example, a LiDAR scan detected a 2.3mm lateral displacement of the sleeper, which the U-Net++ classification module identified as a loose fastener.

[0111] The structural parameters of the semantic segmentation network (U-Net++) are shown below.

[0112] model = UnetPlusPlus(

[0113] backbone='ResNet34'

[0114] input_shape = (1024, 1024, 3)

[0115] classes=6 #Track / Catnet / Turnout / Fastener / Ballast / Other )

[0117] The fastener friction coefficient is adjusted through the building information model (BIM model) in the digital twin platform. Based on the adjusted fastener friction coefficient, thermal-mechanical coupling simulation is performed through the physical model in the digital twin platform to obtain the track thermal expansion coefficient and current track stress field distribution data.

[0118] The thermal coupling simulation code is as follows:

[0119] graph TB

[0120] A[input state vector]-->B(ANSYS solver)

[0121] B-->C{deformation prediction}

[0122] -->|>3mm| D[Trigger MPC control]

[0123] C-->|≤3mm|E[Generate monitoring report]

[0124] Step S50: Based on the spatiotemporal correlation characteristic matrix and the simulation results, an adaptive control module is used to generate an optimization control instruction and an alarm message.

[0125] In some specific embodiments, step S50 includes:

[0126] Step S501, obtaining historical track data within a preset time period;

[0127] Step S502: Based on the historical track data and the simulation results, a track deformation prediction value and a confidence interval of the track deformation prediction value are obtained through a long sequence prediction model;

[0128] Step S503: When the track deformation prediction value is not within the confidence interval, the spatiotemporal correlation characteristic matrix and the simulation results are optimized and solved by the target optimization function included in the adaptive control module to obtain an optimized control instruction;

[0129] Step S504: querying a corresponding relationship table between track deformation values ​​and deformation levels based on the track deformation prediction value, determining the track deformation level, and obtaining corresponding warning information based on the track deformation level.

[0130] In this example, using a preset timeframe of 72 hours, simulation results (track thermal expansion coefficient and current track stress field distribution data) and 72 hours of historical track data were input into a long-sequence prediction model to obtain track deformation predictions and confidence intervals. The long-sequence prediction model (LSTM + Attention) was built using the Transformer model architecture.

[0131] The data structure code for inputting the long sequence prediction model is shown below.

[0132] input_data = {

[0133] 'temp_history':np.array([-15,-18,...,20]), #Track data sequence for the past 72 hours

[0134] 'Stress_distribution':(256,256)tensor,#Current track stress field distribution

[0135] 'Material_params':{'thermal_expansion':11.7e-6} # Rail thermal expansion coefficient

[0136] }

[0137] When the track deformation prediction value is not within the confidence interval of the track deformation prediction value, the objective optimization function included in the adaptive control module is used to optimize the spatiotemporal correlation feature matrix and simulation results to obtain the optimized control instructions. The objective optimization function is:

[0138]

[0139] Where u represents the set of control variables, including train speed v, braking force B, and turnout compensation angle θ; J represents the total cost function, which measures the weighted sum of safety, equipment loss, and energy consumption. represents the minimum total cost corresponding to the set of control variables, represents the longitudinal impulse force when the step size of the model predictive control rolling optimization time domain is k, represents the wheel-rail wear index when the time domain step size of the model predictive control rolling optimization is k, represents the energy consumption when the step size of the model predictive control rolling optimization horizon is k, 、 and represents the weight coefficient (α+β+γ=1), and k represents the receding optimization step size of the model predictive control. Preferably, α=0.6, β=0.3, and γ=0.1.

[0140] It represents the longitudinal impulse force when the step length of the model predictive control rolling optimization time domain is k, and the calculation formula is , where m is the mass of the train, a is the acceleration, and c is the air resistance coefficient. is the train speed;

[0141] The wheel-rail wear index when the time step of the rolling optimization of the model predictive control is k is calculated based on the Archard model. , Fn is the normal force, s is the sliding distance, and H is the material hardness;

[0142] represents the energy consumption when the step size of the model predictive control rolling optimization horizon is k, , where η is the braking energy recovery efficiency, B is the braking force, and v is the train speed.

[0143] Among them, the constraints of the spatiotemporal correlation feature matrix M are: ; is the correlation feature matrix of the kth time step, is the correlation feature matrix of the k-1th time step.

[0144] Simulation result constraints: , the thermal stress of the track does not exceed the yield strength.

[0145] Represents the threshold value of the change rate of the spatiotemporal feature matrix , determined by the fatigue properties of the track material;

[0146] represents the track thermal stress output from the thermal-mechanical coupling simulation, is the yield strength of the rail, for example, Q345 steel has a yield strength of 345 MPa.

[0147] Input data flow to the target optimization function, the input data flow includes:

[0148] 1. Multi-source perception module: real-time collection of vibration, temperature, deformation and other data (frequency 10kHz~30Hz).

[0149] 2. Model prediction: Based on LSTM or Kalman filtering, generate the next 6-step deformation prediction value Dpred and confidence interval [D low , D high ].

[0150] 3. Thermal-mechanical coupling simulation: Real-time calculation of rail thermal stresses using finite element analysis software (ANSYS) solver .

[0151] Time-align and pre-process the input data stream:

[0152] 1. All input data streams are synchronized with the GPS clock (error ≤ 0.1ms) to ensure alignment within the optimization period (≤ 100ms).

[0153] 2. Data preprocessing (filtering, dimensionality reduction) compresses the raw data into a 128-dimensional state variable x∈R128, which contains key features (vibration spectrum, temperature gradient, deformation vector).

[0154] Real-time guarantee of rolling optimization cycle (≤100ms):

[0155] Calculation process:

[0156] 1. Data collection and prediction (20ms):

[0157] Multi-source sensor data acquisition (≤5ms);

[0158] The model predicts the next 6 steps of deformation and generates confidence intervals (≤10ms);

[0159] Thermal-mechanical coupling simulation calculates the current stress field (≤5ms).

[0160] 2. Optimize question construction (10ms):

[0161] Define the objective function J(u) and constraints (deformation threshold, stress limit);

[0162] The 128-dimensional state variable x is used as the input of the optimization model.

[0163] 3. Optimized solution (60ms):

[0164] Use the SQP algorithm to solve the constrained optimization problem and generate the next 6 control instructions u1, u2, ..., u6.

[0165] 4. Command issuance (10ms):

[0166] Only the first step command u1 is applied (such as adjusting the braking force or the turnout angle);

[0167] The remaining instructions u2, ..., u6 are used as initial guesses for the next cycle.

[0168] Key design:

[0169] Model lightweighting: Prediction models and simulators use simplified versions (such as reduced-order models) to ensure computational speed.

[0170] Hardware acceleration: Compress thermal simulation and optimization solution time through FPGA or GPU parallel computing.

[0171] Dynamic adjustment of the control time domain (6 steps):

[0172] Time domain division:

[0173] Each step corresponds to an optimization period (100ms), and the total duration of the control time domain is T = 6 × 100ms = 600ms.

[0174] In each optimization, the predictions and constraints for the next 6 steps are updated based on the latest data.

[0175] Rolling strategy:

[0176] Initial optimization window: t0~t0+600ms;

[0177] Next cycle window: t0+100ms~t0+700ms, and so on.

[0178] Adaptive Adjustment:

[0179] If the confidence interval width Dhigh−Dlow>3mm, the control time domain is automatically expanded to 8 steps (to increase fault tolerance).

[0180] To optimize control instructions, specifically, 1. Input data preprocessing:

[0181] The spatiotemporal correlation feature matrix M∈RN×6×128 is expanded into M′∈R6N×128 according to the time step;

[0182] Key parameters extracted from thermal-mechanical coupling simulation results: thermal expansion coefficient α thermal , maximum stress σ max , temperature field distribution T(x, y);

[0183] 2. Optimize variable definitions:

[0184] Control variable u = [v, B, θ], where:

[0185] v: target train speed (range 0≤v≤120km / h);

[0186] B: braking torque (range 0≤B≤500kN);

[0187] θ: turnout compensation angle (-5°≤θ≤+5°);

[0188] 3. Objective function construction:

[0189] .

[0190] 4 Constraint loading:

[0191] Spatiotemporal correlation constraints: (Orbital deformation rate limit);

[0192] Thermal stress constraints: ≤345MPa (Q345 steel yield strength);

[0193] Dynamic constraint: |a|≤0.3g (acceleration limit, g=9.8m / s 2 );

[0194] 5. Optimization solution:

[0195] The sequential quadratic programming (SQP) algorithm is used to solve the constrained nonlinear optimization problem. The initial guess value u0 comes from the historical optimal instruction library.

[0196] Output optimization variables , which is sent to the train control system and turnout actuator as an optimization control instruction.

[0197] Furthermore, after receiving the control instructions, the control instructions (such as route arrangement and signal state change instructions) need to be verified by the deployed blockchain (Hyperledger Fabric) evidence system. Node consensus verification ensures that control instructions cannot be tampered with.

[0198] The blockchain verification process code is as follows:

[0199] Function executecommand(bytes memory cmd)public {

[0200] require(validateHash(cmd.hash),"Invalid hash");

[0201] require(checksignatures(cmd.sigs,2),"Insufficient signature")

[0202] emit CommandExecuted(cmd); / / Command execution event

[0203] }

[0204] cmd.hash is the simulated hash value (SHA-256), and cmd.sigs is the edge node signature (must be greater than or equal to 2).

[0205] After obtaining the track deformation prediction value, it is also necessary to query the correspondence table between the track deformation value and the deformation level based on the track deformation prediction value, determine the track deformation level, and obtain corresponding alarm information based on the track deformation level.

[0206] The corresponding relationship between track deformation value and deformation level is shown in Table 1.

[0207] Table 1

[0208]

[0209] If the track deformation level is level 1, the warning light will illuminate, the web interface will display a trend curve, and the AR model will be highlighted. A log will be recorded, prompting the start of a manual inspection plan.

[0210] If the track deformation reaches Level 2, the warning light illuminates, and a pop-up alert appears in the red area of ​​the HUD (Head-Up Display) in the 3D situation icon. The system performs a self-check and activates the preparatory control strategy. HUD (Head-Up Display) is a display technology that projects information in front of the user's line of sight, allowing them to view it without looking down.

[0211] If the track deformation level is level three, the alarm light will light up, the full screen will flash, the alarm voice will be broadcast, the speed will be automatically limited, and the control strategy will be optimized through the model predictive control (MPC) algorithm.

[0212] If the track deformation level is level four, the emergency light will come on, the power supply to non-critical equipment will be cut off, and the EMU emergency brake will be triggered.

[0213] In another embodiment, based on the track deformation prediction value and its confidence interval, combined with real-time environmental parameters (including track temperature gradient ΔT, stress field distribution σ max and the main frequency offset of the vibration spectrum Δf), determine the track deformation level and generate corresponding alarm information.

[0214] For example, the warning light is blue, the track deformation is 12 mm, the temperature gradient is 35°C / m, and the trend curve AR model is highlighted on the web interface.

[0215] The alarm light turns yellow, the track deformation is 23mm, the vibration energy increases by 20%, and a HUD pop-up window with the red area of ​​the 3D situation icon is displayed;

[0216] The alarm light turns orange, the deformation is 35mm, the stress is >250MPa, the full screen flashes and the alarm voice is broadcast;

[0217] The alarm light turns red for emergency alarm. If the deformation is greater than 5mm and the communication is interrupted for more than 5s, the power supply to non-critical equipment will be cut off, triggering the emergency braking of the EMU.

[0218] In this embodiment, raw track data is collected through a multi-source perception module; the raw track data is processed using a bilateral filter and a region growing method to obtain track data containing semantic labels; the track data is feature extracted through a data fusion center to construct a spatiotemporal correlation feature matrix; based on the track data, a thermal coupling simulation is performed through a digital twin decision platform to obtain simulation results; based on the spatiotemporal correlation feature matrix and the simulation results, an adaptive control module is used to generate optimized control instructions and alarm information. Through this embodiment, high-precision multi-source perception, intelligent decision analysis, and digital twin simulation are integrated. Through the deep integration of BIM modeling, 3D physical simulation, real-time data monitoring, and predictive maintenance technology, three-dimensional perception and dynamic risk control of all elements, all time periods, and all dimensions of the train operating environment are achieved, which significantly improves the reliability of railway infrastructure and the safety of train operation, and solves the problems of perception system fragmentation, insufficient intelligence in diagnostic decision-making, and passive risk management in related technologies.

[0219] Optionally, in one embodiment, after step S50, the following steps are included:

[0220] Get historical track data within a preset time period;

[0221] Based on the historical track data and the simulation results, a track deformation prediction value and a confidence interval of the track deformation prediction value are obtained through a long sequence prediction model;

[0222] Building a distributed diagnostic model based on a federated learning framework, and aggregating the distributed diagnostic model through a central server to obtain a global model;

[0223] When the difference between the vibration frequencies and amplitudes of the plurality of track components detected based on the global model is less than a preset value, the root cause probabilities of the failures of the plurality of track components are inferred using a knowledge graph and a root cause analysis method;

[0224] Based on the track deformation prediction value, the confidence interval of the track deformation prediction value and the root cause probability, the risk probability distribution of the track is obtained.

[0225] In this embodiment, taking the preset time length of 72 hours as an example, the simulation results (track thermal expansion coefficient and current track stress field distribution data) and 72 hours of historical track data are input into the long sequence prediction model to obtain the track deformation prediction value and the confidence interval of the track deformation prediction value.

[0226] A distributed diagnostic model is built based on the federated learning framework (FATE platform), and the distributed diagnostic model is aggregated through the central server to obtain a global model.

[0227] Specifically, the central server distributes the parameters of the distributed diagnostic model to each client. The client independently trains the model using local data and updates the parameters (e.g., by gradient descent). The server collects the updated parameters from the clients and takes a weighted average based on the amount of data to generate an initial global model. This process is repeated until the model converges, resulting in the final global model.

[0228] When the final global model detects that the difference between the vibration frequency and amplitude of multiple track components is less than a preset value, root cause analysis and knowledge graphs are used to infer the root cause probability of the failure of these multiple track components. For example, when the final global model detects similar vibration characteristics in multiple track components, that is, the difference between the vibration frequency and amplitude of these multiple track components is less than a preset value, the knowledge graph is used to search the historical case library. Combined with root cause analysis and environmental parameters such as the current temperature of -4°C and 85% humidity, the probability of fastener loosening caused by frost heave is inferred to be 82%. Knowledge graphs (KGs) are a key component of artificial intelligence technology. They are a structured, semantic knowledge representation method that helps computers understand and process human language. Root cause analysis (RCA) is a structured problem-solving method used to gradually identify and resolve the root cause of a problem.

[0229] Based on the track deformation prediction value, the confidence interval of the track deformation prediction value and the root cause probability, the calculation process of the track risk probability distribution is obtained:

[0230] 1. Input parameter definition:

[0231] Track deformation prediction value (Dpred): The predicted value of track deformation in a certain period of time in the future (unit: mm), such as Dpred=5.2mm.

[0232] Confidence interval (CI): The statistical range of the deformation prediction value, assuming a normal distribution, for example, 95% CI = [4.0, 6.4] mm (mean μ = 5.2, standard deviation σ = 0.6).

[0233] Root cause probability (P(Ri)): The set of root cause probabilities inferred by the Bayesian network, such as: P(R1=fastener loosening)=0.62, P(R2=ballast settlement)=0.28, P(R3=other)=0.10.

[0234] 2. Risk threshold setting:

[0235] Safety threshold (D th ): The maximum value allowed for track deformation, set according to the track type (such as D th =6.0mm). Risk level classification:

[0236] Low risk: Dpred < 4.0 mm; medium risk: 4.0 ≤ Dpred < 6.0 mm; high risk: Dpred ≥ 6.0 mm.

[0237] 3. Calculate the basic risk probability:

[0238] Based on the normal distribution of track deformation prediction value D∼N (μ=5.2, σ2=0.62), the deformation exceeding the threshold D is calculated. th The probability P is:

[0239] .

[0240] Where Φ is the cumulative function of the standard normal distribution.

[0241] 4. Root cause condition risk correction:

[0242] Different root causes have different impacts on track deformation, and the risk probability needs to be adjusted:

[0243] Loose fasteners (R1): prone to rapid deformation, risk factor ;

[0244] Ballast settlement (R2): deformation develops slowly, risk factor ;

[0245] Other (R3): Risk factor .

[0246] Conditional risk probability:

[0247] .

[0248] 6. Risk probability distribution output:

[0249] The final risk probability distribution after root cause correction is divided into levels:

[0250] ;

[0251] ;

[0252] .

[0253] Optionally, in one embodiment, after step S50, the method further includes:

[0254] When an alarm is received, the location coordinates of the faulty track component are queried through the digital twin platform based on the alarm information, and a maintenance work order for the faulty track component is generated using the equipment remaining service life prediction model;

[0255] The location coordinates and the maintenance work order are sent to a preset terminal.

[0256] In this embodiment, a geographic information system is integrated with a building information model (BIM) model, enabling bidirectional querying of equipment location coordinates (WGS84) and attribute data. Therefore, upon receiving an alarm, the location coordinates of the faulty track component can be queried based on the predicted track deformation value. Using the equipment's remaining useful life prediction model, a maintenance work order for the faulty track component is generated. The coordinates and work order are then sent to a pre-defined terminal, allowing maintenance personnel to promptly perform maintenance on the faulty track component.

[0257] In a second aspect, an embodiment of the present invention further provides a rail infrastructure safety assurance system.

[0258] In one embodiment, referring to Figure 2 , Figure 2 This is a schematic diagram of the components of an embodiment of the rail infrastructure safety assurance system of the present invention. Figure 2 As shown in Figure 1, the rail infrastructure safety assurance system includes: multi-source perception module, data fusion hub, digital twin platform and adaptive control module.

[0259] A multi-source perception module configured to collect raw track data;

[0260] A data fusion hub is configured to process the raw track data using a bilateral filter and a region growing method to obtain track data containing semantic labels; perform feature extraction on the track data to construct a spatiotemporal correlation feature matrix;

[0261] A digital twin decision platform is configured to perform a thermal-mechanical coupling simulation based on the track data to obtain simulation results, wherein the simulation results include track thermal expansion coefficient and current track stress field distribution data;

[0262] The adaptive control module is configured to generate optimized control instructions and warning information based on the spatiotemporal correlation characteristic matrix and the simulation results.

[0263] Optionally, in one embodiment, the multi-source perception module includes a fiber grating sensor, a laser radar, an interferometric synthetic aperture radar, a multispectral camera, and an infrared thermal imager;

[0264] The original track data includes: track stress, deformation, temperature, terrain and environmental image data.

[0265] Optionally, in one embodiment, the data fusion hub is specifically configured to:

[0266] performing noise reduction and data smoothing on the track data using a wavelet transform and a Kalman filter algorithm to obtain first track data;

[0267] performing dimensionality reduction on the first track data using principal component analysis and a t-distributed random neighbor embedding algorithm to obtain second track data;

[0268] Extract key features of the second track data using a spatiotemporal graph convolutional network to obtain feature track data;

[0269] Based on the characteristic trajectory data, a spatiotemporal graph is constructed using a graph neural network library;

[0270] A spatiotemporal correlation feature matrix is ​​generated based on the characteristic trajectory data and the spatiotemporal graph.

[0271] Optionally, in one embodiment, the digital twin decision platform is specifically configured to:

[0272] Using a semantic segmentation network to identify and classify the track data to determine the faulty component;

[0273] Adjusting the parameters of the faulty component through a building information model in a digital twin platform;

[0274] Based on the adjusted parameters, thermal-mechanical coupling simulation is performed to obtain the track thermal expansion coefficient and current track stress field distribution data.

[0275] Optionally, in one embodiment, the simulation results include the rail thermal expansion coefficient and current rail stress field distribution data, and the adaptive control module is specifically configured to:

[0276] Get historical track data within a preset time period;

[0277] Based on the historical track data and the simulation results, a track deformation prediction value and a confidence interval of the track deformation prediction value are obtained through a long sequence prediction model;

[0278] When the track deformation prediction value is not within the confidence interval, optimizing and solving the spatiotemporal correlation characteristic matrix and the simulation results through the target optimization function included in the adaptive control module to obtain an optimized control instruction;

[0279] Based on the track deformation prediction value, a corresponding relationship table between track deformation values ​​and deformation levels is queried to determine the track deformation level, and corresponding warning information is obtained based on the track deformation level.

[0280] Optionally, in one embodiment, the objective optimization function is:

[0281]

[0282] In the formula, u represents the set of control variables, J represents the total cost function, represents the minimum total cost corresponding to the set of control variables, represents the longitudinal impulse force when the step size of the model predictive control rolling optimization time domain is k, represents the wheel-rail wear index when the time domain step size of the model predictive control rolling optimization is k, represents the energy consumption when the step size of the model predictive control rolling optimization horizon is k, 、 and represents the weight coefficient, and k represents the step size of the rolling optimization time domain of the model predictive control.

[0283] Optionally, in one embodiment, the rail infrastructure safety assurance system further includes a risk probability distribution calculation module configured to:

[0284] Get historical track data within a preset time period;

[0285] Based on the historical track data and the simulation results, a track deformation prediction value and a confidence interval of the track deformation prediction value are obtained through a long sequence prediction model;

[0286] Building a distributed diagnostic model based on a federated learning framework, and aggregating the distributed diagnostic model through a central server to obtain a global model;

[0287] When the difference between the vibration frequencies and amplitudes of the plurality of track components detected based on the global model is less than a preset value, the root cause probabilities of the failures of the plurality of track components are inferred using a knowledge graph and a root cause analysis method;

[0288] A risk probability distribution of the track is obtained based on the track deformation prediction value, the confidence interval of the track deformation prediction value, and the root cause probability.

[0289] Optionally, in one embodiment, the rail infrastructure safety assurance system further includes an alarm module configured to:

[0290] When an alarm is received, the location coordinates of the faulty track component are queried through the digital twin platform based on the alarm information, and a maintenance work order for the faulty track component is generated using the equipment remaining service life prediction model;

[0291] The location coordinates and the maintenance work order are sent to a preset terminal.

[0292] Among them, the functional implementation of each module in the above-mentioned rail infrastructure safety assurance system corresponds to the various steps in the above-mentioned rail infrastructure safety assurance method embodiment, and their functions and implementation processes will not be repeated here one by one.

[0293] In a third aspect, an embodiment of the present invention further provides an electronic device, the structure of which is as follows: Figure 3 As shown, it includes: a memory and a processor, and the processor is used to read and execute the computer program stored in the memory to implement the aforementioned rail infrastructure safety assurance method.

[0294] In a fourth aspect, an embodiment of the present invention further provides a computer storage medium, wherein the computer storage medium stores computer executable instructions, and when the computer executable instructions are executed, the aforementioned rail infrastructure safety assurance method is implemented.

[0295] Finally, it should be noted that some of the processes described in the embodiments of the present invention include multiple operations or steps that appear in a specific order. However, it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present invention, or may be executed in parallel. The sequence numbers of the operations are only used to distinguish different operations and do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0296] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for ensuring the safety of railway infrastructure, characterized in that: The method comprises: Collect raw track data through multi-source perception modules; Processing the original track data using a bilateral filter and a region growing method to obtain track data containing semantic labels, wherein the semantic labels include track, fastener, and ballast; The orbital data is subjected to noise reduction and data smoothing using a wavelet transform and a Kalman filter algorithm to obtain first orbital data; the first orbital data is subjected to dimensionality reduction using a principal component analysis method and a t-distributed random neighbor embedding algorithm to obtain second orbital data; key features of the second orbital data are extracted using a spatiotemporal graph convolutional network to obtain characteristic orbital data; a spatiotemporal graph is constructed based on the characteristic orbital data using a graph neural network library; a spatiotemporal graph convolution operation is performed on the characteristic orbital data using the spatiotemporal graph to obtain time series features of each node; feature aggregation is performed on the time series features of each node to obtain a node-level feature vector; and a spatiotemporal correlation feature matrix is ​​constructed based on the node-level feature vector. Based on the track data, a thermal-mechanical coupling simulation is performed through a digital twin decision platform to obtain simulation results, wherein the simulation results include track thermal expansion coefficient and current track stress field distribution data; Acquiring historical track data within a preset time period; obtaining a track deformation prediction value and a confidence interval of the track deformation prediction value using a long sequence prediction model based on the historical track data and the simulation results; when the track deformation prediction value is not within the confidence interval, optimizing and solving the spatiotemporal correlation feature matrix and the simulation results using a target optimization function included in an adaptive control module to obtain an optimized control instruction; querying a correspondence table between track deformation values ​​and deformation levels based on the track deformation prediction value to determine the track deformation level, and obtaining corresponding alarm information based on the track deformation level; The objective optimization function is: Where, represents the set of control variables, represents the total cost, represents the minimum total cost corresponding to the set of control variables, The step length of the rolling optimization time domain of the model predictive control is The longitudinal impulse force, The step length of the rolling optimization time domain of the model predictive control is The wheel-rail wear index at The step length of the rolling optimization time domain of the model predictive control is Energy consumption when 、 and represents the weight coefficient, Represents the receding optimization horizon step size of model predictive control.

2. The rail infrastructure safety assurance method according to claim 1, characterized in that: The multi-source perception module includes a fiber grating sensor, a lidar, an interferometric synthetic aperture radar, a multispectral camera, and an infrared thermal imager; The original track data includes: track stress, deformation, temperature, terrain and environmental image data.

3. The rail infrastructure safety assurance method according to claim 1, characterized in that: Based on the track data, a thermal-mechanical coupling simulation is performed through a digital twin decision platform to obtain simulation results, including: Using a semantic segmentation network to identify and classify the track data to determine the faulty component; Adjusting the parameters of the faulty component through a building information model in a digital twin platform; Based on the adjusted parameters, thermal-mechanical coupling simulation is performed to obtain the track thermal expansion coefficient and current track stress field distribution data.

4. The rail infrastructure safety assurance method according to claim 1, characterized in that: After performing thermal-mechanical coupling simulation based on the track data through the digital twin decision platform and obtaining simulation results, the method includes: Get historical track data within a preset time period; Based on the historical track data and the simulation results, a track deformation prediction value and a confidence interval of the track deformation prediction value are obtained through a long sequence prediction model; Building a distributed diagnostic model based on a federated learning framework, and aggregating the distributed diagnostic model through a central server to obtain a global model; When the difference between the vibration frequencies and amplitudes of the plurality of track components detected based on the global model is less than a preset value, the root cause probabilities of the failures of the plurality of track components are inferred using a knowledge graph and a root cause analysis method; A risk probability distribution of the track is obtained based on the track deformation prediction value, the confidence interval of the track deformation prediction value, and the root cause probability.

5. The rail infrastructure safety assurance method according to claim 1, characterized in that: After generating the optimization control instruction and the warning information through the adaptive control module based on the spatiotemporal correlation feature matrix and the simulation results, the method further includes: When an alarm is received, the location coordinates of the faulty track component are queried through the digital twin platform based on the alarm information, and a maintenance work order for the faulty track component is generated using the equipment remaining service life prediction model; The location coordinates and the maintenance work order are sent to a preset terminal.

6. A rail infrastructure safety assurance system, characterized in that: The system is used to implement the steps of the method according to any one of claims 1 to 5, and the system includes: a multi-source perception module, a data fusion hub, a digital twin platform and an adaptive control module. A multi-source perception module configured to collect raw track data; A data fusion hub is configured to process the raw track data using a bilateral filter and a region growing method to obtain track data containing semantic labels; perform feature extraction on the track data to construct a spatiotemporal correlation feature matrix; The data fusion hub is specifically configured to: performing noise reduction and data smoothing on the orbital data using a wavelet transform and a Kalman filter algorithm to obtain first orbital data; performing dimensionality reduction on the first orbital data using a principal component analysis method and a t-distributed random neighbor embedding algorithm to obtain second orbital data; performing key feature extraction on the second orbital data using a spatiotemporal graph convolutional network to obtain characteristic orbital data; constructing a spatiotemporal graph based on the characteristic orbital data using a graph neural network library; and generating a spatiotemporal correlation feature matrix based on the characteristic orbital data and the spatiotemporal graph; A digital twin decision platform is configured to perform a thermal-mechanical coupling simulation based on the track data to obtain simulation results, wherein the simulation results include track thermal expansion coefficient and current track stress field distribution data; The adaptive control module is configured to generate optimized control instructions and warning information based on the spatiotemporal correlation characteristic matrix and the simulation results.

7. The system according to claim 6, characterized in that The digital twin decision-making platform is specifically configured to: Using a semantic segmentation network to identify and classify the track data to determine the faulty component; Adjusting the parameters of the faulty component through a building information model in a digital twin platform; Based on the adjusted parameters, thermomechanical coupling simulation is performed to obtain the track thermal expansion coefficient and current track stress field distribution data.

8. The system according to claim 6, wherein: The adaptive control module is specifically configured to: Get historical track data within a preset time period; Based on the historical track data and the simulation results, a track deformation prediction value and a confidence interval of the track deformation prediction value are obtained through a long sequence prediction model; When the track deformation prediction value is not within the confidence interval, optimizing and solving the spatiotemporal correlation characteristic matrix and the simulation results through the target optimization function included in the adaptive control module to obtain an optimized control instruction; Based on the track deformation prediction value, a corresponding relationship table between track deformation values ​​and deformation levels is queried to determine the track deformation level, and corresponding warning information is obtained based on the track deformation level.

9. An electronic device, characterized in that: include: memory and processor; The processor is configured to read and execute the computer program stored in the memory to implement the steps of the rail infrastructure safety assurance method according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which, when executed, implement the steps of the rail infrastructure safety assurance method according to any one of claims 1 to 5.

Images