Method and device for automatically updating service state of infrastructure of high-speed railway line

Abstract

The invention discloses a method and a device for automatically updating the service state of infrastructure of a high-speed railway line. The method comprises the following steps: acquiring acquired data and performing data synchronization; identifying cracks in the track image data; extracting track geometric deformation characteristics in the track point cloud data; performing data fusion to obtain a freeze-thaw cycle frequency matrix and an axle load distribution load matrix; constructing a structure state model, and calculating crack expansion data; track smoothness and track durability are evaluated; constructing a three-dimensional twinborn model; an automatic updating model is obtained; outputting a visual evaluation result by using the automatic updating model and generating a maintenance strategy; and carrying out maintenance based on the maintenance strategy, and uploading recheck data to the automatic updating model after maintenance so as to update the model. According to the method, the detection efficiency of the service state of the high-speed railway line infrastructure can be remarkably improved, the model can be dynamically updated in real time, the model accuracy is high, and powerful technical support is provided for intelligent operation and maintenance of the high-speed railway infrastructure.

Description

Technical Field

[0001] The present application belongs to the field of high-speed railway detection technology, and more specifically, relates to a method and device for automatically updating the service status of high-speed railway line infrastructure. Background Art

[0002] High-speed railways are the lifeblood of China's transportation system. The long-term, safe service of their infrastructure (such as tracks, bridges, tunnels, and roadbeds) is crucial for ensuring efficient train operations. As my country's high-speed rail mileage surpasses 40,000 kilometers, infrastructure is facing increasingly complex challenges in service condition monitoring. These challenges include: 1. Increased dynamic damage: Factors such as high-frequency train vibration, thermal deformation, and groundwater erosion are leading to frequent problems such as concrete cracking, rail wear, and tunnel lining loss. 2. Explosive data collection: New technologies such as laser scanning, drone inspections, and smart sensors generate massive amounts of data (e.g., terabytes of point clouds, images, and vibration signals daily). 3. Evolved operational and maintenance decision-making: A shift from "periodic maintenance" to "condition-based maintenance" requires real-time assessment of structural health and risk prediction.

[0003] Currently, the industry generally adopts two types of high-speed railway service status inspection methods, but both have significant technical bottlenecks: 1) Traditional manual inspection and static models: ① Data lag: Relying on manual visual inspection (such as crack ruler measurement), the inspection cycle is as long as 1-3 months, and it is unable to capture sudden damage (such as roadbed settlement after heavy rain); ② Model rigidity: BIM models established based on historical data cannot be automatically updated, and the assessment fails when new building materials are applied (such as carbon fiber reinforcement) or the environment changes (such as freeze-thaw cycles); 2) Single-source data-driven assessment system: ① Data fragmentation: Laser scanning, image recognition, and sensor monitoring operate independently. For example, laser point clouds can capture structural deformation but cannot identify corrosion inside cracks; cameras can detect surface cracks but have difficulty quantifying concrete strength degradation; ② Lack of interactivity: Assessment results are output as PDF reports, and it is impossible to dynamically mark risk areas or associate maintenance plans in the 3D model.

[0004] Therefore, there is an urgent need for a method that can solve the problems of rigid indicators and data fragmentation in existing methods, and can significantly improve the accuracy of high-speed rail infrastructure status detection and the efficiency of operation and maintenance throughout the entire life cycle. Summary of the Invention

[0005] In response to the above technical problems, the present application provides a method and device for automatically updating the service status of high-speed railway line infrastructure.

[0006] The technical solutions provided in this application are as follows: In a first aspect, a method for automatically updating the service status of high-speed railway line infrastructure is provided, comprising: Acquire collected data and synchronize it to obtain synchronized data including track point cloud data, track image data, temperature and humidity data, and vibration data; Identify cracks in track image data and output crack data; Extract track geometric deformation features from track point cloud data; The track geometry deformation characteristics, temperature and humidity data, and vibration data are integrated to obtain the freeze-thaw cycle frequency matrix and axle load distribution matrix. Based on the crack data, a structural state model is constructed and crack extension data is calculated; Evaluate track smoothness based on track geometry and vibration data; evaluate track durability based on crack growth data; Build a 3D twin model based on synchronized data and dynamically correlate crack data, crack propagation data, and track geometry deformation characteristics; Continuously import new data into the 3D twin model, and optimize model parameters and adjust thresholds in real time to automatically update the model. Utilize the automatically updated model to output visual assessment results and generate maintenance strategies; Repairs are performed based on the maintenance strategy, and post-repair re-inspection data is uploaded to the automatic update model to update the model.

[0007] In a possible implementation, the fracture data includes the length, width, direction, and confidence level of the fracture.

[0008] In one possible implementation, the method for extracting track geometric deformation features from track point cloud data includes: (1) Data preprocessing Point cloud registration: Use the ICP algorithm to align the current scan point cloud with the design BIM model, with a registration error of <1mm; (2) Geometric deformation calculation (2.1) Orbital height deviation extraction: Reference surface fitting: perform least squares plane fitting on the designed rail surface to generate the theoretical rail surface equation; Deviation calculation: Calculate the vertical distance from each laser point cloud to the fitting plane, and take the maximum value as the height deviation of the track segment; (2.2) Extraction of tunnel lining displacement: Benchmark ring fitting: Select the point cloud data of the undeformed section of the tunnel and fit the benchmark cross section; Displacement mapping: Compare the current scan section with the reference ring and calculate the radial displacement; (3) Precision control By adjusting the number of ICP registration iterations or the point cloud sampling density, the calculation accuracy can be controlled within the range of ±1mm.

[0009] In one possible implementation, the method of fusing track geometric deformation characteristics, temperature and humidity data, and vibration data to obtain a freeze-thaw cycle frequency matrix and an axle load distribution matrix includes: 1. Construct a freeze-thaw cycle frequency matrix: Based on temperature and humidity data, periodically calculate the monthly or quarterly cumulative freeze-thaw times to generate a freeze-thaw cycle frequency matrix; 2. Construct an axle weight distribution load matrix: Calculate vibration energy based on vibration data and extract vibration frequency characteristics. Use vibration energy and frequency characteristics, combined with a vehicle model database, to match actual axle weights. Count the passing frequency of different axle weights over time to generate an axle weight distribution load matrix.

[0010] In one possible implementation, the calculation formula for the crack extension data is as follows:

[0011] Where a is the current length of the crack; N is the number of train passes; da / dN is the crack extension per load cycle; ΔK is the stress intensity factor amplitude, and the calculation formula is: ; Y is the geometric correction factor; s is the dynamic stress amplitude, ; E 0 is the initial elastic modulus of track concrete; N i is the passing frequency of the i-th axle load train in the axle load distribution matrix; W v,i is the reference vibration energy corresponding to the i-th type axle load; W v0 is the vibration energy of the reference axle load; W v is the vibration energy; A is the track cross-sectional area; C and m are the material constants of concrete; η is the material attenuation coupling coefficient;

[0012] E 0 is the initial elastic modulus of track concrete; E(t) : elastic modulus after t years; RH : annual average relative humidity; RH 0: initial annual average relative humidity; T : average annual temperature; T 0: reference temperature; α : basic decay rate; β : Humidity sensitivity coefficient; c : Temperature sensitivity coefficient; D: number of freeze-thaw cycles; d: freeze-thaw attenuation coefficient.

[0013] In one possible implementation, the track smoothness evaluation formula is as follows:

[0014] Where P is the smoothness index, represents the standard deviation of track geometry deformation, ; y i : No. i The deformation of the sampling points; : mean value of shape variable; N : the number of sampling points of the track segment; represents vibration energy, ; a k : instantaneous value of vibration acceleration at the kth sampling point; M : number of sampling points; Δt : sampling interval; m and n are weight coefficients; The durability evaluation formula is as follows:

[0015] Wherein, L is the durability index; is the crack growth rate, ; is the dth day; da / dN is the crack extension per load cycle; N d is the average number of trains passing per day; l is the material attenuation rate, calculated as: ; x 、 y Represents the weight coefficient.

[0016] In one possible implementation, the method for constructing a three-dimensional twin model based on synchronized data and dynamically associating crack data, crack propagation data, and track geometric deformation characteristics includes: Utilize 3D modeling tools to synchronize and dynamically correlate crack data, crack propagation data, and track geometry as input; Associating crack data, crack extension data, and track geometry with BIM model vertices or faces to achieve a one-to-one mapping between geometry and inspection data. Perform dynamic risk visualization rendering.

[0017] In one possible implementation, the method of continuously importing new data into the 3D twin model and performing real-time model parameter optimization and threshold adjustment to automatically update the model includes: Continuously import new data into the 3D twin model; The random forest model is used as a classifier for the 3D twin model to integrate multi-source data to output the service performance level of the infrastructure; Use a lightweight optimizer to update the weights of the random forest model for incremental learning; Based on sliding window statistics, dynamically calculate the mean and standard deviation and adjust the threshold range; Detect and process abnormal data for automatic update models.

[0018] In a possible implementation, the visual assessment result includes a risk heat map and a crack dynamic prediction curve; the risk heat map includes high-risk areas and warning areas; The maintenance strategy includes: Scenario 1: Trigger conditions: crack width > 0.25mm and deformation > 5mm, Vibration energy exceeds baseline by 50%; Risk level: High; Maintenance measures: Speed-restricted operation + emergency repair; Materials and / or equipment: Rapid-setting concrete, epoxy resin; Response time ≤ 24 hours; Priority: P0; Scenario 2: Trigger conditions: 0.2mm<crack width≤0.25mm and humidity>80%, Deformation 3-5mm; Risk level: Medium; Maintenance measures: Planned repair + temporary reinforcement; Materials and / or equipment: Carbon fiber cloth, anchor bolts; Response time ≤ 7 days; Priority: P1; Scenario 3: Triggering conditions: crack width ≤ 0.2mm but expansion rate > 0.01mm / day, Humidity >70% for 30 days; Risk level: Low; Maintenance measures: Enhanced monitoring + preventive maintenance; Materials and / or equipment: Anti-corrosion coatings, sealants; Response time ≤ 30 days; Priority: P2; Scenario 4: Trigger condition: deformation 1-3mm with no acceleration trend, The vibration energy fluctuation is within ±20% of the baseline; Risk level: Observation; Maintenance measures: Regular review + data tracking; Materials and / or equipment mentioned: None; Response time ≤ 90 days; Priority: P3.

[0019] Secondly, a device for automatically updating the service status of high-speed railway line infrastructure is provided, including: Multi-source data acquisition module: used to acquire and synchronize data, obtaining synchronized data including track point cloud data, track image data, temperature and humidity data, and vibration data; An identification module is used to identify cracks in the track image data and output crack data; Extraction module, used to extract track geometric deformation features from track point cloud data; The fusion module is used to fuse the track geometry deformation characteristics, temperature and humidity data, and vibration data to obtain the freeze-thaw cycle frequency matrix and the axle load distribution matrix; A calculation module is used to construct a structural state model based on crack data and calculate crack extension data; An evaluation module is used to evaluate track smoothness based on track geometry and vibration data, and to evaluate track durability based on crack growth data. A model building module is used to build a 3D twin model based on synchronized data and dynamically associate crack data, crack propagation data, and track geometry deformation characteristics; The update module is used to continuously import new data into the 3D twin model and optimize the model parameters and adjust the threshold in real time to obtain an automatically updated model; The output module is used to output visual evaluation results and generate maintenance strategies using the automatically updated model.

[0020] The feedback optimization module is used to perform maintenance based on the maintenance strategy, and upload the re-inspection data after maintenance to the automatic update model to update the model.

[0021] The beneficial effects of this application are as follows: (1) The method provided in this application adopts a multi-dimensional data dynamic fusion method, which realizes the crack-deformation-environment multi-parameter coupling analysis by fusing track point cloud data, track image data, sensor data and crack and extension data; (2) The method provided in this application dynamically updates the evaluation rules and thresholds based on the FTRL optimizer, adapts to material aging and sudden loads, and realizes incremental learning and automatic updating of model data. It can also further update the model in combination with the re-inspection data after maintenance to achieve data closed-loop feedback. (3) The method provided in this application uses a WebGL engine to render risk heat maps in real time, associate maintenance work orders with spare parts inventory, and visualize assessment results; (4) The method provided in this application is scalable across scenarios, supporting multiple types of facilities such as bridges, tunnels, and rails, and collaboratively constructing a global state knowledge base in the cloud; (5) The method provided in this application can significantly improve the efficiency of detecting the service status of high-speed railway line infrastructure. The model can be dynamically updated in real time, and the model accuracy is high, providing strong technical support for the intelligent operation and maintenance of high-speed railway infrastructure. (6) This application also provides a device based on the above method, which can also realize automatic updating and detection of the service status of high-speed railway line infrastructure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flowchart of a method for automatically updating the service status of high-speed railway line infrastructure provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of a device for automatically updating the service status of high-speed railway line infrastructure provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The content of this application is further described below with reference to specific embodiments, but the content of this application is not limited thereto.

[0024] The current methods for detecting the service status of high-speed railways have problems such as rigid indicators and data fragmentation, low accuracy in detecting the status of high-speed railway infrastructure, and low efficiency in operation and maintenance throughout the entire life cycle.

[0025] In view of this, the present application provides a method for automatically updating the service status of high-speed railway line infrastructure, comprising: S101: Acquire collected data and synchronize the data to obtain synchronized data including track point cloud data, track image data, temperature and humidity data, and vibration data.

[0026] In one possible implementation, the data synchronization method includes: (1) Time synchronization: All detection equipment has a built-in high-precision clock module, which performs millisecond-level time synchronization through GPS / Beidou satellite signals to ensure that the acquisition timestamps of laser, image, and vibration data are consistent.

[0027] The timestamp format is unified to Coordinated Universal Time (UTC), with a deviation of less than 0.1 seconds, to avoid cross-system data association errors.

[0028] (2) Spatial alignment: The laser scanning point cloud and the infrastructure BIM model are aligned through feature matching algorithms, for example, using inherent features such as track sleeper spacing and bridge bolt hole positions for spatial alignment.

[0029] After registration, the data is converted to a unified coordinate system (such as the line design coordinate system), and the spatial error is controlled within 3mm to ensure accurate mapping of parameters such as crack location and deformation.

[0030] It is understandable that the collected data is pre-processed to facilitate its application in subsequent data processing, including: Point cloud data denoising: Removes noise data (such as flying insects and raindrops) from the point cloud, preserving true structural surface information. Implementation: Based on the density distribution characteristics of the point cloud, it automatically identifies and filters discrete noise points, preserving the continuous surface point cloud. The clean point cloud data is output for deformation calculations (such as track elevation deviation and tunnel lining displacement).

[0031] Temperature, humidity, and vibration data filtering: This eliminates random noise from sensor signals and highlights valid features (such as wheel-rail impact vibration). Sliding average processing is performed on vibration signals to smooth high-frequency noise while retaining low-frequency structural response components. Transient outliers (such as transient equipment failures) are removed from temperature and humidity data to generate a continuous and smooth environmental change curve.

[0032] S102: Identify cracks in the track image data and output crack data.

[0033] In one possible implementation, the method of identifying cracks in track image data and outputting crack data includes: Automatically identify cracks and quantify geometric parameters from orbital image data.

[0034] For example, a pre-trained AI model (such as YOLOv8) is loaded to perform real-time inference on the image, select the crack area and calculate the actual physical size (camera parameters and shooting distance need to be calibrated), output the crack parameters (length, width, direction) and confidence level, and only retain detection results with a confidence level higher than 90%.

[0035] S103: Extracting track geometric deformation features from the track point cloud data.

[0036] In one possible implementation, the method of S103 includes: S103a, data preprocessing Point cloud registration: The ICP (Iterative Closest Point) algorithm is used to align the current scan point cloud with the design BIM model, with a registration error of <1mm.

[0037] S103b, Geometric deformation calculation (1) Track height deviation extraction: Reference surface fitting: perform least squares plane fitting on the designed track surface (BIM model) to generate the theoretical track surface equation; Deviation calculation: Calculate the vertical distance from each laser point cloud to the fitting plane, and take the maximum value as the height deviation of the track segment.

[0038] Calculation formula: ( i =1, 2, ..., N ) in: Axe + By + Cz + D =0: fitting plane equation, representing the theoretical design plane of the track, A, B, C and D are the coefficients obtained by fitting; ( x i , y i , z i ): track surface point cloud coordinates obtained by laser scanning; Δ h : Track height deviation (unit: mm), that is, the maximum vertical distance from the point cloud to the fitting plane.

[0039] (2) Extraction of tunnel lining displacement: Benchmark ring fitting: Select the point cloud data of the undeformed section of the tunnel and fit the benchmark cross section (circular or horseshoe shape); Displacement mapping: Compare the current scan section with the reference ring and calculate the radial displacement (e.g. radial displacement of a point = current radius - reference radius).

[0040] S103b, precision control By adjusting the number of ICP registration iterations (default 100) or the point cloud sampling density (5mm → 1mm), the calculation accuracy can be controlled within the range of ±1mm. For example, reducing the number of iterations to 50 relaxes the accuracy to ±2mm, but the calculation speed increases by 40%.

[0041] Exemplary: (1) Scenario: Scanning point cloud of the K123+500 section of a high-speed railway tunnel (1 million points).

[0042] (2) Processing flow: After point cloud denoising, 980,000 points are retained; ICP registration error 0.8mm; Calculated lining convergence displacement: left wall +1.2mm, right wall -0.9mm, total convergence 2.1mm.

[0043] (3) Result: The error is 0.1mm compared with the manual re-measurement result (2.0mm) of the total station, which meets the accuracy requirement of ±1mm.

[0044] S104: Fusing the track geometric deformation characteristics, temperature and humidity data, and vibration data to obtain a freeze-thaw cycle frequency matrix and an axle load distribution matrix.

[0045] In one possible implementation, S104 includes: S104a, constructing a freeze-thaw cycle frequency matrix: Based on the temperature and humidity data, periodically counting the monthly or quarterly cumulative freeze-thaw times to generate a freeze-thaw cycle frequency matrix.

[0046] For example, Data input: ① Temperature data: daily minimum / maximum temperature collected by temperature and humidity sensors; ② Humidity data: continuous monitoring value of relative humidity (RH).

[0047] Calculation rules: Single-day freeze-thaw determination: When the minimum temperature of the day is ≤0℃ and the maximum temperature is >0℃, and the average humidity is ≥75%, it is counted as one effective freeze-thaw cycle; periodic statistics: The number of freeze-thaw cycles is accumulated on a monthly / quarterly basis to generate a freeze-thaw cycle frequency matrix (for example, the number of freeze-thaw cycles per month in a tunnel lining area is D: 5 times).

[0048] It should be noted that the freeze-thaw cycle frequency matrix is ​​directly involved in the calculation of the concrete elastic modulus attenuation formula.

[0049] S104b. Construct an axle weight distribution load matrix: Calculate vibration energy based on vibration data and extract vibration frequency characteristics; use vibration energy and frequency characteristics, combined with the vehicle model database, to match actual axle weights; count the passing frequencies of different axle weights according to time periods to generate an axle weight distribution load matrix.

[0050] Exemplarily, the method for constructing the axle load distribution matrix is ​​as follows: Data input: ① Vibration signal: time domain waveform collected by track vibration sensor (sampling rate 1kHz); ② Train operation log: vehicle type, formation, and passing time.

[0051] Implementation steps: ① Feature extraction: Vibration energy integration: Calculate the train wheel-rail impact energy (energy integration in the frequency band 80-200Hz); Peak frequency analysis: FFT extracts the main frequency component (e.g., 25-ton axle load corresponds to 120Hz±5Hz); ② Axle load estimation: Based on vibration energy and frequency characteristics, combined with the vehicle model database (e.g., CR400AF axle weight 17 tons, CRH380A axle weight 16 tons), the actual axle load is matched; ③ Distribution matrix generation: Count the passing frequencies of trains with different axle loads (16t, 17t, 25t) on a daily / weekly basis to generate an axle load-frequency distribution matrix (for example, a certain track section has an average of 25t axle load trains passing through it every day). 50 times, N i : The passing frequency of the i-th axle load train in the axle load distribution matrix).

[0052] It should be noted that the axial load distribution matrix drives the vibration energy calculation and affects the dynamic stress amplitude in the crack growth prediction model. s .

[0053] S105. Based on the crack data, a structural state model is constructed to calculate crack extension data.

[0054] In one possible implementation, the calculation formula for the crack extension data is as follows:

[0055] in, a : Current length of the crack, mm; N is the number of train passes; da / dN : Crack expansion per load cycle (mm / time); C, m: Material constants (typical values ​​for concrete , m=3.5); or : Material attenuation coupling coefficient (0.8, experimental calibration value).

[0056] DK : Stress intensity factor amplitude (MPa· ), the calculation formula is ; Y : Geometric correction factor (1.12, for surface cracks); s : Dynamic stress amplitude (MPa), converted from vibration energy:

[0057] E 0 is the initial elastic modulus of track concrete; N i is the passing frequency of the i-th axle load train in the axle load distribution matrix; W v,i is the reference vibration energy corresponding to the i-th axle load (experimental calibration value); W v0 The vibration energy of the reference axle load (e.g. 16t, experimental calibration value); Wv It is vibration energy; ; a k : The instantaneous value of the vibration acceleration at the kth sampling point (unit: m / s²), that is, the amplitude of the vibration signal collected by the sensor at a discrete time point; M : Number of sampling points, when the sampling rate is 1kHz, M=1000 in 1 second; Δt : sampling interval (0.001 seconds); A is the cross-sectional area of ​​the track; Material performance model: Concrete elastic modulus attenuation formula: Calculate the annual attenuation rate based on humidity-temperature historical data;

[0058] E 0 is the initial elastic modulus of track concrete; E(t) : elastic modulus of track concrete after t years (initial value E0); RH : Annual average relative humidity (reference humidity RH0=60%); RH 0: initial annual average relative humidity; T : annual average temperature (reference temperature T0 = 20℃); α : basic attenuation rate (0.5% / year, experimental calibration value); β : Humidity sensitivity coefficient (0.2, humidity increases by 10% with a decay rate of +0.5%); c : Temperature sensitivity coefficient (-0.015 / °C, attenuation rate +1.5% for every 1°C increase in temperature).

[0059] D: Number of freeze-thaw cycles (If the monthly freeze-thaw cycle of a tunnel is D=8, the material attenuation rate will increase by an additional 0.3%×8=2.4%.) d: Freeze-thaw attenuation coefficient (experimental calibration value, such as 0.3% / time).

[0060] S106. Evaluate track smoothness based on track geometric deformation characteristics and vibration data; evaluate track durability based on crack expansion data.

[0061] In one possible implementation, the track smoothness evaluation formula is as follows:

[0062] Ride comfort index threshold setting (dynamically updated): ≤0.5 is excellent grade A, 0.5-0.7 is qualified grade B, and >0.7 is high-risk grade C.

[0063] Where P is the smoothness index, Indicates the standard deviation of track geometric deformation (unit: mm), ; y i : No. i The deformation variables of each sampling point (such as height deviation, gauge change); : mean value of shape variable; N : The number of sampling points in the track segment, such as the number of sampling points per 10-meter track segment (when the laser scanning density is 500 points / m², N=5000); represents vibration energy, ; a k : The instantaneous value of the vibration acceleration at the kth sampling point (unit: m / s²), that is, the amplitude of the vibration signal collected by the sensor at a discrete time point; M : Number of sampling points, when the sampling rate is 1kHz, M=1000 in 1 second; Δt : sampling interval (0.001 seconds); m and n are weight coefficients (default m=0.6, n=0.4), which are adjusted dynamically according to the line level.

[0064] In one possible implementation, the durability evaluation formula is as follows:

[0065] Wherein, L is the durability index; is the crack growth rate, ; is the dth day; da / dN is the crack extension per load cycle; N d is the average number of trains passing per day (e.g. 50 times / day); l is the material attenuation rate (unit: % / year), calculated using the following formula: ; x 、 y Represents the weight coefficient, (default x =0.7, y =0.3).

[0066] S107. Construct a three-dimensional twin model based on the synchronized data and dynamically associate the crack data, crack extension data, and track geometric deformation characteristics.

[0067] In one possible implementation, S107 includes: S107a, using a three-dimensional modeling tool with synchronized data and dynamically associated crack data, crack propagation data, and track geometry as input; S107b, associating the crack data, crack extension data, and track geometric deformation features with the BIM model vertices or faces to achieve a one-to-one mapping between geometry and detection data; S107c, perform dynamic risk visualization rendering.

[0068] An exemplary method for constructing a 3D twin model includes the following steps: (1) Data integration and model loading 1.1 Input data: Static data: Infrastructure BIM models (railways, bridges, tunnels, etc.), including initial geometry, material properties, and structural topology; Dynamic data: pre-processed deformation (laser scanning), crack parameters (image recognition), environmental loads (temperature, humidity, vibration energy).

[0069] 1.2 Modeling Tools: WebGL engine: Build lightweight 3D rendering scenes based on the Three.js framework; Data binding: Associate dynamic data with BIM model vertices / faces to achieve a one-to-one mapping of "geometry-detection data".

[0070] (2) Dynamic risk visualization rendering 2.1 Rendering rules: Color mapping: Red highlights areas with crack width > 0.25 mm and deformation > 5 mm, and yellow highlights areas where vibration energy exceeds the limit; Dynamic Labeling: Superimpose text labels on the model surface (e.g., "Crack 0.28mm, predicted to expand to 0.35mm in 30 days").

[0071] 2.2 Predictive rendering: The direction of crack expansion is indicated by an arrow symbol, and the length of the arrow is proportional to the predicted expansion amount.

[0072] S108. Continuously import new data into the three-dimensional twin model, and perform parameter optimization and threshold adjustment of the model in real time to obtain an automatically updated model.

[0073] In one possible implementation, S108 includes: S108a, continuously importing new data into the 3D twin model; S108b. Use the random forest model as a classifier for the 3D twin model to integrate multi-source data to output the service performance level of the infrastructure; S108c, using a lightweight optimizer to update the weights of the random forest model for incremental learning.

[0074] As can be understood, in S108c, the random forest model is used as the core classifier for state-performance evaluation, integrating multi-source data (deformation, cracks, and environment) to output the service performance level (safety / warning / high risk) of the infrastructure, as shown in Table 2.

[0075] Table 2 Related parameters of random forest

[0076] Furthermore, the incremental learning method in S108c is as follows: Step 1: The FTRL (Follow-the-Regularized-Leader) optimizer is used to update feature weights online. The formula is:

[0077] w i (t) : The weight of the i-th feature at time t (e.g. the initial value of the crack width weight is 0.7); g i (t) : gradient (calculate the derivative of the loss function based on new data); or : learning rate (default 0.01); l : L1 regularization coefficient (default 0.1); sgn(): Signum function.

[0078] Step 2: When the new data distribution deviation exceeds 20% (KL divergence test), full data retraining is triggered, updating the decision tree structure and splitting nodes. For example, if recent crack width data for a tunnel is concentrated between 0.2 and 0.25 mm (the original threshold was 0.3 mm), the model uses incremental learning to reduce the crack width weight (0.7 to 0.6) and add a temperature gradient feature weight (0.05 to 0.1).

[0079] S108d. Based on sliding window statistics, dynamically calculate the mean and standard deviation and adjust the threshold range; Exemplarily, the steps of S108d are as follows: Based on sliding window statistics (data distribution over the past 90 days), the mean and standard deviation are dynamically calculated, and the threshold range is adjusted (for example, deformation threshold = historical mean + 3 times the standard deviation). For example, if recent crack width data is concentrated between 0.2 and 0.25 mm (the original threshold is 0.3 mm), the threshold is automatically lowered to 0.25 mm. This is shown in Table 3.

[0080] Table 3 Threshold adjustment

[0081] Threshold adjustment formula:

[0082] μ 滑动窗口 : The mean value of the data within the sliding window (e.g., the mean crack width over the past 90 days is 0.22 mm); σ 滑动窗口 : Standard deviation within the sliding window (e.g. 0.02mm); k: Adjustment coefficient (default is 3, which can be adjusted dynamically according to the risk level).

[0083] For example: Input data: Crack width dataset for the past 90 days [0.20, 0.21, 0.23, 0.22, 0.24]; Calculate the statistic: μ 滑动窗口 =0.22mm, σ 滑动窗口 =0.015mm; Threshold adjustment: New threshold = 0.22 + 3 × 0.015 = 0.265 mm → Automatically adjusted down to 0.25 mm (rounded to 0.05 mm accuracy).

[0084] S108e. Detect and process abnormal data of the automatic update model.

[0085] Further, the steps of S108e are as follows: (1) Deviation check: If the new data deviates from the historical pattern by more than 20% (e.g., a sudden increase of 50% in humidity), it is marked as abnormal data; (2) Manual review: Abnormal data is pushed to the operation and maintenance platform, triggering drone re-inspection or on-site verification; (3) Model rollback: If the misjudgment rate suddenly increases by more than 10% after the update, it will automatically switch to the previous stable version.

[0086] S109: Utilize the automatic update model to output visual evaluation results and generate maintenance strategies.

[0087] In one possible implementation, the visual assessment result includes a risk heat map and a crack dynamic prediction curve; the risk heat map includes high-risk areas and warning areas.

[0088] Furthermore, the risk heat map marks areas of different risks by color.

[0089] For example, in the risk heat map, high-risk areas (cracks > 0.25 mm, deformation > 5 mm) are highlighted in red; warning areas (vibration energy exceeds the baseline by 20%, humidity > 80%) are marked in yellow.

[0090] Furthermore, the crack dynamic prediction curve shows the changing trends of key indicators over the next 30 days. Key indicators include: crack parameters, deformation parameters, material properties, and environmental loads.

[0091] For example, the crack dynamic prediction curve is shown in Table 4.

[0092] Table 4 Related indicators of crack dynamic prediction curve

[0093] In one possible implementation, the maintenance strategy includes: Scenario 1: Trigger conditions: crack width > 0.25mm and deformation > 5mm, Vibration energy exceeds baseline by 50%; Risk level: High; Maintenance measures: Speed-restricted operation + emergency repair; Materials and / or equipment: Rapid-setting concrete, epoxy resin; Response time ≤ 24 hours; Priority: P0; Scenario 2: Trigger conditions: 0.2mm<crack width≤0.25mm and humidity>80%, Deformation 3-5mm; Risk level: Medium; Maintenance measures: Planned repair + temporary reinforcement; Materials and / or equipment: Carbon fiber cloth, anchor bolts; Response time ≤ 7 days; Priority: P1; Scenario 3: Triggering conditions: crack width ≤ 0.2mm but expansion rate > 0.01mm / day, Humidity >70% for 30 days; Risk level: Low; Maintenance measures: Enhanced monitoring + preventive maintenance; Materials and / or equipment: Anti-corrosion coatings, sealants; Response time ≤ 30 days; Priority: P2; Scenario 4: Trigger condition: deformation 1-3mm with no acceleration trend, The vibration energy fluctuation is within ±20% of the baseline; Risk level: Observation; Maintenance measures: Regular review + data tracking; Materials and / or equipment mentioned: None; Response time ≤ 90 days; Priority: P3.

[0094] See Table 5 for details.

[0095] Table 5 Maintenance strategy

[0096] S110 , performing maintenance based on the maintenance strategy, and uploading post-maintenance re-inspection data to the automatic update model to update the model.

[0097] In a possible implementation, the method of uploading the post-repair review data to the automatic update model to update the model includes: (1) Use the re-inspection data to update the data in the 3D twin model to ensure the timeliness of the data.

[0098] (2) Update model weights.

[0099] It should be noted that the model weight is updated using the method described in S108.

[0100] The content of this application is further described below with reference to more specific embodiments.

[0101] Take tunnel lining crack detection and maintenance as an example: 1. Data collection: The laser scanner detects lining cracks with a length of 2.1m, the industrial camera identifies a width of 0.28mm, and the temperature and humidity sensor monitors humidity at 85%.

[0102] 2. Time and space synchronization: GPS timing aligns data timestamps, and the ICP algorithm maps the crack location to the BIM model (error <3mm).

[0103] 3. Edge preprocessing: After denoising the laser point cloud, a deformation of 0.8 mm was extracted, and the YOLOv8 model confirmed the crack with a confidence level of 92%.

[0104] 4. Fusion modeling: Correlate crack width with humidity data and predict that the crack will expand to 0.35mm after 30 days, determining a durability grade of C.

[0105] 5. Dynamic update: Based on recent similar data, the model automatically lowers the crack width threshold from 0.3mm to 0.25mm.

[0106] The following describes the device for automatically updating the service status of high-speed railway line infrastructure provided by the present invention. The device for automatically updating the service status of high-speed railway line infrastructure described below and the method for automatically updating the service status of high-speed railway line infrastructure described above can be referenced to each other.

[0107] Figure 2 Schematic diagram of the structure of the device for automatically updating the service status of high-speed railway line infrastructure provided by an embodiment of the present invention. Figure 2 As shown, it includes: a multi-source data acquisition module 201, an identification module 202, an extraction module 203, a fusion module 204, a calculation module 205, an evaluation module 206, a model building module 207, an update module 208, an output module 209 and a feedback optimization module 210, wherein: The multi-source data acquisition module 201 is used to acquire and synchronize data to obtain synchronized data including track point cloud data, track image data, temperature and humidity data, and vibration data; Identification module 202, for identifying cracks in track image data and outputting crack data; Extraction module 203, used to extract track geometric deformation features from track point cloud data; Fusion module 204 is used to fuse the track geometry deformation characteristics, temperature and humidity data, and vibration data to obtain the freeze-thaw cycle frequency matrix and the axle load distribution matrix; The calculation module 205 is used to construct a structural state model based on the crack data and calculate the crack extension data; Evaluation module 206, for evaluating track smoothness based on track geometric deformation characteristics and vibration data; and evaluating track durability based on crack growth data; A model building module 207 is used to build a three-dimensional twin model based on the synchronized data and dynamically associate the crack data, crack propagation data and track geometric deformation characteristics; The update module 208 is used to continuously import new data into the 3D twin model and perform parameter optimization and threshold adjustment of the model in real time to obtain an automatically updated model; The output module 209 is used to output the visual evaluation results and generate the maintenance strategy using the automatic update model.

[0108] The feedback optimization module 210 is used to perform maintenance based on the maintenance strategy, and upload the re-inspection data after maintenance to the automatic update model to update the model.

[0109] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0110] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for automatically updating the service status of high-speed railway line infrastructure, characterized in that: include: Acquire collected data and synchronize it to obtain synchronized data including track point cloud data, track image data, temperature and humidity data, and vibration data; Identify cracks in track image data and output crack data; Extract track geometric deformation features from track point cloud data; The track geometry deformation characteristics, temperature and humidity data, and vibration data are integrated to obtain the freeze-thaw cycle frequency matrix and axle load distribution matrix. Based on the crack data, a structural state model is constructed and crack extension data is calculated; Evaluate track smoothness based on track geometry and vibration data; Evaluate track durability based on crack growth data; Build a 3D twin model based on synchronized data and dynamically correlate crack data, crack propagation data, and track geometry deformation characteristics; Continuously import new data into the 3D twin model, and optimize model parameters and adjust thresholds in real time to automatically update the model. Utilize the automatically updated model to output visual assessment results and generate maintenance strategies; Repairs are performed based on the maintenance strategy, and post-repair re-inspection data is uploaded to the automatic update model to update the model.

2. The method for automatically updating the service status of high-speed railway line infrastructure according to claim 1, characterized in that: The fracture data includes the length, width, direction and confidence of the fracture.

3. The method for automatically updating the service status of high-speed railway line infrastructure according to claim 1, characterized in that: The method for extracting track geometric deformation features from track point cloud data comprises: (1) Data preprocessing Point cloud registration: Use the ICP algorithm to align the current scan point cloud with the design BIM model, with a registration error of <1mm; (2) Geometric deformation calculation (2.1) Orbital height deviation extraction: Reference surface fitting: perform least squares plane fitting on the designed rail surface to generate the theoretical rail surface equation; Deviation calculation: Calculate the vertical distance from each laser point cloud to the fitting plane, and take the maximum value as the height deviation of the track segment; (2.2) Extraction of tunnel lining displacement: Benchmark ring fitting: Select the point cloud data of the undeformed section of the tunnel and fit the benchmark cross section; Displacement mapping: Compare the current scan section with the reference ring and calculate the radial displacement; (3) Precision control By adjusting the number of ICP registration iterations or the point cloud sampling density, the calculation accuracy can be controlled within the range of ±1mm.

4. The method for automatically updating the service status of high-speed railway line infrastructure according to claim 1, characterized in that: The method of fusing track geometric deformation characteristics, temperature and humidity data, and vibration data to obtain a freeze-thaw cycle frequency matrix and an axle load distribution matrix includes:

1. Construct a freeze-thaw cycle frequency matrix: Based on temperature and humidity data, periodically calculate the monthly or quarterly cumulative freeze-thaw times to generate a freeze-thaw cycle frequency matrix; 2. Construct an axle weight distribution load matrix: Calculate vibration energy based on vibration data and extract vibration frequency characteristics. Use vibration energy and frequency characteristics, combined with a vehicle model database, to match actual axle weights. Count the passing frequency of different axle weights over time to generate an axle weight distribution load matrix.

5. The method for automatically updating the service status of high-speed railway line infrastructure according to claim 1, characterized in that: The calculation formula of the crack extension data is as follows: Where a is the current length of the crack; N is the number of train passes; da / dN is the crack extension per load cycle; ΔK is the stress intensity factor amplitude, and the calculation formula is: ; Y is the geometric correction factor; σ is the dynamic stress amplitude, ; E 0 is the initial elastic modulus of track concrete; N i is the passing frequency of the i-th axle load train in the axle load distribution matrix; W v,i is the reference vibration energy corresponding to the i-th type axle load; W v0 is the vibration energy of the reference axle load; W v is the vibration energy; A is the track cross-sectional area; C and m are the material constants of concrete; η is the material attenuation coupling coefficient; E 0 is the initial elastic modulus of track concrete; E(t) : elastic modulus after t years; RH : annual average relative humidity; RH 0: initial annual average relative humidity; T : average annual temperature; T 0: reference temperature; α : basic decay rate; β : Humidity sensitivity coefficient; γ : Temperature sensitivity coefficient; D: number of freeze-thaw cycles; d: freeze-thaw attenuation coefficient.

6. The method for automatically updating the service status of high-speed railway line infrastructure according to claim 1, characterized in that: The track smoothness evaluation formula is as follows: Where P is the smoothness index, represents the standard deviation of track geometry deformation, ; y i : No. i The deformation of the sampling points; : mean value of shape variable; N : the number of sampling points of the track segment; represents vibration energy, ; a k : instantaneous value of vibration acceleration at the kth sampling point; M : number of sampling points; Δt : sampling interval; m and n are weight coefficients; The durability evaluation formula is as follows: Wherein, L is the durability index; is the crack growth rate, ; is the dth day; da / dN is the crack extension per load cycle; N d is the average number of trains passing per day; λ is the material attenuation rate, calculated as: ; x 、 y Represents the weight coefficient.

7. The method for automatically updating the service status of high-speed railway line infrastructure according to claim 1, characterized in that: The method for constructing a three-dimensional twin model based on synchronized data and dynamically associating crack data, crack extension data, and track geometric deformation characteristics includes: Utilize 3D modeling tools to synchronize and dynamically correlate crack data, crack propagation data, and track geometry as input; Associating crack data, crack extension data, and track geometry with BIM model vertices or faces to achieve a one-to-one mapping between geometry and inspection data. Perform dynamic risk visualization rendering.

8. The method for automatically updating the service status of high-speed railway line infrastructure according to claim 1, characterized in that: The method of continuously importing new data into the 3D twin model and performing real-time model parameter optimization and threshold adjustment to automatically update the model includes: Continuously import new data into the 3D twin model; The random forest model is used as a classifier for the 3D twin model to integrate multi-source data to output the service performance level of the infrastructure; Use a lightweight optimizer to update the weights of the random forest model for incremental learning; Based on sliding window statistics, dynamically calculate the mean and standard deviation and adjust the threshold range; Detect and process abnormal data for automatic update models.

9. The method for automatically updating the service status of high-speed railway line infrastructure according to claim 1, characterized in that: The visual assessment results include a risk heat map and a crack dynamic prediction curve; the risk heat map includes high-risk areas and early warning areas; The maintenance strategy includes: Scenario 1: Trigger conditions: crack width > 0.25mm and deformation > 5mm, Vibration energy exceeds baseline by 50%; Risk level: High; Maintenance measures: Speed ​​limit operation + emergency repair; Materials and / or equipment: Quick-setting concrete, epoxy resin; Response time ≤ 24 hours; Priority: P0; Scenario 2: Trigger conditions: 0.2mm<crack width≤0.25mm and humidity>80%, Deformation 3-5mm; Risk level: Medium; Maintenance measures: Planned repair + temporary reinforcement; Materials and / or equipment: Carbon fiber cloth, anchor bolts; Response time ≤ 7 days; Priority: P1; Scenario 3: Triggering conditions: crack width ≤ 0.2mm but expansion rate > 0.01mm / day, Humidity >70% for 30 days; Risk level: Low risk; Maintenance measures: Enhanced monitoring + preventive maintenance; Materials and / or equipment: Anti-corrosion coatings, sealants; Response time ≤ 30 days; Priority: P2; Scenario 4: Trigger condition: deformation 1-3mm with no acceleration trend, The vibration energy fluctuation is within ±20% of the baseline; Risk level: Observation; Maintenance measures: Regular review + data tracking; Materials and / or equipment mentioned: None; Response time ≤ 90 days; Priority: P3.

10. A device for automatically updating the service status of high-speed railway line infrastructure, characterized by comprising: Multi-source data acquisition module: used to acquire and synchronize data, obtaining synchronized data including track point cloud data, track image data, temperature and humidity data, and vibration data; An identification module is used to identify cracks in the track image data and output crack data; Extraction module, used to extract track geometric deformation features from track point cloud data; The fusion module is used to fuse the track geometry deformation characteristics, temperature and humidity data, and vibration data to obtain the freeze-thaw cycle frequency matrix and the axle load distribution matrix; A calculation module is used to construct a structural state model based on crack data and calculate crack extension data; An evaluation module is used to evaluate track smoothness based on track geometry and vibration data, and to evaluate track durability based on crack growth data. A model building module is used to build a 3D twin model based on synchronized data and dynamically associate crack data, crack propagation data, and track geometry deformation characteristics; The update module is used to continuously import new data into the 3D twin model and optimize the model parameters and adjust the threshold in real time to obtain an automatically updated model; The output module is used to output visual evaluation results and generate maintenance strategies using the automatically updated model. The feedback optimization module is used to perform maintenance based on the maintenance strategy, and upload the re-inspection data after maintenance to the automatic update model to update the model.

Images