Method, system and equipment for monitoring health state of infrastructure along high-speed rail and medium

Abstract

The invention discloses a high-speed rail infrastructure health state monitoring method, system and device and a medium, and relates to the technical field of high-speed rail infrastructure monitoring. The method comprises the following steps: constructing a monitoring framework based on a monitoring category; arranging a data acquisition scheme based on the monitoring framework, and obtaining monitoring data; predicting the health state of the infrastructure along the high-speed rail according to the monitoring data and the state monitoring model, and determining a risk prediction result; the risk prediction result comprises a potential risk area and a corresponding type; the state monitoring model comprises a prediction model and a key area monitoring model based on a prediction result. The method can improve the precision of an abnormal state prediction scheme, and improves the data transmission and processing efficiency. And meanwhile, flaw detection train monitoring items are expanded, and the monitoring precision is enhanced, so that the health state of the whole infrastructure along the high-speed railway is monitored.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of high-speed rail infrastructure monitoring, in particular to a high-speed rail infrastructure health state monitoring method, system, device and medium. BACKGROUND

[0002] As of the end of 2023, the operating mileage of China's railways reached 159,000 kilometers, of which high-speed railways accounted for 45,000 kilometers. With the continuous acceleration of high-speed railway construction and the continuous improvement of operating mileage, the monitoring level of high-speed rail infrastructure is also constantly innovating and improving. Currently, there are three ways to monitor the health of high-speed rail infrastructure: manual periodic inspection, sensor fixed-point monitoring, and defect detection train inspection. Manual periodic inspection can flexibly adjust the detection range and focus according to the on-site situation, but the labor and time costs are high; sensor fixed-point monitoring reduces the investment in manual inspection and has real-time monitoring capabilities, but it is mostly limited to monitoring a single bridge, tunnel, roadbed, or a part of the track, and the overall line health monitoring scheme is relatively insufficient; defect detection train inspection has a wide coverage and high monitoring accuracy, which helps to monitor the overall high-speed rail infrastructure, but frequent scheduling of defect detection trains increases the line scheduling pressure and occupies a large amount of window time, in addition, the monitoring items of the existing defect detection train are still not comprehensive enough to achieve overall line monitoring. How to reasonably schedule the defect detection train and add more monitoring items is an important problem that needs to be solved. To achieve a scientific scheduling scheme for the defect detection train and expand the monitoring items, the main difficulties lie in the following aspects:

[0003] (1) Lack of high-precision abnormal state prediction scheme. Current abnormal state prediction models are mostly based on a single data source, lacking the use of multi-source data for infrastructure overrun prediction monitoring, relying solely on a single data source to determine the overrun state, making the prediction accuracy of existing models insufficient; (2) Lack of overall high-speed rail infrastructure health monitoring scheme. Due to the current focus on individual areas or single indicators in the monitoring of infrastructure state overrun, there is a lack of referenceable scheme for adding more monitoring items to the defect detection train; (3) Lack of efficient processing and transmission of monitoring data. The load impact of high-speed train operation and the network conditions in remote areas limit the processing and transmission of data, making it challenging to process and transmit data. SUMMARY

[0004] The present application relates to the technical field of high-speed rail infrastructure monitoring, in particular to a high-speed rail infrastructure health state monitoring method, system, device and medium.

[0005] To achieve the above-mentioned purpose, the present application provides the following scheme:

[0006] A method for monitoring the health status of infrastructure along a high-speed railway line, comprising:

[0007] Establish a monitoring framework consisting of four monitoring categories; the monitoring categories include structural health monitoring of infrastructure along the high-speed railway line, roadbed condition monitoring, track condition monitoring, and external environment and geological disaster monitoring along the high-speed railway line;

[0008] Arrange a data collection plan based on the monitoring framework and obtain monitoring data;

[0009] The health status of infrastructure along the high-speed railway is predicted based on monitoring data and a status monitoring model to determine risk prediction results; the risk prediction results include potential risk areas and corresponding types; the status monitoring model includes a prediction model and a key area monitoring model based on the prediction results.

[0010] Optionally, the construction of a monitoring framework consisting of four monitoring categories includes:

[0011] Constructing a functional layer for structural health monitoring of infrastructure along the high-speed railway based on the infrastructure's underlying state data and structural degradation data; the underlying state data includes deformation data, crack data, and corrosion data; and the structural degradation data includes geological changes, temperature changes, and train vibrations.

[0012] Construct a functional layer for roadbed condition monitoring based on the left-right and vertical displacement and deformation of the high-speed railway roadbed;

[0013] Constructing a functional layer for track condition monitoring based on track-related data; the relevant data is used to reflect the smoothness and surface damage of the track; the relevant data includes track geometry data, crack data, and wear data;

[0014] A functional layer for monitoring the external environment and geological disasters along the high-speed railway is constructed based on environmental geological data; the environmental geological data includes extreme weather, landslide data, mud-rock flow data and earthquake data.

[0015] Optionally, arranging a data collection scheme based on the monitoring framework specifically includes:

[0016] Deploy sensors along the high-speed rail line, leaving the fixed-point sensors on the original line unchanged, and deploy perception sensors on the flaw detection train;

[0017] Based on the deployed sensors, monitoring data is collected and processed using the adaptive decomposition algorithm and SVM model to obtain the monitoring data to be transmitted;

[0018] Data transmission is carried out using a multi-level transmission method that combines 5G communication network and satellite communication.

[0019] Optionally, the fixed-point sensor includes a temperature sensor, a humidity sensor, a wind speed sensor and a GNSS receiving device; the perception sensor includes a lidar, a depth camera, an ultrasonic sensor, a vibration sensor and a mobile GNSS device.

[0020] Optionally, before predicting the health status of infrastructure along the high-speed rail line based on the monitoring data and the condition monitoring model, the method further includes: building and training the condition monitoring model, with the specific steps being:

[0021] Constructing a pre-trained network based on an LSTM network and a CNN network, optimizing initial parameters of the pre-trained network using historical data, and determining the optimized network as a prediction model; the historical data includes external environmental data and infrastructure overrun data;

[0022] The prediction model is used to make preliminary predictions, identify preliminary key areas, and build a key area monitoring model based on the in-depth inspection data of the preliminary key areas. Finally, the key monitoring data output by the key area monitoring model is fed back into the prediction model for dynamic optimization and iteration to obtain the final status monitoring model.

[0023] The present invention also provides a health status monitoring system for infrastructure along a high-speed railway, comprising:

[0024] A functional construction unit for constructing a monitoring framework consisting of four monitoring categories, including structural health monitoring of infrastructure along the high-speed railway, roadbed condition monitoring, track condition monitoring, and external environment and geological disaster monitoring along the high-speed railway;

[0025] A data monitoring unit, configured to arrange a data collection plan based on the monitoring framework and obtain monitoring data;

[0026] The status assessment unit is used to predict the health status of infrastructure along the high-speed railway based on monitoring data and a status monitoring model, and determine risk prediction results; the risk prediction results include potential risk areas and corresponding types; the status monitoring model includes a prediction model and a key area monitoring model based on the prediction results.

[0027] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned method for monitoring the health status of infrastructure along the high-speed railway.

[0028] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned method for monitoring the health status of infrastructure along a high-speed railway.

[0029] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0030] The present invention discloses a method, system, equipment and medium for monitoring the health status of infrastructure along a high-speed railway line. The method includes constructing a monitoring framework based on monitoring categories; wherein the monitoring categories include structural health monitoring of infrastructure along the high-speed railway line, roadbed condition monitoring, track condition monitoring and external environment and geological disaster monitoring along the high-speed railway line; arranging a data acquisition scheme based on the monitoring framework and acquiring monitoring data; predicting the health status of infrastructure along the high-speed railway line based on the monitoring data and the condition monitoring model, and determining the risk prediction results; the risk prediction results include potential risk areas and corresponding types; the condition monitoring model includes a prediction model and a key area monitoring model based on the prediction results. The present invention can improve the accuracy of the abnormal state prediction scheme and enhance the efficiency of data transmission and processing; at the same time, it expands the flaw detection train monitoring project and enhances the monitoring accuracy, thereby realizing the monitoring of the overall health status of infrastructure along the high-speed railway line. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in 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 paying any creative work.

[0032] Figure 1 This is the overall framework diagram of the health monitoring method for infrastructure along the high-speed railway in this embodiment;

[0033] Figure 2 This is the monitoring category framework in this embodiment;

[0034] Figure 3 This is a schematic diagram of the sensor layout in this embodiment;

[0035] Figure 4 This is the data transmission scheme in this embodiment;

[0036] Figure 5 This is the data preprocessing solution in this embodiment;

[0037] Figure 6 This is the infrastructure status monitoring model along the high-speed railway in this embodiment;

[0038] Figure 7 This is the prediction model framework in this embodiment;

[0039] Figure 8 This is the key area monitoring framework based on the prediction results in this embodiment. DETAILED DESCRIPTION

[0040] 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.

[0041] The purpose of the present invention is to provide a method, system, equipment and medium for monitoring the health status of infrastructure along a high-speed railway, which can monitor the health status of the entire infrastructure along the high-speed railway.

[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] The present invention provides a method for monitoring the health status of infrastructure along a high-speed railway, comprising:

[0044] Step 100: Construct a monitoring framework consisting of four monitoring categories; wherein the monitoring categories include structural health monitoring of infrastructure along the high-speed railway, roadbed condition monitoring, track condition monitoring, and external environment and geological disaster monitoring along the high-speed railway.

[0045] Step 200: Arrange a data collection plan based on the monitoring framework and obtain monitoring data.

[0046] Step 300: Predict the health status of infrastructure along the high-speed railway based on monitoring data and a status monitoring model to determine risk prediction results; the risk prediction results include potential risk areas and corresponding types; the status monitoring model includes a prediction model and a key area monitoring model based on the prediction results.

[0047] As a specific implementation method, Figures 1-8As shown in the figure, this solution combines external environmental data, geological conditions, and train operation information, employing a combination of prediction and focused monitoring of abnormal areas to monitor the overall health of infrastructure along high-speed rail lines. This solution first utilizes historical environmental data and infrastructure overrun data to construct an infrastructure health prediction model. This model is then trained to establish a relationship between the external environment and facility overrun status. Environmental data is collected in real time using various fixed-point sensors along the high-speed rail line. This data is transmitted to the edge or cloud via 5G or Beidou communications. After processing, it is input into the prediction model to predict the health of the infrastructure and promptly identify potential overrun risks. If a facility overrun risk is predicted, this information is fed back to the dispatching center. The center then modifies the operation plan based on the abnormal risk and dispatches flaw detection trains equipped with various sensors to conduct inspections in specific areas and locations. The prediction model is then refined and optimized using this data feedback, thereby improving prediction accuracy.

[0048] Through a closed-loop process of "model training - data collection - state prediction - key monitoring - data recording - model iteration," this solution establishes a mechanism for continuous monitoring and dynamic optimization of the health status of infrastructure along high-speed rail lines, ensuring infrastructure safety and operational stability. The technical framework of this invention is divided into three major components: a monitoring category layer, a data acquisition layer, and a state monitoring layer. These layers are responsible for the classification and monitoring of monitored objects, the collection of environmental data, and the application and iterative updating of prediction and detection models, respectively. This comprehensively improves the scientific nature and accuracy of infrastructure monitoring along high-speed rail lines.

[0049] The overall framework of this method is as follows Figure 1 As shown. The overall framework of the present invention consists of three core parts: monitoring category layer, data perception layer, and status monitoring layer. The monitoring category layer divides the monitoring content of infrastructure into four categories: structural health monitoring, roadbed status monitoring, track status monitoring, and external environment and geological disaster monitoring. According to different monitoring objects and needs, corresponding technical means are used for real-time evaluation. The data perception layer includes three links: sensor deployment, data processing, and data transmission. It collects and processes environmental data along the high-speed rail in real time, such as temperature and humidity, wind speed, precipitation, etc., as well as infrastructure over-limit data, to ensure the accuracy and completeness of the monitoring data. The status monitoring layer combines prediction and key monitoring models, trains historical data through the LSTM network, realizes the prediction of infrastructure status and the identification of potential risks, and cooperates with regular inspections and verifications to form a closed-loop monitoring and early warning mechanism, thereby ensuring the safety and stable operation of high-speed rail infrastructure. The specific steps are as follows:

[0050] S100 monitoring category layer. According to the high-speed railway operation safety monitoring content in the "High-speed Railway Infrastructure Operation Status Detection Management Measures", the monitoring objects and contents of the infrastructure along the high-speed railway are divided. The present invention constructs a systematic monitoring framework. It mainly monitors the following four types of objects. Specifically: Figure 2 As shown, S101 infrastructure structural health monitoring along the high-speed railway; S102 roadbed condition monitoring; S103 track condition monitoring; S104 external environment and geological disaster monitoring along the high-speed railway.

[0051] Structural health monitoring of infrastructure along the S101 line focuses on monitoring the deterioration and damage of high-speed rail infrastructure caused by external factors during its service life. This monitoring focuses on deformation, displacement, cracks, and excessive stress in bridges, tunnels, and other facilities. Over time, the structural bearing capacity may decrease, leading to cracks, displacement, and other defects, seriously impacting the safety and service life of the facilities. Continuous monitoring of these potential issues can promptly identify health risks and ensure the safety and stability of high-speed rail operations.

[0052] S102 Roadbed Condition Monitoring. As a critical component of high-speed rail infrastructure, roadbed monitoring focuses on assessing its morphology, stability, and bearing capacity to ensure the safety of the overall structure. Key areas include monitoring roadbed deformation, specifically focusing on the stability of the foundation and soil, analyzing whether there are risks of slippage or settlement, and monitoring roadbed cracks to promptly detect their initiation and expansion, thereby identifying potential damage risks. These monitoring measures play a vital role in ensuring the durability and stability of high-speed rail roadbed structures.

[0053] S103 Track Condition Monitoring. Regarding the safety and stability of tracks, it mainly covers five monitoring aspects. Track geometry parameter monitoring: ensuring that the gauge, displacement, etc. meet the standards to avoid track deviation affecting train safety; smoothness monitoring detects track fluctuations, reduces vibration and wear, and improves comfort; structural integrity monitoring: through technologies such as ultrasonic and laser scanning, monitors cracks and damage in the track structure to ensure stability; damage monitoring: real-time monitoring of cracks, corrosion and other problems to prevent failures; stress monitoring: assessing the stress changes of the track under load to prevent deformation and fatigue damage. Through the comprehensive application of these five monitoring aspects, the track condition can be comprehensively assessed to ensure the safety of train operation.

[0054] External environmental and geological hazard monitoring along the S104 high-speed railway line. This monitoring focuses on comprehensively assessing external factors affecting high-speed train operations and the risks of natural disasters. External environmental monitoring includes real-time monitoring of natural conditions such as rainfall, snowfall, strong winds, and water levels, which directly impact the safety of high-speed train operations. Geological hazard monitoring focuses on early warning and monitoring of hazards such as landslides, collapses, mudslides, and earthquakes. By identifying potential risks early, it provides timely and effective safety assurance for high-speed railway operations.

[0055] S200 data perception framework: Based on the monitoring category framework of S100, this paper designs a data perception framework suitable for monitoring the health status of infrastructure along high-speed railways. Specifically:

[0056] S201 sensor layout plan, such as Figure 3 As shown, the layout scheme of the present invention mainly arranges sensors from two levels: one is to install sensors at fixed points along the high-speed railway line, and the other is to arrange sensors on the flaw detection train.

[0057] (1) Install sensors at fixed points along the high-speed rail line

[0058] To achieve comprehensive monitoring of infrastructure along high-speed rail lines, this solution leverages sensors already installed along existing lines. These sensors are broadly divided into two types: environmental monitoring sensors and condition monitoring sensors. Environmental monitoring sensors include sensors for temperature, humidity, wind speed, and wind pressure, while condition monitoring sensors include vibration sensors, displacement sensors, and GNSS receivers. These sensors collect real-time environmental and facility status data, providing fundamental support for facility health. These sensors are also relatively easy to install on new lines, effectively reducing construction complexity and installation risks.

[0059] (2) Installing sensors on the flaw detection train

[0060] In order to improve the flexibility and coverage of monitoring while reducing the deployment requirements of fixed-point sensors. The present invention adds some new sensors to the existing flaw detection train, such as carrying laser radar, depth camera, vibration sensor and mobile GNSS receiver and other equipment, so that the flaw detection train can conduct dynamic detection of infrastructure along the high-speed rail while inspecting the tracks. Lidar can be used to build a three-dimensional model of the facilities along the line, depth camera and ultrasonic sensor can help detect cracks and damage of facilities such as bridges and tunnels, vibration sensor monitors the coupled vibration response between trains, tracks and infrastructure, GNSS receiver tracks the spatial displacement state of the train and evaluates the roadbed settlement. Through the deployment of sensors on the flaw detection train, efficient inspection of facilities along the line can be achieved, especially those areas that are difficult to monitor at the location of fixed-point sensors, providing more comprehensive monitoring data and support.

[0061] S202 Data transmission scheme. Figure 4 As shown in the figure, this solution aims to ensure efficient and stable transmission of monitoring data along the high-speed rail line, using 5G and Beidou satellite communication technologies to guarantee data reliability and real-time performance. Data collected by fixed-point sensors along the line is transmitted to the monitoring center via optical fiber or wireless networks. After data exchange and processing, it is ultimately delivered to the terminal. Data acquired by sensors on the flaw detection train is transmitted in real time to the ground receiving terminal via the onboard communication system, and then to the terminal via the data exchange platform. To address the strong electromagnetic interference and signal attenuation in the high-speed rail environment, the system implements data caching and redundant transmission mechanisms to ensure data integrity and accuracy during transmission. This ensures that both fixed-point data and train data can be transmitted to the terminal stably and in real time, supporting subsequent analysis and prediction.

[0062] S203 Data processing scheme. The present invention mainly processes the monitored time-frequency domain signal data, which mainly includes two parts: environmental noise removal and abnormal data removal to ensure the accuracy and integrity of the monitoring data. The specific scheme is as follows Figure 5 shown.

[0063] (1) Environmental noise removal

[0064] The time-frequency domain signal data monitored by this invention includes two categories: environmental data collected by sensors, such as temperature, humidity, and wind speed, and structural health monitoring data, such as displacement, vibration, and stress. Common denoising methods for these data include EMD denoising and wavelet threshold denoising. The former removes high-frequency noise through decomposition but can easily lead to signal distortion. The latter, while offering good time-frequency localization, is difficult to determine optimal parameters. Therefore, this paper utilizes a denoising method that combines EMD with wavelet thresholding to more effectively remove noise while preserving useful signals.

[0065] This environmental noise removal method consists of four parts: EMD decomposition, wavelet decomposition, wavelet reconstruction and EMD reconstruction. In the EMD decomposition part, the original signal is decomposed into a series of intrinsic mode function (IMF) components with different frequencies and a residual component. The decomposition formula of the time-frequency domain signal data is:

[0066]

[0067] Where, X r (t) is different time-frequency domain signal data, p i (t) is the i-th IMF component, q n (t) is the residual, and n is the number of IMF components.

[0068] In the wavelet decomposition part, the IMF components at different frequencies obtained by EMD decomposition are subjected to wavelet decomposition to decompose the signal into low-frequency components and high-frequency components. Because noise mainly exists in high-frequency components, a threshold is set for high-frequency components, and the coefficients with absolute values ​​less than the threshold are set to zero. These coefficients are considered to correspond to noise, and the coefficients greater than the threshold are retained. The high-frequency components after wavelet decomposition and denoising are reconstructed, and the reconstruction formula is:

[0069]

[0070] Where, X r '(t) is the reconstructed high frequency component, α J (k) is the low-frequency approximation coefficient of the Jth layer, d j (k) is the low-frequency approximation coefficient of the Jth layer, ψ j,k (t) is the wavelet function of the jth layer, φ J,k (t) is the scaling function of the Jth layer.

[0071] Finally, in the EMD reconstruction part, the high-frequency IMF components reconstructed by wavelet and the low-frequency IMF components obtained by EMD decomposition in the first step are fused and reconstructed to obtain the denoised time-frequency domain signal data.

[0072] (2) Abnormal data elimination

[0073] In the status monitoring of infrastructure along the high-speed railway, sensor data may be abnormal due to complex external environmental factors and the influence of electromagnetic interference and air vibration generated by train operation. For example, common abnormal conditions of acceleration sensors include data loss, data jumps, abnormal trend changes, etc. These abnormal data are widely present in various time periods and locations along the line, affecting the accurate assessment of the real-time health status of high-speed railway infrastructure, and may make it impossible to timely grasp the dynamic performance and potential safety hazards of the structure. Therefore, it is extremely important to detect abnormal data and eliminate it in a timely manner. The present invention uses support vector machines (SVM) to identify and eliminate abnormal data.

[0074] like Figure 5 As shown in the figure, SVM is used to remove abnormal data, and the denoised data is first standardized to normalize the data to a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0075]

[0076] In the formula, μ is the mean of the data, and σ is the standard deviation of the data.

[0077] The goal of training an SVM model using normalized data is to enclose most normal data within a boundary, leaving only a few data points outside the boundary. These values ​​outside the boundary are considered outliers. The model training is performed according to the following formula.

[0078]

[0079] Where w is the normal vector of the hyperplane, which is expressed as i '={x1',x2',x3',...,x n '} data set, find a hyperplane w so that normal data is on one side of the hyperplane as much as possible, and abnormal data is on the other side.

[0080] At the same time, select a suitable kernel function for the SVM model to train. After obtaining the trained model, for any data set X i The data point x in n ', input it into the model and get the predicted value g(x):

[0081] g(x)=SVM(x)-ρ

[0082] Where ρ is the bias term and SVM(x) is the trained SVM model.

[0083] If g(x)>0, it is a normal value, otherwise it is an abnormal value and should be removed.

[0084] S300 model method framework. The model method framework of this solution is based on the idea of ​​"prediction first and then key area monitoring". By combining data-driven prediction models with real-time monitoring data, dynamic feedback on the over-limit status of infrastructure along the high-speed rail line can be achieved. The overall framework of the model method is shown in the figure below. Figure 6 As shown in the figure. First, an infrastructure status prediction model is constructed using historical monitoring data, geological conditions, and train operation information. Real-time sensor data is input into the model to predict the current structural health of the infrastructure. If the model identifies an over-limit risk, an early warning mechanism is triggered and abnormal areas are marked. Subsequently, flaw detection trains equipped with high-precision sensors are dispatched to monitor key areas of predicted abnormalities, acquiring high-frequency, fixed-point, and accurate data. The causes of abnormalities are analyzed and the infrastructure status is confirmed. Finally, key monitoring results and real-time data are fed back into the prediction model for correction and optimization, forming a closed-loop process of "status prediction → key monitoring → data feedback → model optimization." This continuously improves the accuracy and robustness of the prediction model, enabling precise monitoring and safety assurance of infrastructure along the high-speed rail line.

[0085] S301 Prediction Model Framework. In order to more accurately complete the task of predicting the status of infrastructure along the high-speed rail line, the present invention designs an improved unsupervised abnormal state prediction model based on the LSTM network, such as Figure 7 As shown in Figure 2, the framework consists of two parts: feature extraction module and anomaly prediction module.

[0086] (1) Feature extraction module

[0087] The feature extraction module includes two stages: feature preprocessing and feature processing. In the feature pre-extraction stage, in order to more effectively capture the time series characteristics in the data, the present invention uses a temporal convolutional neural network (TCN) to pre-extract time domain information before feature extraction. TCN uses one-dimensional convolution operations combined with causal convolution and dilated convolution to extract multi-scale time-dependent features, thereby enhancing the model's ability to express temporal information and providing richer feature representations for subsequent feature processing and prediction. In the feature extraction stage, the LSTM network is used for temporal feature extraction, and the CNN network is used for spatial feature extraction.

[0088] ① Temporal feature extraction

[0089] like Figure 7 As shown in the figure, in the temporal feature extraction section, the LSTM model, with its excellent memory and generalization capabilities, can effectively capture the key features contained in the time series data of facility status anomalies and handle the complex sequential relationship between the external environment and infrastructure status anomalies. At each time step, the LSTM network calculates the control signals for the input gate, forget gate, and output gate based on the current input data and the hidden state at the previous moment, thereby dynamically adjusting the selection and memory of information in the time series. The mathematical expression relationship between the input, forget, and output gates for temporal feature extraction is:

[0090] The calculation process of the input gate:

[0091] i t =σ(W i ×[h t-1 ,X r '(t)]+b i )

[0092]

[0093] Where W i and W c is the weight matrix of the input gate, b i and b c is the offset of the input gate, is the candidate memory cell state.

[0094] The calculation process of the forget gate:

[0095] f t =σ(W f ×[h t-1 ,X r '(t)]+b f )

[0096] Where W f is the weight matrix of the input gate, b f is the offset of the input gate.

[0097] The calculation process of the output gate:

[0098] o t =σ(W o ×[h t-1 ,X r '(t)]+b o )

[0099] h t =o t ×tanh(C t )

[0100] Overall time iterative calculation process:

[0101]

[0102] Where C t Represents the state of the memory cell at the current time step t.

[0103] ②Spatial feature extraction

[0104] In the high-speed rail infrastructure monitoring data, some data have spatial characteristics. In order to more efficiently extract the spatial features in the infrastructure monitoring data along the high-speed rail, the present invention uses a convolutional neural network (CNN) to perform deep feature extraction on the spatial dimension of the data. Through its convolutional layer and pooling layer structure, CNN can capture the local spatial patterns and features hidden in the monitoring data, such as the changing trend of infrastructure status at different locations and the distribution of abnormal areas. It extracts low-level spatial features of the input data through convolution operations, and further abstracts high-level features in subsequent convolution layers, so that it can characterize the global and local characteristics of the infrastructure status. Ultimately, these spatial features will be passed to the subsequent prediction model for the prediction of facility status and anomaly detection. Its mathematical expression can be expressed as:

[0105]

[0106] Where H(i,j) is the value of the output feature map at position (i,j), σ(·) is the ReLU activation function, and b is the bias term.

[0107] (2) Abnormal prediction module

[0108] The network consists of four main components: encoder, decoder, prediction module and loop critic. During the training phase, the two models of the prediction network that share the encoder are trained simultaneously, where the encoder-decoder loss function ζ pred and the automatic cycle judge loss function ζ judge They are:

[0109]

[0110] Where N is the total number of data samples, the true value of the i-th sample, the predicted value of the i-th sample, and Ω(·) represents the quality score calculation function.

[0111] During the prediction phase, after the time series is processed by the encoder and decoder, a prediction result for the current phase is generated. The prediction result is then evaluated for quality using an automatic loop evaluator. If the current score does not exceed the score of the previous prediction round and the number of loops has not reached the maximum limit, the evaluator uses the current prediction result as the initial state and restarts the next round of calculations. If the score meets the requirement or the number of loops reaches the upper limit, the loop is terminated and the final prediction result is output.

[0112] S302 Key area monitoring module based on prediction results

[0113] Key area monitoring is a key step in further analyzing the abnormal areas identified in the prediction phase. Figure 8 As shown in the figure, when the prediction model detects a facility at a certain location at risk of exceeding limits, it automatically triggers an early warning mechanism, marks the location as an abnormal area, and generates corresponding monitoring task instructions. Based on the task instructions, flaw detection trains or portable inspection equipment equipped with high-precision sensors are dispatched to conduct targeted, high-frequency, and refined monitoring of the abnormal area, acquiring high-resolution, comprehensive monitoring data. The collected data is then subjected to multidimensional analysis, combined with historical data and geological conditions, to deeply explore the potential causes of the abnormal area and comprehensively assess its actual impact on infrastructure safety. Finally, the analysis results are fed back into the prediction model, and the model's parameters and algorithms are adjusted and optimized, forming a closed-loop monitoring and prediction mechanism to continuously improve the accuracy and robustness of the predictions.

[0114] (1) Abnormal area triggering and task scheduling module

[0115] During the prediction phase, if the model identifies an abnormal infrastructure state or a risk of exceeding limits, the key area monitoring process is triggered and the area is marked as an abnormal area. Based on the spatial location, risk level, and infrastructure type of the abnormal area, monitoring task instructions are generated and a flaw detection train equipped with more sensors is dispatched to perform precise monitoring tasks. This can be expressed mathematically as:

[0116]

[0117] In the formula, R(A i ) represents area A i The risk score of is, and T is the risk threshold.

[0118] (2) High-precision monitoring module

[0119] The high-precision monitoring module is the data acquisition part and core of the key area monitoring framework. This module uses various dynamic point sensors mounted on the flaw detection vehicle to collect point cloud data, acoustic wave information, image data, vibration frequency, GNSS and other multi-dimensional data to implement fixed-point, high-frequency and detailed monitoring of specifically marked abnormal areas.

[0120] The point cloud data obtained by the flaw detection train is mainly used to construct a three-dimensional model of the abnormal area, showing the geometric structure and morphology of the area, so as to analyze the subtle deformation and structural defects of the facilities in the abnormal area; the acquired acoustic wave information can be used to detect and evaluate the internal structure of the facilities in the abnormal area, and to identify whether there are abnormalities such as cracks, voids and other internal defects; the acquired image data is mainly used for the visual detection of surface damage, cracks, corrosion and other problems. The image information can be used to accurately identify and locate surface defects, and provide basic data for subsequent trend analysis; the collected vibration frequency data is used to judge the dynamic performance and stability of the facilities in the abnormal area; the collected GNSS data can be used to track displacement changes such as settlement and offset in the abnormal area.

[0121] After data collection is completed, all information will be immediately transmitted to the data analysis module for further processing and analysis.

[0122] (3) Data analysis and abnormality confirmation module

[0123] The data analysis and anomaly confirmation module is mainly responsible for in-depth processing and anomaly confirmation of the collected monitoring data set S. For the data collected in real time by the flaw detection train, a comprehensive analysis of the temporal and spatial characteristics is carried out to extract key features to determine whether there is an abnormal state, including the type and severity of the abnormality. The risk assessment model f is established based on the infrastructure requirements in the Railway Technical Management Regulations (High-speed Railway Section) risk , judge the abnormal risk level R(A i ).

[0124] R(A i )=f risk (S)

[0125] The output abnormal risk level index provides a decision basis for further optimization and risk control of the facility state. Finally, the results of the key area monitoring are fed back to the dispatch center to support the development of repair plans.

[0126] (4) Data feedback and model optimization module

[0127] The analysis results of the key area monitoring and the environmental information data collected by the real-time fixed-point sensor are fed back to the prediction model together. The weights and hyperparameters of the model are adjusted through supervised learning, and the model structure and parameters are iteratively optimized, so as to improve the identification accuracy of the prediction model for abnormal states and make it more adaptive to the dynamic changes of the complex environment along the high-speed rail. The mathematical expression of the model update can be expressed as:

[0128]

[0129] In the formula, ζ is the loss function, Y is the true value, is the model parameter.

[0130] In the specification, each embodiment is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between each embodiment can be referred to each other.

[0131] In this paper, specific examples are used to illustrate the principles and implementation methods of the application. The above description of the embodiments is only used to help understand the core idea of the application; at the same time, for those skilled in the art, according to the idea of the application, the specific implementation method and application range will be changed. In summary, the content of the specification should not be understood as a limitation of the application.

Claims

1. A method for monitoring the health status of infrastructure along a high-speed railway, characterized in that: include: Establish a monitoring framework consisting of four monitoring categories; the monitoring categories include structural health monitoring of infrastructure along the high-speed railway line, roadbed condition monitoring, track condition monitoring, and external environment and geological disaster monitoring along the high-speed railway line; Arrange a data collection plan based on the monitoring framework and obtain monitoring data; The health status of infrastructure along the high-speed railway is predicted based on monitoring data and a status monitoring model to determine risk prediction results; the risk prediction results include potential risk areas and corresponding types; the status monitoring model includes a prediction model and a key area monitoring model based on the prediction results.

2. The method for monitoring the health status of infrastructure along a high-speed railway according to claim 1, characterized in that: The construction of a monitoring framework consisting of four monitoring categories includes: Constructing a functional layer for structural health monitoring of infrastructure along the high-speed railway based on the infrastructure's underlying state data and structural degradation data; the underlying state data includes deformation data, crack data, and corrosion data; and the structural degradation data includes geological changes, temperature changes, and train vibrations. Construct a functional layer for roadbed condition monitoring based on the left-right and vertical displacement and deformation of the high-speed railway roadbed; Constructing a functional layer for track condition monitoring based on track-related data; the relevant data is used to reflect the smoothness and surface damage of the track; the relevant data includes track geometry data, crack data, and wear data; A functional layer for monitoring the external environment and geological disasters along the high-speed railway is constructed based on environmental geological data; the environmental geological data includes extreme weather, landslide data, mud-rock flow data and earthquake data.

3. The method for monitoring the health status of infrastructure along a high-speed railway according to claim 1, characterized in that: Arrange a data collection plan based on the monitoring framework, specifically including: Deploy sensors along the high-speed rail line, leaving the fixed-point sensors on the original line unchanged, and deploy perception sensors on the flaw detection train; Based on the deployed sensors, monitoring data is collected and processed using the adaptive decomposition algorithm and SVM model to obtain the monitoring data to be transmitted; Data transmission is carried out using a multi-level transmission method that combines 5G communication network and satellite communication.

4. The method for monitoring the health status of infrastructure along a high-speed railway according to claim 3, characterized in that: The fixed-point sensors include temperature sensors, humidity sensors, wind speed sensors and GNSS receivers; the perception sensors include lidars, depth cameras, ultrasonic sensors, vibration sensors and mobile GNSS devices.

5. The method for monitoring the health status of infrastructure along a high-speed railway according to claim 1, characterized in that: Before predicting the health status of infrastructure along the high-speed rail line based on monitoring data and the condition monitoring model, the following steps are required: building and training the condition monitoring model. Constructing a pre-trained network based on an LSTM network and a CNN network, optimizing initial parameters of the pre-trained network using historical data, and determining the optimized network as a prediction model; the historical data includes external environmental data and infrastructure overrun data; The prediction model is used to make preliminary predictions, identify preliminary key areas, and build a key area monitoring model based on the in-depth inspection data of the preliminary key areas. Finally, the key monitoring data output by the key area monitoring model is fed back into the prediction model for dynamic optimization and iteration to obtain the final status monitoring model.

6. A health status monitoring system for infrastructure along a high-speed railway, characterized in that: include: A functional construction unit for constructing a monitoring framework consisting of four monitoring categories, including structural health monitoring of infrastructure along the high-speed railway, roadbed condition monitoring, track condition monitoring, and external environment and geological disaster monitoring along the high-speed railway; A data monitoring unit, configured to arrange a data collection plan based on the monitoring framework and obtain monitoring data; The status assessment unit is used to predict the health status of infrastructure along the high-speed railway based on monitoring data and a status monitoring model, and determine risk prediction results; the risk prediction results include potential risk areas and corresponding types; the status monitoring model includes a prediction model and a key area monitoring model based on the prediction results.

7. An electronic device, characterized in that: It includes a memory and a processor, the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the health status monitoring method for infrastructure along a high-speed railway according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that It stores a computer program, which, when executed by a processor, implements the method for monitoring the health status of infrastructure along a high-speed railway as described in any one of claims 1 to 5.

Images