High-speed rail earthquake emergency disposal equipment linkage control method and device, equipment and medium

Abstract

The application relates to a high-speed rail earthquake emergency disposal equipment linkage control method, device, equipment and medium. The method comprises the following steps: firstly, collecting multi-source vibration and geological displacement information of key nodes along the line, removing interference noise through a filtering algorithm, and generating a purified earthquake precursor signal data set. Then, a feature vector is extracted from the data set, the preliminary intensity is quantitatively calculated, and after the threshold value, a warning signal containing a time stamp, a warning level and the like is generated through multi-dimensional verification. Then, based on the warning signal feature parameters and the geological data, the earthquake wave propagation mode is analyzed by using deep learning, the track displacement and crack detection results are integrated, and a section damage level sequence sorted by mileage is generated. Finally, a priority list is constructed according to the sequence, train shutdown scheduling and rescue resource allocation are guided, and an optimized emergency disposal scheme is obtained. The method realizes efficient response of the whole chain from earthquake precursor monitoring to emergency disposal, and improves the safety guarantee capability and emergency disposal efficiency of high-speed rail in response to earthquake disasters.

Description

Technical Field

[0001] This invention belongs to the field of railway transportation technology, and in particular relates to the linkage control method, device, equipment and medium for high-speed railway earthquake emergency response equipment. Background Technology

[0002] As a high-capacity, high-speed rail transit system, the operational safety of high-speed rail is closely related to the geological environment along its route. Earthquakes, as sudden geological disasters, can cause damage to infrastructure such as track deformation and bridge damage, posing a direct threat to train operation safety. Current earthquake monitoring technologies largely rely on independent seismic networks, which lack sufficient coordination with equipment along the high-speed rail line. Data processing is also lagging, making it difficult to quickly generate targeted emergency response plans. Furthermore, the removal of train interference and environmental noise from vibration signals is limited, resulting in low accuracy in identifying earthquake precursor signals, affecting the timeliness of early warnings and the accuracy of responses, and failing to meet the high reliability requirements of high-speed rail for earthquake emergency response. Summary of the Invention

[0003] Therefore, it is necessary to provide a method, device, equipment, and medium for the linkage control of high-speed rail earthquake emergency response equipment that can improve the safety assurance capability and emergency response efficiency of high-speed rail in response to earthquake disasters.

[0004] Firstly, this application provides a method for the coordinated control of high-speed rail earthquake emergency response equipment, including:

[0005] Multi-source vibration data and geological displacement information of key nodes along the high-speed railway were obtained. Train operation interference and environmental noise were removed by filtering algorithm to obtain a purified earthquake precursor signal dataset.

[0006] Feature vectors are extracted from the signal dataset to determine the preliminary intensity of earthquake precursors. When the preliminary intensity exceeds a preset threshold, a warning signal is obtained by performing multi-dimensional parameter verification analysis based on the feature vectors. The feature vectors include amplitude frequency, propagation velocity, and signal period.

[0007] By using deep learning algorithms to analyze the wave propagation patterns in complex geological environments along the route, the characteristic parameters associated with the early warning signals and geological data are analyzed. By integrating track displacement monitoring data and crack detection results, the damage level sequence of each section is obtained.

[0008] A priority ranking list is constructed based on the damage level sequence to guide train suspension scheduling and rescue resource allocation, resulting in an optimized emergency response plan.

[0009] In one embodiment, multi-source vibration data and geological displacement information of key nodes along the high-speed railway are acquired. A filtering algorithm is then used to remove train operation interference and environmental noise, resulting in a purified earthquake precursor signal dataset, including:

[0010] Sensors deployed at key nodes along the high-speed railway line collect multi-source vibration data and geological displacement information in real time; key nodes include bridge bearings, tunnel entrances and exits, track beds, and stress concentration areas in fault fracture zones along the line.

[0011] A filtering algorithm is used to denoise the multi-source vibration data to remove the periodic vibration interference generated by train operation, thus obtaining the first vibration dataset. The filtering algorithm includes a combination of adaptive notch filtering and wavelet threshold denoising.

[0012] Low-frequency components are extracted from the first vibration dataset based on a preset frequency threshold, potential earthquake precursor signals are screened out, and a purified second vibration dataset is generated.

[0013] Based on geological displacement information, the second vibration dataset is compensated for spatial coordinates and amplitude using a preset correction model to obtain a corrected earthquake precursor signal dataset. The correction model uses the least squares method to compensate for spatial attenuation of the vibration signal and correct baseline drift.

[0014] In one embodiment, feature vectors are extracted from the signal dataset to determine the preliminary intensity of earthquake precursors. When the preliminary intensity exceeds a preset threshold, a warning signal is obtained by performing multi-dimensional parameter verification analysis based on the feature vectors, including:

[0015] Feature vectors, including amplitude frequency, propagation speed, and signal period, are extracted from the signal dataset to construct a feature matrix.

[0016] The preliminary intensity of earthquake precursors is determined by quantifying the feature matrix using a pre-set intensity assessment model, which is based on a BP neural network.

[0017] When the initial intensity exceeds the preset threshold, a time-domain feature set is generated based on the feature vector using time-domain analysis, and then converted into a frequency-domain feature set through Fourier transform.

[0018] The frequency domain feature set is matched with the preset earthquake precursor model library to obtain the corresponding earthquake precursor model parameters. The earthquake precursor model library contains historical earthquake cases classified and stored by magnitude and focal depth. Each earthquake precursor model contains a typical frequency domain feature vector and the corresponding geological response parameters.

[0019] Multi-dimensional parameter verification is performed based on earthquake precursor model parameters to generate a verification result set; the multi-dimensional parameter verification includes wave velocity consistency verification and amplitude attenuation law verification.

[0020] The verification result set is correlated with the historical earthquake database to determine the earthquake early warning level and generate an early warning signal containing timestamp, early warning level, source characteristics and monitoring location information; the historical earthquake database contains magnitude, epicenter location, time of occurrence, peak ground acceleration and infrastructure damage records.

[0021] In one embodiment, the preliminary intensity of the earthquake precursor is calculated using the following formula:

[0022]

[0023] Where I represents the initial intensity of the earthquake precursor, standardized to a value of 0-10, with 10 corresponding to the highest risk, and σ(·) represents the Sigmoid activation function. f represents the peak amplitude frequency in the characteristic matrix, v represents the propagation speed in the characteristic matrix, and c represents the signal periodicity stability coefficient in the characteristic matrix. The value range is [0,1]. ω1, ω2, and ω3 represent weight parameters, which are determined through training with historical data. b represents the bias term, which is trained to make the output fit the preset intensity range.

[0024] In one embodiment, deep learning algorithms are used to analyze wave propagation patterns in complex geological environments along the route, based on the characteristic parameters associated with the early warning signals and geological data. This is combined with track displacement monitoring data and crack detection results to obtain a damage level sequence for each section, including:

[0025] An analytical model is constructed using deep learning algorithms based on the characteristic parameters associated with the early warning signals and geological data. The characteristic parameters include wave velocity and amplitude characteristics; the geological data includes stratigraphic lithology and fault distribution.

[0026] The analysis model is used to simulate the propagation mode of seismic waves in complex geological environments along the route, calculate the vibration response values ​​of infrastructure, and generate a potential damage distribution map.

[0027] A weighted fusion algorithm was used to integrate potential damage distribution maps, track displacement monitoring data, and crack detection results to obtain multi-source data fusion results. Track displacement monitoring data included the lateral and longitudinal displacement of the track and the time series change rate. Crack detection results included crack length, maximum width, and propagation rate.

[0028] Based on the results of multi-source data fusion, a comprehensive damage index is calculated for sections along the high-speed railway. The level is divided according to the score interval, and a damage level sequence is generated in mileage order.

[0029] In one embodiment, the comprehensive damage index is calculated using the following formula:

[0030] D = 0.6 × P + 0.25 × DG +0.15×D L

[0031] Where D represents the comprehensive damage index of the high-speed rail section, and P represents the risk score of the section corresponding to the potential damage distribution map. G Indicates the quantitative score of orbital displacement. d h d represents the lateral displacement. v D represents the longitudinal displacement, r represents the time-series rate of change, and D L This indicates a quantitative score for crack detection. L represents the crack length, W represents the maximum width, and V represents the propagation rate.

[0032] In one embodiment, a priority ranking list is constructed based on the damage level sequence to guide train suspension scheduling and rescue resource allocation, resulting in an optimized emergency response plan, including:

[0033] Damage level data is extracted from the damage level sequence; the damage level data includes damage level labels for each segment and the corresponding mileage interval.

[0034] The random forest machine learning algorithm is used to classify and train the damage level data, with the section damage level, train density and peak passenger flow period as input features, and the output priority ranking list.

[0035] Train cancellation dispatch instructions are generated based on a priority sorting list combined with the real-time train timetable; the dispatch instructions clearly indicate the cancellation order, start and end times of cancellation, and alternative connection schemes for each train.

[0036] The system invokes a priority sorting list and shutdown dispatch instructions, associates with a preset rescue resource database, and generates a rescue resource allocation plan. The rescue resource allocation plan includes resource types and the mileage range of the corresponding allocation area. The rescue resource database includes information on engineering vehicles, repair personnel, and material reserves.

[0037] Based on the rescue resource allocation scheme, the shortest emergency path for each allocation area is calculated using a path optimization algorithm. By integrating path information with resource allocation sequence, an emergency response plan is obtained that includes repair sequence, resource arrival time, and path planning.

[0038] Secondly, this application also provides a linkage control device for high-speed rail earthquake emergency response equipment, the device including:

[0039] The signal purification module is used to acquire multi-source vibration data and geological displacement information of key nodes along the high-speed railway, and remove train operation interference and environmental noise through filtering algorithms to obtain a purified earthquake precursor signal dataset.

[0040] The early warning determination module is used to extract feature vectors from the signal dataset to determine the preliminary intensity of earthquake precursors. When the preliminary intensity exceeds a preset threshold, a warning signal is obtained by performing multi-dimensional parameter verification analysis based on the feature vectors. The feature vectors include amplitude frequency, propagation velocity, and signal period.

[0041] The damage rating module is used to analyze the wave propagation patterns in complex geological environments along the route using deep learning algorithms based on the characteristic parameters associated with the early warning signals and geological data. It integrates track displacement monitoring data and crack detection results to obtain the damage level sequence for each section.

[0042] The emergency dispatch module is used to construct a priority ranking list based on the damage level sequence to guide train suspension scheduling and rescue resource allocation, resulting in an optimized emergency response plan.

[0043] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0044] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method.

[0045] The aforementioned high-speed railway earthquake emergency response equipment linkage control method, device, computer equipment, and storage medium first collect multi-source vibration data and geological displacement information from key nodes along the high-speed railway line. A filtering algorithm is then used to remove train operation interference and environmental noise, generating a purified earthquake precursor signal dataset, providing a high-quality data foundation for subsequent analysis. Next, amplitude frequency, propagation velocity, and signal period feature vectors are extracted from the signal dataset. Quantitative calculations determine the preliminary intensity of the earthquake precursor. When the intensity exceeds a preset threshold, a warning signal containing timestamps, warning levels, and other information is generated after multi-dimensional parameter verification, achieving accurate risk identification. Subsequently, based on the characteristic parameters associated with the warning signal and geological data, deep learning algorithms are used to analyze seismic wave propagation patterns, integrate track displacement monitoring data and crack detection results, and generate a segment damage level sequence ordered by mileage, completing a refined assessment of the damage degree. Finally, a priority ranking list is constructed based on the damage level sequence to guide train suspension scheduling and rescue resource allocation, ultimately resulting in an optimized emergency response plan. This method improves the accuracy of earthquake precursor identification through signal purification, ensures the reliability of early warning through multi-dimensional verification, achieves accurate damage level classification by combining deep learning, and optimizes emergency resource allocation based on priority ranking. It effectively solves the problems of insufficient data interference removal, weak early warning linkage, and poor targeting of response plans in existing technologies, and realizes efficient response across the entire chain from earthquake precursor monitoring to emergency response, thereby improving the safety guarantee capability and emergency response efficiency of high-speed rail in response to earthquake disasters. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 A flowchart of the linkage control method for high-speed railway earthquake emergency response equipment provided in an embodiment of the present invention;

[0048] Figure 2 The structural block diagram of the linkage control device for high-speed rail earthquake emergency response equipment provided in the embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] In one embodiment, such as Figure 1 As shown, this application provides a method for the linkage control of high-speed railway earthquake emergency response equipment, which may include the following steps:

[0051] Step S101: Obtain multi-source vibration data and geological displacement information of key nodes along the high-speed railway line, and remove train operation interference and environmental noise through filtering algorithms to obtain a purified earthquake precursor signal dataset.

[0052] Specifically, sensors deployed at key nodes along the high-speed railway line (such as bridge bearings, tunnel entrances and exits, track beds, and stress concentration areas in fault fracture zones) collect multi-source vibration data (such as ground vibration acceleration and structural vibration displacement) and geological displacement information (such as ground subsidence and fault activity) in real time. After data collection, a combination algorithm of adaptive notch filtering and wavelet threshold denoising is used to process the raw data, specifically removing periodic vibration interference and environmental noise generated by train operation, ultimately generating a purified earthquake precursor signal dataset.

[0053] Step S102: Extract feature vectors from the signal dataset to determine the preliminary intensity of earthquake precursors. When the preliminary intensity exceeds a preset threshold, perform multi-dimensional parameter verification analysis based on the feature vectors to obtain an early warning signal. The feature vectors include amplitude frequency, propagation velocity, and signal period.

[0054] Specifically, from the purified earthquake precursor signal dataset, core feature vectors such as peak amplitude frequency, seismic wave propagation velocity, and signal periodic stability coefficient are extracted according to preset feature extraction rules. These vectors are then arranged by nodes and time sequence to construct a feature matrix. A preset intensity assessment model based on a BP neural network is used to quantify the feature matrix, obtaining the preliminary intensity of earthquake precursors from magnitude 0 to 10. When the preliminary intensity exceeds a preset threshold (e.g., magnitude 6.5), time-domain analysis is used to generate time-domain feature sets such as peak factor and kurtosis value. These are then converted into frequency-domain feature sets such as dominant frequency and spectral entropy through Fourier transform. After matching with a preset earthquake precursor model library and performing multi-dimensional verification such as wave velocity consistency and amplitude attenuation law, a warning signal containing timestamp, warning level, source characteristics, and monitoring location is generated.

[0055] Step S103: Analyze the wave propagation patterns in the complex geological environment along the route using deep learning algorithms based on the characteristic parameters associated with the early warning signals and geological data, and integrate track displacement monitoring data and crack detection results to obtain the damage level sequence for each section.

[0056] Based on characteristic parameters such as wave velocity and amplitude associated with the early warning signals, as well as geological data such as lithology and fault distribution along the route, an analysis model is constructed using deep learning algorithms such as the improved U-Net. This model simulates the propagation path, refraction and reflection patterns, and vibration energy attenuation process of seismic waves in complex geological environments, calculating the vibration response values ​​of infrastructure such as bridges, tunnels, and railways, and generating a potential damage distribution map with a spatial resolution of 10m × 10m. A weighted fusion algorithm (potential damage distribution accounting for 60%, track displacement for 25%, and crack detection for 15%) is used to integrate the potential damage distribution map, track lateral / longitudinal displacement and time-series change rate data, and crack length / width / propagation rate data to obtain multi-source data fusion results. Finally, a comprehensive damage index is calculated by section and classified into levels, generating a damage level sequence arranged in mileage order.

[0057] Step S104: Construct a priority ranking list based on the damage level sequence to guide train suspension scheduling and rescue resource allocation, thereby obtaining an optimized emergency response plan.

[0058] Damage level labels (e.g., emergency repair, priority maintenance) and corresponding mileage intervals for each section are extracted from the damage level sequence. A random forest machine learning algorithm is used, with section damage level, train density, and peak passenger flow periods as input features for classification training, outputting a list sorted by emergency priority. Combined with real-time train timetables, dispatch instructions are generated based on the priority ranking list, including the order of train cancellations, start and end times, and alternative connection schemes. This priority list and dispatch instructions are then invoked, linked to a pre-set rescue resource database containing information on engineering vehicles, repair personnel, and material reserves, to generate a resource allocation plan with clearly defined resource types and allocation area mileage ranges. Finally, Dijkstra's path optimization algorithm is used to calculate the shortest emergency path, integrating path information with resource allocation timing to form an optimized emergency response plan that includes repair sequence, resource arrival time, and path planning.

[0059] The aforementioned high-speed railway earthquake emergency response equipment linkage control method first collects multi-source vibration data and geological displacement information from key nodes along the high-speed railway line. A filtering algorithm is then used to remove train operation interference and environmental noise, generating a purified earthquake precursor signal dataset, providing a high-quality data foundation for subsequent analysis. Next, amplitude frequency, propagation velocity, and signal period feature vectors are extracted from the signal dataset. Quantitative calculations determine the preliminary intensity of the earthquake precursor. When the intensity exceeds a preset threshold, a warning signal containing timestamps, warning levels, and other information is generated after multi-dimensional parameter verification, achieving accurate risk identification. Subsequently, based on the characteristic parameters associated with the warning signal and geological data, deep learning algorithms are used to analyze seismic wave propagation patterns, integrate track displacement monitoring data and crack detection results, and generate a segment damage level sequence ordered by mileage, completing a refined assessment of the damage degree. Finally, a priority ranking list is constructed based on the damage level sequence to guide train suspension scheduling and rescue resource allocation, ultimately resulting in an optimized emergency response plan. This method improves the accuracy of earthquake precursor identification through signal purification, ensures the reliability of early warning through multi-dimensional verification, achieves accurate damage level classification by combining deep learning, and optimizes emergency resource allocation based on priority ranking. It effectively solves the problems of insufficient data interference removal, weak early warning linkage, and poor targeting of response plans in existing technologies, and realizes efficient response across the entire chain from earthquake precursor monitoring to emergency response, thereby improving the safety guarantee capability and emergency response efficiency of high-speed rail in response to earthquake disasters.

[0060] In one embodiment, acquiring multi-source vibration data and geological displacement information of key nodes along the high-speed railway line, and removing train operation interference and environmental noise through a filtering algorithm to obtain a purified earthquake precursor signal dataset may include the following steps:

[0061] Step S201 involves real-time collection of multi-source vibration data and geological displacement information using sensors deployed at key nodes along the high-speed railway line; key nodes include bridge bearings, tunnel entrances and exits, track beds, and stress concentration areas in fault fracture zones along the line.

[0062] Key nodes, such as bridge bearings, are crucial supporting components of the bridge structure; their vibration and displacement reflect the overall stress and stability of the bridge. Tunnel entrances and exits are weak points in the tunnel structure, easily deformed by factors such as earthquakes. The track bed is the foundation for train operation, and its condition changes are closely related to train safety. Stress concentration areas along the fault fracture zone are geologically active locations, more likely to generate earthquake-related signals. Sensors collect multi-source vibration data and geological displacement information from these key nodes in real time.

[0063] Step S202: A filtering algorithm is used to denoise the multi-source vibration data to remove periodic vibration interference generated by train operation, thus obtaining the first vibration dataset. The filtering algorithm includes a combination of adaptive notch filtering and wavelet threshold denoising.

[0064] Preferably, the acquired multi-source vibration data contains various interferences and noises, with periodic vibration interference from train operation being one of the main sources. To obtain more accurate seismic correlation signals, filtering algorithms are needed to denoise the data. The filtering algorithm used here is a combination of adaptive notch filtering and wavelet thresholding. Adaptive notch filtering automatically adjusts the notch frequency according to the characteristics of the input signal, effectively suppressing periodic interference at specific frequencies generated by train operation. Wavelet thresholding decomposes the signal into different scales using wavelet transform, and then removes noise signals by setting an appropriate threshold, retaining useful signal components. After processing by these two filtering algorithms, a first vibration dataset is obtained, which is cleaner than the original acquired data, reducing the impact of train operation interference and other environmental noise.

[0065] Step S203: Based on a preset frequency threshold, low-frequency components are extracted from the first vibration dataset to screen out potential earthquake precursor signals and generate a purified second vibration dataset.

[0066] Earthquake precursor signals typically exhibit certain frequency characteristics, generally exhibiting relatively low frequencies. Based on this characteristic, after obtaining the first vibration dataset, low-frequency components are extracted from it using a preset frequency threshold. This preset frequency threshold, determined based on extensive research and practical experience, helps us filter out components more likely to be earthquake precursor signals. By extracting the low-frequency components, a purified second vibration dataset is generated, containing signals that are more likely to be potential earthquake-related precursor signals.

[0067] Step S204: Based on geological displacement information, spatial coordinates and amplitude compensation are performed on the second vibration dataset using a preset correction model to obtain the corrected earthquake precursor signal dataset; the correction model performs spatial attenuation compensation and baseline drift correction of the vibration signal using the least squares method.

[0068] Due to variations in sensor installation locations and geological conditions, the acquired vibration signals may exhibit spatial deviations, and their amplitude may also change due to the propagation path and medium. To more accurately analyze and utilize these signals, it is necessary to use geological displacement information as a benchmark and apply a pre-defined calibration model to compensate for the spatial coordinates and amplitude of the second vibration dataset. The pre-defined calibration model employs the least squares method for spatial attenuation compensation and baseline drift correction of the vibration signals. The least squares method is a mathematical optimization technique that finds the most suitable functional relationship for the data by minimizing the sum of squared errors. This model allows for the correction of the spatial location of the vibration signals, making them more reflective of the actual geological conditions, while also adjusting the amplitude to compensate for potential attenuation during signal propagation, thus obtaining a corrected earthquake precursor signal dataset.

[0069] Specifically, firstly, sensors deployed at key nodes along the high-speed railway, such as bridge supports, tunnel entrances and exits, track beds, and stress concentration areas in fault fracture zones, collect multi-source vibration data and geological displacement information in real time, ensuring data coverage of critical infrastructure parts and geologically risky areas. Then, a combined algorithm of adaptive notch filtering and wavelet threshold denoising is used to reduce noise in the vibration data, specifically removing periodic interference from train operation to generate a first vibration dataset. Based on a preset frequency threshold, low-frequency components are extracted from the first vibration dataset to screen for potential earthquake precursor signals, resulting in a purified second vibration dataset. Finally, using geological displacement information as a benchmark, a preset correction model based on the least squares method is used to perform spatial coordinate matching, amplitude attenuation compensation, and baseline drift correction on the second vibration dataset, ultimately obtaining a corrected earthquake precursor signal dataset.

[0070] This embodiment ensures targeted data acquisition through precise placement of key nodes, effectively eliminates train operation interference and environmental noise using a combined filtering algorithm, focuses on earthquake precursor features by extracting low-frequency components, and achieves spatial and amplitude compensation of the signal by combining a correction model based on geological displacement information. This significantly improves the signal-to-noise ratio and accuracy of earthquake precursor signals. It solves problems such as incomplete interference removal in vibration signals, significant environmental noise impact, and signal spatial deviation, providing high-quality basic data support for subsequent earthquake precursor identification and early warning analysis, and ensuring the reliability of input data for the early warning model.

[0071] In one embodiment, feature vectors are extracted from a signal dataset to determine the preliminary intensity of earthquake precursors. When the preliminary intensity exceeds a preset threshold, a warning signal is obtained by performing multi-dimensional parameter verification analysis based on the feature vectors. This may include the following steps:

[0072] Step S301: Extract feature vectors including amplitude frequency, propagation speed and signal period from the signal dataset and construct a feature matrix.

[0073] Preferably, key feature parameters are extracted from the corrected earthquake precursor signal dataset according to preset rules: amplitude frequency (in Hz, taken as the peak frequency in the signal spectrum), propagation velocity (in km / s, calculated by the time difference between adjacent sensor signals), and signal period (in s, taken as the stability coefficient of the signal waveform repetition period), forming a three-dimensional feature vector. The feature vectors of multiple key nodes in different time windows are arranged in rows to construct a feature matrix with dimensions of "number of nodes × number of time windows × feature dimension", realizing the structured integration of signal features.

[0074] Step S302: Quantitatively calculate the feature matrix using a preset intensity assessment model to determine the preliminary intensity of the earthquake precursor; the preset intensity assessment model is constructed based on a BP neural network.

[0075] Furthermore, the preset intensity assessment model is a three-layer BP neural network structure (number of input layer nodes = feature dimension, number of hidden layer nodes = feature dimension × 2, number of output layer nodes = 1), and the activation function is the Sigmoid function. After inputting the feature matrix into the model, the output value is calculated through forward propagation, and after standardization, the preliminary intensity of earthquake precursors from magnitude 0 to 10 (magnitude 10 corresponds to the highest risk) is obtained. The model is trained and optimized using a historical earthquake signal dataset (containing 500+ known earthquake cases), and the weight parameters and bias terms are determined iteratively through the backpropagation algorithm.

[0076] Step S303: When the initial intensity exceeds the preset threshold, a time-domain feature set is generated based on the feature vector using time-domain analysis, and then converted into a frequency-domain feature set through Fourier transform.

[0077] When the initial intensity exceeds a preset threshold (e.g., magnitude 6.5), time-domain analysis is performed based on feature vectors: time-domain statistics such as peak value, kurtosis, and root mean square are calculated to generate a time-domain feature set; then, the time-domain signal is converted to a frequency-domain signal using a Fast Fourier Transform (FFT), extracting parameters such as dominant frequency, spectral energy distribution, and bandwidth to form a frequency-domain feature set. The combination of time- and frequency-domain features can comprehensively capture the time-domain waveform characteristics and frequency-domain energy distribution patterns of earthquake precursor signals, improving the completeness of feature recognition.

[0078] Step S304: Perform pattern matching between the frequency domain feature set and the preset earthquake precursor pattern library to obtain the corresponding earthquake precursor pattern parameters; the earthquake precursor pattern library contains historical earthquake cases classified and stored by magnitude and focal depth, and each earthquake precursor pattern contains a typical frequency domain feature vector and the corresponding geological response parameters.

[0079] The frequency domain feature set is compared and matched with a pre-defined earthquake precursor model library. This model library stores historical earthquake cases categorized by magnitude (e.g., below magnitude 3, 3-5, and above magnitude 5) and focal depth (e.g., shallow, intermediate, and deep). Each model contains typical frequency domain feature vectors (e.g., dominant frequency, spectral peak value) and geological response parameters (e.g., ground vibration attenuation coefficient) for the corresponding earthquake type. Feature similarity is calculated using the cosine similarity formula to match the closest historical model and obtain the corresponding earthquake precursor model parameters.

[0080] Step S305: Perform multi-dimensional parameter verification based on earthquake precursor model parameters and generate a verification result set; the multi-dimensional parameter verification includes wave velocity consistency verification and amplitude attenuation law verification.

[0081] Multi-dimensional parameter verification is achieved through quantitative calculations: wave velocity consistency verification compares the deviation between the measured propagation velocity and the theoretical wave velocity in the model library (allowable deviation ≤15%); amplitude attenuation law verification is performed by fitting the attenuation curve of signal amplitude with propagation distance, and the deviation from the formation attenuation model in the model library must be ≤20%. Upon successful verification, a verification result set containing deviation values ​​and pass rates is generated; if it fails, a signal re-examination process is triggered to eliminate interfering signals and ensure the reliability of precursor parameters.

[0082] Step S306: The verification result set is correlated with the historical earthquake database to determine the earthquake early warning level and generate an early warning signal containing timestamp, early warning level, source characteristics and monitoring location information; the historical earthquake database contains magnitude, epicenter location, time of occurrence, peak ground acceleration and infrastructure damage records.

[0083] The historical earthquake database stores earthquake records within a 200km radius along the earthquake line over the past 50 years. Fields include magnitude, epicenter latitude and longitude, time of occurrence, peak ground acceleration (GFA) (in gal), and extent of infrastructure damage. The verification result set is correlated with records of similar earthquakes in the database (magnitude deviation ≤ 0.5, focal depth deviation ≤ 10km). Statistical analysis is used to determine the earthquake early warning level (Level I-IV, with Level I being the most urgent). The final generated warning signal is structured data, containing timestamps accurate to milliseconds, warning level, focal characteristics (magnitude, depth, estimated epicenter location), and latitude and longitude information of the monitoring nodes.

[0084] Specifically, based on the signal dataset, amplitude frequency, propagation velocity, and signal period are first extracted to construct a feature matrix. The feature matrix is ​​then quantified using a pre-defined intensity assessment model based on a BP neural network to determine the preliminary intensity of the earthquake precursor. When the preliminary intensity exceeds a preset threshold, a time-domain feature set is generated based on the feature vectors through time-domain analysis, and then a frequency-domain feature set is obtained through Fourier transform. The frequency-domain feature set is then matched with a pre-defined earthquake precursor model library (historical cases are stored according to magnitude and focal depth, including typical frequency-domain feature vectors and geological response parameters) to obtain the earthquake precursor model parameters. Multi-dimensional parameter verification of wave velocity consistency and amplitude attenuation patterns is then performed to generate a verification result set. Finally, the verification result set is correlated with a historical earthquake database (including records of magnitude, epicenter location, etc.) to determine the earthquake early warning level and generate an early warning signal containing timestamps, warning levels, focal characteristics, and monitoring location information.

[0085] This embodiment achieves structured integration of signal parameters through feature matrix construction, improves the accuracy of preliminary intensity assessment with the help of a BP neural network model, enhances the feature recognition capability of precursor signals through time-frequency domain transformation and pattern library matching, further filters interference signals through multi-dimensional parameter verification, and ensures the scientific nature of warning level determination by combining historical database correlation analysis. The final generated warning signal contains key information elements, effectively solving the problems of incomplete feature extraction, large intensity assessment deviation, and insufficient warning accuracy in traditional warnings.

[0086] In one embodiment, the preliminary intensity of the earthquake precursor can be calculated using the following formula:

[0087]

[0088] Where I represents the initial intensity of the earthquake precursor, standardized to a value of 0-10, with 10 corresponding to the highest risk, and σ(·) represents the Sigmoid activation function. f represents the peak amplitude frequency in the characteristic matrix, v represents the propagation speed in the characteristic matrix, and c represents the signal periodicity stability coefficient in the characteristic matrix. The value range is [0,1]. ω1, ω2, and ω3 represent weight parameters, which are determined through training with historical data. b represents the bias term, which is trained to make the output fit the preset intensity range.

[0089] Preferably, the formula for constructing the feature matrix is:

[0090]

[0091] Where F represents the 3×(n×m) dimension feature matrix, n represents the total number of key monitoring nodes deployed along the high-speed rail line, m represents the number of time-series data sets collected by each node (e.g., generating one data set every 5 minutes), fi,j v represents the peak amplitude frequency of the i-th data set at the i-th monitoring node. i,j c represents the seismic wave propagation velocity of the i-th data set at the i-th monitoring node. i,j This represents the signal periodic stability coefficient of the i-th data set at the i-th monitoring node. The value range is 0-1.

[0092] This embodiment introduces the Sigmoid activation function to achieve standardized mapping of the output value (0-10 points). Key parameters such as peak amplitude frequency, propagation velocity, and signal periodicity stability coefficient in the feature matrix are incorporated into the quantitative calculation. Combined with weight parameters and bias terms determined through training with historical data, it achieves a precise characterization of earthquake precursor intensity. The formula highlights the differences in the contribution of different feature parameters to precursor intensity through weight allocation. Utilizing the nonlinear transformation characteristics of the Sigmoid function, it adapts to the preset range of intensity assessment, ensuring that the output results reflect both the comprehensive influence of feature parameters and a clear risk level correspondence. This provides a quantitative and comparable intensity index for subsequent early warning threshold judgment, effectively improving the objectivity and consistency of preliminary earthquake precursor identification.

[0093] In one embodiment, the wave propagation pattern in the complex geological environment along the route is analyzed using a deep learning algorithm based on the characteristic parameters associated with the early warning signal and geological data. The damage level sequence for each section is obtained by integrating track displacement monitoring data and crack detection results. This process may include the following steps:

[0094] Step S401: Based on the characteristic parameters associated with the early warning signal and geological data, an analysis model is constructed using a deep learning algorithm; the characteristic parameters include wave velocity and amplitude characteristics; the geological data includes stratigraphic lithology and fault distribution.

[0095] Using dynamic parameters such as wave velocity and amplitude characteristics associated with early warning signals, and static geological data such as lithology of strata along the route (e.g., lithological classification of sedimentary and igneous rocks) and fault distribution (e.g., fault strike, dip angle, and activity), an analysis model is constructed using deep learning algorithms such as an improved U-Net. The model learns the mapping relationship between geological parameters and seismic wave propagation characteristics through multi-layer convolution and deconvolution structures. The feature parameters, as dynamic inputs, reflect the inherent properties of seismic waves, while the geological data, as static inputs, characterize the differences in the propagation medium. The combination of these two approaches enables the model to simulate complex geological environments.

[0096] Step S402: Use the analysis model to simulate the seismic wave propagation mode under complex geological conditions along the route, calculate the vibration response value of the infrastructure, and generate a potential damage distribution map.

[0097] By inputting geological data along the route (such as stratum thickness in different sections and fault fracture zone range) and early warning signal characteristic parameters (such as seismic wave velocity and peak amplitude) into the constructed analysis model, the propagation path, refraction and reflection patterns, and energy attenuation process of seismic waves in complex geological structures are simulated. By calculating the vibration response values ​​of infrastructure such as bridge bearing vibration acceleration, track bed displacement, and tunnel lining strain, the response values ​​are converted into spatial distribution data according to risk thresholds (e.g., vibration acceleration > 0.1g indicates high risk). This generates a potential damage distribution map with mileage as the horizontal axis and risk level as the vertical axis, visually presenting the estimated damage degree of each section.

[0098] Step S403: A weighted fusion algorithm is used to integrate the potential damage distribution map, track displacement monitoring data, and crack detection results to obtain multi-source data fusion results; the track displacement monitoring data includes the track lateral and longitudinal displacement and the time series change rate; the crack detection results include crack length, maximum width, and propagation rate.

[0099] A weighted fusion algorithm was employed to integrate three types of data: risk scores from potential damage distribution maps (weight 0.6), overlaid with track displacement monitoring data (weight 0.25, including quantified values ​​of lateral and longitudinal displacement and their temporal rates of change), and crack detection results (weight 0.15, including measured values ​​of crack length, maximum width, and propagation rate). During the fusion process, data of different dimensions were converted to score values ​​ranging from 0 to 100 through normalization, and then weighted summation was performed to obtain the multi-source data fusion result. This approach enabled complementary verification between predicted and measured data, improving data reliability.

[0100] Step S404: Calculate the comprehensive damage index for the high-speed rail section based on the multi-source data fusion results, classify the level standards according to the score intervals, and generate a damage level sequence arranged in mileage order.

[0101] Based on the results of multi-source data fusion, a quantitative score for each section along the high-speed railway is calculated using a comprehensive damage index formula (the formula includes a weighted calculation of potential damage score, track displacement score, and crack detection score). The scores are then categorized into risk levels according to preset ranges (e.g., 80-100 points for emergency repair, 60-79 points for priority maintenance, 30-59 points for routine inspection, and 0-29 points for no significant impact). The calculation results are then correlated with the corresponding section mileage to generate a damage level sequence arranged by mileage, clearly defining the risk level for different mileage ranges.

[0102] Specifically, using characteristic parameters such as wave velocity and amplitude associated with early warning signals, as well as geological data such as strata lithology and fault distribution, as inputs, a deep learning algorithm is used to construct an analysis model. This model accurately simulates the propagation path, energy attenuation law, and interaction mechanism of seismic waves in the complex geological environment along the route, and calculates the vibration response values ​​of structures such as bridges, tunnels, and tracks, generating a potential damage distribution map reflecting the spatial risk distribution. Based on this, a weighted fusion algorithm is used to integrate the potential damage distribution map (weight 0.6), track lateral / longitudinal displacement and time-series change rate data (weight 0.25), and crack length / maximum width / expansion rate data (weight 0.15) to obtain multi-source data fusion results. Then, the risk of each section is quantified using a comprehensive damage index formula, and the level standards are divided according to the score interval, finally generating a damage level sequence arranged in mileage order.

[0103] This embodiment achieves accurate simulation of seismic wave propagation under complex geological conditions through deep learning algorithms, improving the scientific rigor of potential damage prediction. Weighted fusion of multi-source data effectively integrates predicted and measured data, overcoming the limitations of single data sources. Quantitative calculation and classification of the comprehensive damage index make risk assessment more objective and operable. It solves problems such as insufficient model accuracy, low data utilization, and ambiguous classification in traditional geological risk assessment, providing reliable technical support for accurate identification and risk ranking of earthquake damage along high-speed railway lines, ensuring the targeted and effective nature of subsequent emergency response.

[0104] In one embodiment, the comprehensive damage index can be calculated using the following formula:

[0105] D = 0.6 × P + 0.25 × D G +0.15×D L

[0106] Where D represents the comprehensive damage index of the high-speed rail section, and P represents the risk score of the section corresponding to the potential damage distribution map. G Indicates the quantitative score of orbital displacement. d h d represents the lateral displacement. v D represents the longitudinal displacement, r represents the time-series rate of change, and D L This indicates a quantitative score for crack detection. L represents the crack length, W represents the maximum width, and V represents the propagation rate.

[0107] This embodiment achieves a multi-dimensional quantitative assessment of seismic damage along high-speed railway lines by integrating risk scores from potential damage distribution maps, quantitative scores from track displacement, and quantitative scores from crack detection. In the formula, the track displacement score is calculated using a weighted average of lateral displacement, longitudinal displacement, and time-series change rates, accurately reflecting the degree of deformation of the track structure. The crack detection score, based on a comprehensive consideration of crack length, maximum width, and propagation rate, effectively captures the structural damage status of the infrastructure. This formula, through structured parameter fusion and quantitative calculation, avoids the limitations of single-indicator assessments, enabling the comprehensive damage index to objectively reflect the actual damage level of the section. This provides a scientific basis for classifying damage levels according to score intervals, improving the accuracy and consistency of seismic damage assessment.

[0108] In one embodiment, constructing a priority ranking list based on the damage level sequence to guide train shutdown scheduling and rescue resource allocation results in an optimized emergency response plan, which may include the following steps:

[0109] Step S501: Extract damage level data from the damage level sequence; the damage level data includes damage level labels for each segment and corresponding mileage intervals.

[0110] Preferably, the core damage information for each section specifically includes damage level labels (such as "emergency repair" or "priority maintenance" classification indicators) and corresponding mileage intervals (the starting and ending mileages accurate to the hundred-meter level, such as K123+200-K125+500). Through structured data extraction, the continuous risk distribution along the line is transformed into discretized segment data that can be directly used for subsequent analysis, clarifying the degree of damage in different spatial ranges.

[0111] Step S502: The random forest machine learning algorithm is used to classify and train the damage level data, with the section damage level, train density and peak passenger flow period as input features, and the output priority ranking list is generated.

[0112] A random forest machine learning algorithm was used to classify and train the extracted damage level data. The algorithm's input features included the section damage level (quantified as 1-4, with level 1 corresponding to the highest risk), train density (the number of trains passing through the section per hour), and peak passenger flow periods (e.g., 7:00-9:00, 17:00-19:00). Through ensemble learning of multiple decision trees, the algorithm learned the correlation between features and emergency priority, outputting a priority ranking list sorted by urgency. Each entry in the list included the section mileage, priority score (0-100 points), and ranking result, ensuring that high-risk and high-impact sections were prioritized for handling.

[0113] Step S503: Generate train cancellation dispatch instructions based on the priority sorting list and the real-time train timetable; the dispatch instructions clearly indicate the cancellation order, cancellation start and end time and alternative connection scheme for each train.

[0114] Based on a priority ranking list and combined with real-time train timetables (including planned transit times, current locations, and speeds for each train), a conflict detection algorithm generates train stoppage dispatch instructions. The instructions clearly indicate the stoppage order for each train (trains involved in higher-priority sections are stopped first), the start and end times of the stoppage (accurate to the minute, e.g., 10:23-12:45), and alternative connection options (e.g., transferring to a bus at the nearest station, or having subsequent trains turn back). The generation of dispatch instructions must avoid large-scale train delays and balance safety requirements with transportation efficiency.

[0115] Step S504: Invoke the priority sorting list and shutdown dispatch instruction, associate with the preset rescue resource database, and generate a rescue resource allocation plan. The rescue resource allocation plan includes resource types and the mileage range of the corresponding allocation area. The rescue resource database includes information on engineering vehicles, repair personnel, and material reserves.

[0116] The system invokes a priority sorting list and shutdown dispatch instructions, and matches them with a pre-set rescue resource database. This database stores information such as engineering vehicles (e.g., rail repair vehicles, crane models and real-time locations), repair personnel (classified by skill level and their standby status), and material reserves (e.g., rails, connectors, inventory and storage locations). Through matching calculations between resource demand and supply capacity, a rescue resource allocation plan is generated, clearly defining the allocation area (corresponding mileage range), quantity, and allocation priority of each resource, ensuring that resources are concentrated in high-risk sections.

[0117] Step S505: Based on the rescue resource allocation plan, the shortest emergency path for each allocation area is calculated using a path optimization algorithm. The path information and resource allocation sequence are integrated to obtain an emergency response plan that includes the repair sequence, resource arrival time, and path planning.

[0118] Based on the rescue resource allocation scheme, a weighted road network topology map is first constructed, including the coordinates of reserve points, the geographical boundaries of the allocation area, and the road network attributes along the route (road grade, traffic status, restrictions, etc.). Priority lanes are assigned low weights, and restricted road sections are assigned infinite weights, with key nodes marked with traffic restrictions. An improved Dijkstra's or A* algorithm is used to dynamically avoid real-time obstacles (such as road interruptions), filter unsuitable road sections based on resource type characteristics, and introduce a time decay factor to adjust the weight of congested road sections. Based on the shortest distance, multi-objective optimization is performed by integrating travel time (distance / road speed) and safety factor (avoiding high-risk areas). A comprehensive optimal path is generated by weighted summation (distance and time each accounting for 40%, and safety factor accounting for 20%). Finally, a path scheme including passing nodes, road names, estimated time, and turning prompts is output, and its reliability is verified by comparison with historical cases. Simultaneously, by combining resource scheduling timing (such as personnel assembly time and vehicle travel time), integrating route information and time nodes, an emergency response plan is generated that includes repair sequence (segment operation sequence ordered by priority), resource arrival time (accurate to 15-minute intervals), and detailed route planning (including turning points and mileage markers), thereby achieving efficient resource delivery and orderly operation.

[0119] Specifically, damage level labels and corresponding mileage intervals for each section are extracted from the damage level sequence. Using section damage level, train density, and peak passenger flow periods as input features, a random forest machine learning algorithm is used for classification training, outputting a list sorted by emergency priority. Based on this priority list, and combined with the real-time train timetable, dispatch instructions are generated that clearly define the order of train cancellations, start and end times, and alternative connection schemes for each train. Simultaneously, the priority list and dispatch instructions are invoked, and a pre-set rescue resource database containing information on engineering vehicles, repair personnel, and material reserves is linked to generate a resource allocation scheme that clearly defines resource types and allocation area mileage ranges. Finally, a path optimization algorithm is used to calculate the shortest emergency path for each area, integrating path information with resource allocation timing to form an emergency response plan that includes repair sequence, resource arrival time, and path planning.

[0120] This embodiment utilizes a random forest algorithm to scientifically prioritize emergency responses, and combines this with real-time timetables to ensure the accuracy and feasibility of train suspension scheduling. The integration of rescue resource allocation and route optimization achieves efficient matching of resources and needs, shortening emergency response time. Multi-stage data association and algorithm optimization effectively address issues such as ambiguous priorities, unreasonable resource allocation, and inefficient route planning in traditional emergency dispatching, enabling emergency response plans to accurately match the damage needs of different sections with resource supply capacity, thus improving the systematic nature and effectiveness of high-speed rail earthquake emergency response.

[0121] In one embodiment, such as Figure 2As shown, this application also provides a linkage control device for high-speed railway earthquake emergency response equipment, which may include:

[0122] The signal purification module 601 is used to acquire multi-source vibration data and geological displacement information of key nodes along the high-speed railway, and remove train operation interference and environmental noise through filtering algorithms to obtain a purified earthquake precursor signal dataset.

[0123] The early warning determination module 602 is used to extract feature vectors from the signal dataset to determine the preliminary intensity of earthquake precursors. When the preliminary intensity exceeds a preset threshold, a warning signal is obtained by performing multi-dimensional parameter verification analysis based on the feature vectors. The feature vectors include amplitude frequency, propagation speed and signal period.

[0124] Damage rating module 603 is used to analyze wave propagation patterns in complex geological environments along the route using deep learning algorithms based on the characteristic parameters associated with early warning signals and geological data, and to integrate track displacement monitoring data and crack detection results to obtain a damage level sequence for each section.

[0125] The emergency dispatch module 604 is used to construct a priority ranking list based on the damage level sequence to guide train suspension dispatch and rescue resource allocation, thereby obtaining an optimized emergency response plan.

[0126] The aforementioned high-speed railway earthquake emergency response equipment linkage control device includes a signal purification module deployed at key nodes along the high-speed railway line. This module collects multi-source vibration data and geological displacement information, uses filtering algorithms to remove train operation interference and environmental noise, and generates a purified earthquake precursor signal dataset, providing a high-quality data foundation for subsequent analysis. The early warning judgment module extracts feature vectors such as amplitude frequency, propagation velocity, and signal period from the signal dataset, quantifies and calculates the preliminary intensity of the earthquake precursor, and generates an early warning signal containing key information after multi-dimensional parameter verification when the intensity exceeds a preset threshold. The damage rating module analyzes the seismic wave propagation pattern based on the feature parameters associated with the early warning signal and geological data, integrates track displacement monitoring data and crack detection results, and obtains a segment damage level sequence sorted by mileage. The emergency dispatch module constructs a priority ranking list based on the damage level sequence to guide train suspension scheduling and rescue resource allocation, and generates an optimized emergency response plan by combining route optimization.

[0127] In this embodiment, the various modules work closely together through data flow. The signal purification module improves the signal-to-noise ratio of the data to ensure source reliability, the early warning and judgment module enables accurate risk identification and timely early warning, the damage rating module completes a refined assessment of the degree of damage, and the emergency dispatch module optimizes resource allocation and response efficiency. This effectively solves the problems of large data interference, delayed early warning, vague damage assessment, and unreasonable resource allocation in traditional emergency responses. It achieves a highly efficient response across the entire chain, from earthquake precursor monitoring to emergency response, significantly improving the safety assurance capabilities of high-speed rail in dealing with earthquake disasters and enhancing the scientific and targeted nature of emergency response.

[0128] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0129] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the high-speed rail earthquake emergency response equipment linkage control method, device, equipment, and medium as described above.

[0130] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0131] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0132] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for coordinated control of high-speed railway earthquake emergency response equipment, characterized in that, The method includes: Multi-source vibration data and geological displacement information of key nodes along the high-speed railway were obtained, and train operation interference and environmental noise were removed by filtering algorithm to obtain a purified earthquake precursor signal dataset. Feature vectors are extracted from the signal dataset to determine the preliminary intensity of earthquake precursors. When the preliminary intensity exceeds a preset threshold, a warning signal is obtained by performing multi-dimensional parameter verification analysis based on the feature vectors. The feature vectors include amplitude frequency, propagation velocity, and signal period. The characteristic parameters associated with the warning signal and the geological data are analyzed using deep learning algorithms to determine the wave propagation mode in the complex geological environment along the route. The track displacement monitoring data and crack detection results are integrated to obtain the damage level sequence of each section. Based on the damage level sequence, a priority ranking list is constructed to guide train suspension scheduling and rescue resource allocation, resulting in an optimized emergency response plan.

2. The method according to claim 1, characterized in that, The process involves acquiring multi-source vibration data and geological displacement information from key nodes along the high-speed railway line, removing train operation interference and environmental noise through filtering algorithms, and obtaining a purified earthquake precursor signal dataset, including: Multi-source vibration data and geological displacement information are collected in real time by sensors deployed at key nodes along the high-speed railway; the key nodes include bridge bearings, tunnel entrances and exits, track beds and stress concentration areas of fault fracture zones along the line. The multi-source vibration data is denoised using a filtering algorithm to remove periodic vibration interference generated by train operation, resulting in a first vibration dataset. The filtering algorithm includes a combination of adaptive notch filtering and wavelet threshold denoising. Based on a preset frequency threshold, low-frequency components are extracted from the first vibration dataset, potential earthquake precursor signals are screened out, and a purified second vibration dataset is generated. Based on the geological displacement information, the second vibration dataset is compensated for spatial coordinates and amplitude using a preset correction model to obtain a corrected earthquake precursor signal dataset; the correction model uses the least squares method to compensate for the spatial attenuation of the vibration signal and correct the baseline drift.

3. The method according to claim 1, characterized in that, The step of extracting feature vectors from the signal dataset to determine the preliminary intensity of earthquake precursors, and when the preliminary intensity exceeds a preset threshold, performing multi-dimensional parameter verification analysis based on the feature vectors to obtain an early warning signal, includes: Feature vectors including amplitude frequency, propagation speed and signal period are extracted from the signal dataset to construct a feature matrix; The feature matrix is ​​quantified and calculated using a preset intensity assessment model to determine the preliminary intensity of earthquake precursors; the preset intensity assessment model is constructed based on a BP neural network. When the initial intensity exceeds a preset threshold, a time-domain feature set is generated based on the feature vector using time-domain analysis, and then converted into a frequency-domain feature set through Fourier transform. The frequency domain feature set is matched with a preset earthquake precursor model library to obtain the corresponding earthquake precursor model parameters. The earthquake precursor model library contains historical earthquake cases classified and stored by magnitude and focal depth. Each earthquake precursor model contains a typical frequency domain feature vector and the corresponding geological response parameters. Based on the earthquake precursor model parameters, multi-dimensional parameter verification is performed to generate a verification result set; the multi-dimensional parameter verification includes wave velocity consistency verification and amplitude attenuation law verification. The verification result set is correlated with the historical earthquake database to determine the earthquake early warning level and generate an early warning signal containing timestamp, early warning level, source characteristics and monitoring location information; the historical earthquake database contains magnitude, epicenter location, time of occurrence, peak ground acceleration and infrastructure damage records.

4. The method according to claim 3, characterized in that, The preliminary intensity of the earthquake precursor is calculated using the following formula: Where I represents the initial intensity of the earthquake precursor, standardized to a value of 0-10, with 10 corresponding to the highest risk, and σ(·) represents the Sigmoid activation function. f represents the peak amplitude frequency in the characteristic matrix, v represents the propagation speed in the characteristic matrix, and c represents the signal periodicity stability coefficient in the characteristic matrix. The value range is [0,1]. ω1, ω2, and ω3 represent weight parameters, which are determined through training with historical data. b represents the bias term, which is trained to make the output fit the preset intensity range.

5. The method according to claim 1, characterized in that, The method involves using deep learning algorithms to analyze wave propagation patterns in complex geological environments along the route, based on the characteristic parameters associated with the warning signals and geological data. This integrates track displacement monitoring data and crack detection results to obtain a damage level sequence for each section, including: Based on the characteristic parameters associated with the warning signal and geological data, an analysis model is constructed using a deep learning algorithm; the characteristic parameters include wave velocity and amplitude characteristics; the geological data includes stratigraphic lithology and fault distribution; The analytical model is used to simulate the seismic wave propagation mode under complex geological conditions along the route, calculate the vibration response value of the infrastructure, and generate a potential damage distribution map; A weighted fusion algorithm is used to integrate the potential damage distribution map, track displacement monitoring data, and crack detection results to obtain multi-source data fusion results; the track displacement monitoring data includes the lateral and longitudinal displacement of the track and its temporal change rate; the crack detection results include crack length, maximum width, and propagation rate; Based on the multi-source data fusion results, a comprehensive damage index is calculated for the high-speed rail section, and a damage level sequence is generated by classifying the levels according to the score intervals and arranging them in mileage order.

6. The method according to claim 5, characterized in that, The comprehensive damage index is calculated using the following formula: D=0.6×P+0.25×D G +0.15×D L Where D represents the comprehensive damage index of the high-speed rail section, and P represents the risk score of the section corresponding to the potential damage distribution map. G Indicates the quantitative score of orbital displacement. d h d represents the lateral displacement. v D represents the longitudinal displacement, r represents the time-series rate of change, and D L This indicates a quantitative score for crack detection. L represents the crack length, W represents the maximum width, and V represents the propagation rate.

7. The method according to claim 1, characterized in that, The step of constructing a priority ranking list based on the damage level sequence to guide train suspension scheduling and rescue resource allocation results in an optimized emergency response plan, including: Damage level data is extracted from the damage level sequence; the damage level data includes damage level labels for each segment and corresponding mileage intervals. The random forest machine learning algorithm is used to classify and train the damage level data, with the section damage level, train density and peak passenger flow period as input features, and the output priority ranking list is used. Train cancellation dispatch instructions are generated based on the priority sorting list and the real-time train timetable; the dispatch instructions clearly indicate the cancellation order, cancellation start and end time and alternative connection scheme for each train; The priority sorting list and shutdown scheduling instructions are invoked, and a preset rescue resource database is associated with them to generate a rescue resource allocation plan. The rescue resource allocation plan includes resource types and the mileage range of the corresponding allocation areas. The rescue resource database includes information on engineering vehicles, repair personnel, and material reserves. Based on the aforementioned rescue resource allocation scheme, the shortest emergency path for each allocation area is calculated using a path optimization algorithm. By integrating path information with resource allocation timing, an emergency response plan is obtained that includes repair sequence, resource arrival time, and path planning.

8. A high-speed railway earthquake emergency response equipment linkage control device, characterized in that, The device includes: The signal purification module is used to acquire multi-source vibration data and geological displacement information of key nodes along the high-speed railway, and remove train operation interference and environmental noise through filtering algorithms to obtain a purified earthquake precursor signal dataset. The early warning determination module is used to extract feature vectors from the signal dataset to determine the preliminary intensity of earthquake precursors. When the preliminary intensity exceeds a preset threshold, a multi-dimensional parameter verification analysis is performed based on the feature vectors to obtain an early warning signal. The feature vectors include amplitude frequency, propagation velocity, and signal period. The damage rating module is used to analyze the wave propagation pattern in the complex geological environment along the route using deep learning algorithms on the characteristic parameters associated with the warning signal and geological data, and to integrate track displacement monitoring data and crack detection results to obtain the damage level sequence of each section. The emergency dispatch module is used to construct a priority ranking list based on the damage level sequence to guide train suspension scheduling and rescue resource allocation, thereby obtaining an optimized emergency response plan.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images