Distributed optical fiber sound wave sensing track safety monitoring system

Abstract

The invention relates to the field of rail traffic safety monitoring, and discloses a distributed optical fiber sound wave sensing rail safety monitoring system which comprises a sensing acquisition module, a signal processing module, a feature extraction module, an event recognition module, an event positioning module and an early warning management module. The sensing acquisition module captures a track vibration signal based on a phase optical time domain reflection technology and generates time sequence data; the signal processing module carries out noise suppression and modal decomposition on the collected vibration signals and extracts abnormal signals; the feature extraction module extracts time-frequency features and envelope features of the vibration signals; the event identification module is combined with a deep learning model to classify track abnormal event types; the event positioning module determines the track position of the abnormal event; and the early warning management module generates a multi-stage early warning signal according to the identification result. The method has high precision, real-time performance, robustness and expansibility, and reliable guarantee is provided for operation safety of rail transit.

Description

Technical Field

[0001] The present invention relates to the technical field of rail transit safety monitoring, and particularly to a distributed fiber optic acoustic wave sensing rail transit safety monitoring system. Background Art

[0002] As an important part of modern urban transportation, the operation safety of rail transit is related to the convenience and safety of public travel. However, with the rapid expansion of the rail transit network, the complexity of the rail operation environment has increased significantly, and abnormal events caused by rail vibrations (such as rail loosening, illegal construction, foreign object intrusion, etc.) pose an increasingly prominent threat to rail safety. Traditional rail safety monitoring technologies mainly rely on fixed-point sensors or manual inspections. These methods have problems such as poor real-time performance, limited monitoring range, and high deployment costs, and are difficult to meet the requirements of modern rail transit for efficient and safe monitoring.

[0003] Although existing rail transit safety monitoring systems have made certain progress in some fields, such as using acceleration sensors for vibration monitoring or identifying rail intrusion events through video surveillance, there are still obvious technical defects. On the one hand, the monitoring range of point sensors is limited, making it difficult to achieve full-line coverage of the rail, and at the same time, the signal acquisition is greatly affected by environmental noise, and the monitoring accuracy is difficult to guarantee. On the other hand, traditional video surveillance technology's identification of rail vibration events relies on image processing algorithms. This method has poor applicability in environments with insufficient light such as tunnels and curves, and has insufficient monitoring ability for dynamic vibration signals. In addition, existing systems often lack functions for accurate positioning and classification of abnormal events, and cannot generate early warning information in a timely manner to guide maintenance work. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the present invention provides a distributed fiber optic acoustic wave sensing rail transit safety monitoring system, which realizes high-precision real-time monitoring of abnormal vibration events along the rail, intelligent identification and accurate positioning of event types, and generates real-time warning signals to ensure the operation safety of rail transit.

[0005] To achieve the above object, the present invention is realized through the following technical solutions: A distributed fiber optic acoustic wave sensing rail transit safety monitoring system, including: A sensing and acquisition module, used to capture the vibration signals of the optical fiber along the rail based on the phase optical time domain reflectometry technology and generate time-series vibration data; A signal processing module, used to perform denoising processing on the vibration data and decompose out abnormal vibration signals; A feature extraction module, used to extract time-frequency features and envelope features from the abnormal vibration signals; An event identification module, used to classify rail vibration events based on the extracted features and identify the event types; An event positioning module, configured to determine the specific position of an event on an orbit based on the time delay of an optical signal; An early warning management module, configured to generate an early warning signal and perform information visualization according to the identified event type and position information.

[0006] Preferably, the sensing and acquisition module includes: A pulsed light source, configured to inject high-power optical pulses along an optical fiber; A signal acquisition unit, configured to acquire the backscattered optical signal in the optical fiber; A data conversion unit, configured to convert the acquired scattered optical signal into orbit vibration data.

[0007] Preferably, the signal processing module includes: A noise suppression unit, configured to filter out the background noise in the vibration data; A modal decomposition unit, configured to decompose the vibration signal through variational mode decomposition and extract the abnormal vibration signal.

[0008] Preferably, the feature extraction module includes: A frequency decomposition unit, configured to perform multi-layer band decomposition on the abnormal vibration signal by using wavelet packet decomposition and extract the energy features corresponding to each band; An envelope analysis unit, configured to calculate the instantaneous envelope of the abnormal vibration signal based on the Hilbert transform and extract the dynamic change features of the vibration amplitude; A feature integration unit, configured to combine the extracted band energy features and the dynamic change features of the vibration amplitude into a feature vector and transmit it to the event recognition module.

[0009] Preferably, the frequency decomposition unit extracts the band energy features through the following steps: Using wavelet packet decomposition technology to decompose the abnormal vibration signal into multi-layer band signals, and each layer of band signal is calculated through a recursive formula:

[0010]

[0011] Wherein, is the decomposition signal of the th layer and the th node, and are respectively the low-pass and high-pass coefficients of the decomposition filter, represents the discrete sampling points of the time series, represents the time variable of the signal; Calculate the energy features of the decomposition signal of each band, and the calculation formula of the energy features is:

[0012] Among them, is the energy feature of the th frequency band, represents the th layer, the value of the decomposed signal of the th node at the th sampling point.

[0013] Preferably, the envelope analysis unit extracts the dynamic change characteristics of the vibration amplitude through the following steps: Perform Hilbert transform on the abnormal mode signal to calculate the instantaneous signal:

[0014] Among them, represents the Hilbert transform result of the mode signal , represents the time-domain representation of the abnormal vibration signal, represents the integration variable, represents the time variable, is a mathematical constant; Use the original signal and its Hilbert transform result to calculate the vibration envelope , and the calculation formula of the vibration envelope is:

[0015] Among them, represents the instantaneous envelope of the vibration signal.

[0016] Preferably, the event recognition module includes: A feature input unit, which is used to receive the frequency features and time-domain features output by the feature extraction module and combine them into a multi-dimensional feature vector; A deep learning classification unit, which classifies and identifies the multi-dimensional feature vector based on a convolutional neural network and a long short-term memory network; A classification output unit, which is used to generate the type label and confidence score of the track vibration event according to the deep learning classification result and transmit the result to the event location module.

[0017] Preferably, the event location module includes: A time delay calculation unit, which is used to calculate the optical signal propagation time delay difference caused by the abnormal vibration event in the optical fiber; A position calculation unit, which is used to calculate the specific position of the abnormal vibration event on the track according to the time delay difference and the refractive index of the optical fiber; An event calibration unit for mapping the calculated position to the orbital geographic coordinate system and providing the event positioning result of the orbital point position.

[0018] Preferably, the early warning management module includes: An alarm generation unit for generating multi-level alarm signals according to the event type and location information; A visualization unit for real-time displaying the type, location, and severity of orbital events; A data storage unit for recording event historical information and supporting query and analysis.

[0019] The present invention also provides a distributed fiber optic acoustic wave sensing orbital safety monitoring method, including the following steps: Collecting vibration signals along the orbit using the phase optical time domain reflectometry technique; Suppressing the noise of the collected vibration signals and extracting abnormal vibration signals through modal decomposition; Extracting frequency band energy characteristics using wavelet packet decomposition and calculating the signal envelope based on the Hilbert transform; Inputting the extracted features into a deep learning model to classify and identify the types of orbital abnormal events; Calculating the specific position of the orbital event based on the optical signal delay difference; Generating a warning signal according to the event type and location.

[0020] The present invention provides a distributed fiber optic acoustic wave sensing orbital safety monitoring system. It has the following beneficial effects: 1. The present invention uses the phase optical time domain reflectometry technique to collect the fiber optic vibration signals along the orbit in real time. Through the highly sensitive cooperation of the pulsed light source and the photodetector, it can accurately capture weak vibration changes and avoid external environmental noise interference, thus ensuring high-precision and high signal-to-noise ratio of the vibration signal collection.

[0021] 2. The present invention can effectively remove the background noise in the orbital vibration signals and separate the vibration components related to abnormal events through wavelet noise reduction and modal decomposition methods, ensuring that the processed signals have good representativeness and robustness. The design of the signal processing module can adapt to complex orbital operating environments and improve the recognition accuracy of abnormal signals.

[0022] 3. The feature extraction module of the present invention uses wavelet packet decomposition and Hilbert transform techniques to simultaneously obtain the frequency characteristics and time domain envelope characteristics of the vibration signals, forming a comprehensive feature vector. The multi-dimensional feature information can accurately characterize the characteristics of orbital abnormal vibrations and provide high-quality input data for subsequent classification and recognition.

[0023] 4. The event recognition module of the present invention adopts a deep learning algorithm combining a convolutional neural network and a long short-term memory network, which can accurately classify the types of track abnormal events. The classification model can not only extract the spatial distribution patterns of vibration signals, but also capture the dynamic changes in the time series, effectively improving the accuracy and robustness of track event recognition.

[0024] 5. Through the precise calculation and geographical mapping of the optical signal delay time, the present invention can achieve high-precision positioning of track vibration events. The event positioning module can quickly determine the specific abnormal points on the track and provide the geographical coordinate information of the track, providing key support for the rapid response and processing of abnormal events.

[0025] 6. The early warning management module of the present invention generates real-time early warning signals and classifies them according to the event recognition and positioning results. The early warning information visually presents the types, locations, and severity levels of track abnormal events through a visualization interface, supporting parallel monitoring of multiple events and subsequent query and analysis. The design of this module can significantly improve the response efficiency of track safety monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a schematic diagram of the system architecture of the present invention; Figure 2 is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0028] Please refer to the attached Figure 1 , the present invention provides a distributed fiber optic acoustic wave sensing track safety monitoring system, aiming to identify abnormal events along the track through real-time monitoring and analysis of track vibration signals and perform positioning and early warning.

[0029] As Figure 1 shown, the distributed fiber optic acoustic wave sensing track safety monitoring system of the present invention may include a sensing and acquisition module, a signal processing module, a feature extraction module, an event recognition module, an event positioning module, and an early warning management module. The following will describe each module of the system of the present invention in detail.

[0030] In this embodiment, the sensing and acquisition module is used to capture the real-time vibration signals of the optical fiber along the track based on the phase optical time domain reflectometry (Φ-OTDR), and convert the captured vibration signals into analyzable time-series vibration data. This module monitors the intensity change of the backward scattered optical signal in the optical fiber, senses the phase change of the optical signal caused by vibration, and thus obtains the track vibration information.

[0031] As an option, the sensing and acquisition module includes the following functional units: a pulsed light source unit, an optical signal receiving unit, a data conversion unit, and a time-series data output unit. Each unit is connected by a circuit to achieve signal transmission and processing.

[0032] Specifically, the pulsed light source unit is used to inject high-power pulsed light into the optical fiber laid along the track. Exemplarily, the light source can be a pulsed laser, and its wavelength range can be from 1.3μm to 1.55μm to ensure the high sensitivity and low loss characteristics of the optical fiber communication band. In a possible implementation, the time width of the pulsed light can be adjusted between 10ns and 100ns, depending on the system's spatial resolution requirements.

[0033] The optical signal receiving unit is used to receive the backward scattered optical signal caused by vibration in the optical fiber. It should be noted that the intensity of the backward scattered optical signal is directly related to the phase change caused by vibration in the optical fiber. The backward scattered signal is converted into an electrical signal by a photodetector and further transmitted to the data processing module for analysis. As an example, the photodetector can be a PIN photodiode or an avalanche photodiode to improve the signal detection sensitivity.

[0034] In a possible implementation, the phase change of the vibration signal can be expressed by the following formula:

[0035] where: represents the wavelength of the injected optical wave; represents the refractive index of the optical fiber; represents the length change of the optical fiber caused by vibration at position .

[0036] The data conversion unit is used to digitally process the electrical signals output by the photodetector and generate timing data corresponding to the track vibration signals. Specifically, the data conversion unit includes an analog-to-digital converter (ADC) and a data buffer, which are used to convert continuous analog signals into discrete digital signals and store and transmit the signals. In some embodiments, the data sampling frequency can be set from 10 kHz to 1 MHz to meet the acquisition requirements of different track vibration signals.

[0037] Exemplarily, the timing data output unit outputs the processed vibration signal data to the signal processing module. The output data may include the time, space, and intensity information of the vibration signal, and the format can be a time series point array or spectrogram data. It should be noted that in order to improve the accuracy of the data, the timing data can preliminarily suppress high-frequency noise through a filtering algorithm.

[0038] It can be understood that the sensing and acquisition module can also include various extended functions. For example: In some embodiments, in order to further improve the spatial resolution, the optical fiber can be segmented and calibrated, and the length of each segment of the optical fiber can be between 10m and 50m.

[0039] In another possible implementation, in order to ensure the stability of signal transmission, the system can add an optical amplifier at key nodes of the optical fiber to compensate for the optical loss of the signal.

[0040] It should be emphasized that the design of the sensing and acquisition module is based on the technical solution of the present invention, which can ensure the high-sensitivity capture of track vibration signals. Through the cooperation of a high-power pulsed light source and a high-precision photodetector, this module realizes the precise capture of fiber optic vibration signals along the track and provides high-quality vibration timing data for subsequent modules.

[0041] In this embodiment, the signal processing module is used to perform noise reduction processing on the vibration timing data provided by the sensing and acquisition module and decompose abnormal vibration signals, so as to provide high-quality input signals for the subsequent feature extraction module. Through precise noise suppression and signal decomposition methods, this module can effectively remove background noise and irrelevant signals in the track vibration signals and extract vibration characteristics corresponding to abnormal events.

[0042] As an option, the signal processing module includes a noise suppression unit and a modal decomposition unit, which respectively perform the tasks of noise reduction and decomposition of vibration signals. These two units work together to ensure the accuracy and robustness of signal processing.

[0043] Specifically, the noise suppression unit uses wavelet denoising method to perform multi-resolution decomposition on the vibration signal to remove high-frequency noise and background interference in the signal. In some embodiments, the process of wavelet denoising includes the following steps: First, select an appropriate wavelet basis function (such as Daubechies or Symlet), decompose the input signal, and obtain signal components at different resolutions; then, suppress the high-frequency components through a soft threshold function, and finally reconstruct the denoised signal. It should be noted that the key to wavelet denoising lies in threshold selection. In some embodiments, the threshold can be dynamically adjusted according to the signal-to-noise ratio to optimize the denoising effect.

[0044] In a possible implementation, the signal after wavelet denoising can be expressed as:

[0045] Where: represents the denoised signal; represents the coefficient after wavelet decomposition; represents the soft threshold function; represents the decomposition level.

[0046] It can be understood that wavelet denoising can not only effectively remove high-frequency noise, but also retain the low-frequency information of the signal, thus maximizing the retention of the characteristics of the track vibration signal.

[0047] The mode decomposition unit further decomposes the denoised signal to extract abnormal vibration signals. Specifically, in this embodiment, the variational mode decomposition (VMD) technology is used to decompose the signal into several modal components with narrowband characteristics. The core principle of VMD is to optimize the spectral division and gather the energy of different frequency bands of the signal into independent modes.

[0048] As an implementation, the objective function of mode decomposition can be expressed as:

[0049] Where: represents the th modal signal; represents the th central frequency of the mode; represents the total number of modes; is the unit impulse function; is the time variable.

[0050] Exemplarily, the solution process of VMD includes the following steps: First, set the initial modal signal and center frequency through spectral initialization; then, use the iterative optimization algorithm to gradually adjust the modal signal and frequency to make the spectral distribution between the modes optimal; finally, the output modal signal contains the modal components corresponding to the abnormal vibration signal.

[0051] It should be noted that in some embodiments, in order to improve the stability of modal decomposition, a penalty term can be introduced to constrain the amplitude range of the modal signal. In addition, the total number of modes can be dynamically adjusted according to the characteristics of the actual track vibration signal , to avoid over-decomposition or under-decomposition of the modes.

[0052] In one possible implementation, the abnormal signal after modal decomposition can be used for the subsequent feature extraction module. It can be understood that the low-frequency components in the modal signal usually correspond to the normal running vibration of the track, while the medium-high frequency components may be directly related to abnormal events. Therefore, through the spectral analysis of the modal signal, the accuracy of abnormal event recognition can be further improved.

[0053] As an extension, this module can also include a signal enhancement unit for amplifying weak abnormal events in the vibration signal. In some embodiments, the signal enhancement unit adopts an adaptive filtering method to dynamically adjust the parameters of the filter according to the frequency characteristics of the abnormal signal to enhance the signal strength in a specific frequency band.

[0054] In this embodiment, the feature extraction module is used to extract time-frequency features and envelope features from the abnormal vibration signal separated by the signal processing module as the feature vector input to the event recognition module. Through the processing of this module, the key features of abnormal events can be effectively extracted, providing high-precision data support for subsequent classification and recognition.

[0055] As an option, the feature extraction module includes a frequency decomposition unit and an envelope analysis unit, which are respectively used to extract the frequency features and time-domain features of the signal. These two units work together to ensure that the extracted features are representative and robust.

[0056] Specifically, the frequency decomposition unit adopts wavelet packet decomposition technology to decompose the vibration signal into multiple frequency bands. Wavelet packet decomposition is a multi-resolution analysis method that can simultaneously extract the low-frequency and high-frequency components of the signal through recursive decomposition. In one possible implementation, the calculation process of wavelet packet decomposition is as follows: First, select an appropriate wavelet basis (such as Daubechies or Symlet) to decompose the abnormal vibration signal. The recursive formula for wavelet packet decomposition is:

[0057]

[0058] Wherein: denotes the signal of the -th layer and the -th node; and are the coefficients of the low-pass filter and the high-pass filter, respectively; denotes the index of the discrete sampling point; denotes the time variable.

[0059] Through the above decomposition, signal components in different frequency bands can be obtained. It should be noted that the frequency range corresponding to each component is related to the sampling frequency of the signal and the decomposition layer. In some embodiments, the frequency range of the component can be calibrated through spectrum analysis.

[0060] After the decomposition is completed, the frequency decomposition unit calculates the energy characteristics of each frequency band signal. The calculation formula of the energy characteristics is:

[0061] Wherein: denotes the energy characteristic of the -th frequency band; denotes the value of the -th layer, the -th node at the -th sampling point.

[0062] It can be understood that the output characteristic of the frequency decomposition unit characterizes the energy distribution of the signal in different frequency ranges and provides key features in the frequency dimension for the subsequent classification model.

[0063] The envelope analysis unit is responsible for extracting the time-domain characteristics of the abnormal vibration signal. Specifically, in this embodiment, the envelope analysis unit calculates the instantaneous envelope of the signal through Hilbert transform. The definition formula of Hilbert transform is:

[0064] Wherein: denotes the abnormal vibration signal; denotes the result of the Hilbert transform of the signal; and are the time variable and the integral variable, respectively; is a mathematical constant.

[0065] Through the original signal and the Hilbert transform result , calculate the instantaneous envelope of the signal:

[0066] Where: represents the instantaneous envelope signal.

[0067] It should be noted that the instantaneous envelope describes the variation characteristics of the amplitude of the vibration signal over time. As a possible implementation, the output of the envelope analysis unit can be the statistical characteristics (such as mean, peak, and rate of change) of the envelope signal to enhance the representativeness of the time-domain characteristics.

[0068] Exemplarily, the output characteristics of the frequency decomposition unit and the envelope analysis unit can be combined into a feature vector , where: represents the frequency feature; represents the time-domain feature.

[0069] It can be understood that the feature vector of the feature extraction module provides rich time-frequency information for the event recognition module, thereby improving the accuracy of abnormal event classification.

[0070] To meet the requirements of different track environments, the feature extraction module can also be extended. For example: In some embodiments, the feature extraction module can introduce an adaptive feature selection algorithm to dynamically adjust the extracted feature set according to the statistical characteristics of the track vibration signal.

[0071] In another possible implementation, the frequency-domain and time-domain features can be fused with other signal features (such as phase features or spatial features) to further enhance the representativeness of the feature vector.

[0072] In this embodiment, the event recognition module is used to analyze the feature vector output by the feature extraction module to identify the type of track vibration event. By introducing deep learning technology, the event recognition module can accurately classify track abnormal events (such as track loosening, illegal construction, foreign object intrusion, etc.) and output the event type and confidence level.

[0073] As an option, the event recognition module includes a feature input unit, a deep learning classification unit, and a classification output unit. Each unit works together, and the complete recognition process is formed from the input of the feature vector to the output of the classification result.

[0074] Specifically, the feature input unit is used to receive the multi-dimensional feature vectors generated by the feature extraction module It should be noted that the dimension and quantity of the input features can be dynamically adjusted according to the complexity of the track vibration signal.

[0075] The deep learning classification unit adopts an architecture that combines a convolutional neural network (CNN) and a long short-term memory network (LSTM) to classify and identify the feature vectors. As an example, the convolutional neural network is used to extract the spatial distribution pattern of the feature vectors, and its convolution calculation formula is as follows:

[0076] Where: represents the value of the feature map after convolution; represents the weight of the convolution kernel; represents the value at the corresponding position in the input feature map; is the bias; is the activation function (such as ReLU).

[0077] It can be understood that the role of the convolutional neural network is to extract local features from the multi-dimensional feature vectors, thereby improving the accuracy of event classification.

[0078] In a possible implementation, the long short-term memory network (LSTM) is used to capture the time series pattern of the feature vectors. The state update formula of the LSTM is as follows:

[0079] Where: represents the hidden state at the current moment; represents the hidden state at the previous moment; represents the current input feature; represents the weight matrix; is the bias; is the activation function (such as Sigmoid).

[0080] It should be noted that through the time series modeling of the LSTM, the dynamic changes of the track vibration signal can be captured, thereby improving the ability to judge the event type.

[0081] The classification output unit is used to generate event type labels and confidence scores based on the results of the deep learning model. As an option, the event type labels can include common event types such as track loosening, illegal construction, foreign object intrusion, etc., and the confidence score is used to quantify the reliability of the classification result. In some embodiments, the classification output unit can use the Softmax function to calculate the probability of each event type:

[0082] Where: represents the feature vector belongs to the event type probability; represents the original prediction value of the deep learning model for the event type

[0083] In a possible implementation, the classification output unit can also set an alarm threshold based on the confidence score. For example, when the confidence score of a certain event type exceeds the set value, an alarm signal is triggered and transmitted to the early warning management module.

[0084] It can be understood that the design focus of the event recognition module is on accuracy and real-time performance. To further improve the classification accuracy, in some embodiments, the deep learning classification unit can be extended. For example, multi-model integration technology can be introduced to perform weighted averaging on the prediction results of multiple classification models to obtain a more stable classification result.

[0085] In this embodiment, the event location module is used to determine the specific location of the track vibration event in the optical fiber based on the time delay signal collected by the fiber optic sensing, and map this location to the geographic coordinate system of the track to provide accurate spatial information for subsequent early warning and management. Through precise time delay calculation and geographic mapping, this module can achieve real-time location of abnormal events.

[0086] As an option, the event location module includes a time delay calculation unit, a position calculation unit, and a geographic mapping unit, and each unit works in coordination to gradually complete the calibration of the track position of the event.

[0087] Specifically, the time delay calculation unit is used to calculate the time delay difference of the optical signal propagation caused by the vibration event in the optical fiber. Exemplarily, the calculation formula of the time delay difference is as follows:

[0088] Where: represents the time when the backscattered optical signal reaches the optical detector; ​Represents the time when the pulsed optical signal is emitted from the light source; Represents the propagation time delay of the optical signal.

[0089] It should be noted that the time delay difference is a key characteristic of the optical signal propagating in the optical fiber, and it directly reflects the change in the length of the optical signal propagation path. As a possible implementation, the time delay calculation unit can accurately capture and through a high-speed data acquisition card to ensure high-precision delay calculation.

[0090] The position calculation unit calculates the specific position of the vibration event in the optical fiber based on the time delay difference and the physical characteristics of the optical fiber. Specifically, the event position can be calculated by the following formula:

[0091] Where: Represents the distance of the event in the optical fiber; Is the speed of light; Is the refractive index of the optical fiber; Is the time delay difference of the optical signal propagation.

[0092] It can be understood that since the optical signal needs to propagate back and forth in the optical fiber and the path length is , so the denominator in the formula contains a factor of 2. It should be noted that the refractive index of the optical fiber usually depends on the material of the optical fiber (for example, the of silica optical fiber is usually around 1.45), and can be obtained through experimental calibration.

[0093] In a possible implementation, in order to further improve the positioning accuracy, the position calculation unit can use an interpolation algorithm to refine the time delay data. For example, linear interpolation or cubic spline interpolation is used to fit the delay change of the optical signal, so as to generate higher-resolution position information between the sampling points.

[0094] The geographic mapping unit is used to map the calculated optical fiber position to the geographic coordinate system of the track to generate the event positioning result of the track point position. Specifically, the geographic mapping unit can convert the length coordinates in the optical fiber into the actual geographic coordinates of the track based on the segmented calibration information of the optical fiber layout. In some embodiments, the segmented length of the optical fiber can be between 10 meters and 50 meters, and each segment of the optical fiber is associated with the track position through calibration data.

[0095] It is understandable that in a complex track environment, such as a tunnel or a bend area, there may be a deviation between the laying path of the optical fiber and the actual path of the track. For this reason, the geographical mapping unit can introduce a correction algorithm to adjust the positioning result according to the geometric shape of the track. For example, the position of the optical fiber can be calibrated using the CAD drawings of the track or the GPS positioning data to ensure the accuracy of the positioning result.

[0096] In this embodiment, the early warning management module is used to generate an early warning signal based on the event type output by the event recognition module and the position information provided by the event positioning module, and visually display and manage the early warning information. Through this module, the track abnormal event can be presented to the monitoring personnel in an intuitive way, and corresponding response suggestions can be provided to ensure the safety of track operation.

[0097] As an option, the early warning management module includes an alarm generation unit, an information visualization unit, and a data storage unit. Each unit cooperates through a data interface to achieve the real-time nature of the early warning and the traceability of the information.

[0098] Specifically, the alarm generation unit generates corresponding multi-level early warning signals according to the event type and confidence level provided by the event recognition module, and the track position provided by the event positioning module. Exemplarily, the multi-level early warning signals can be divided into low-level, medium-level, and high-level early warnings. The classification of the early warning can be based on the following logic: If the confidence score of the event recognition module is greater than the set threshold (such as 0.8), and the event type is a high-risk event (such as track loosening or illegal construction), then a high-level early warning is triggered; If the confidence score is between 0.5 and 0.8, and the event type is a medium-risk event, then a medium-level early warning is triggered; If the confidence score is less than 0.5, or the event type is a low-risk event (such as small foreign object intrusion), then a low-level early warning is triggered.

[0099] It should be noted that the alarm generation unit can also generate a positioning early warning signal in combination with the position information provided by the event positioning module. For example, in the case of a high-level early warning, the specific position and geographical coordinates of the track can be marked in the early warning signal to quickly locate the abnormal event.

[0100] The information visualization unit is used to display the early warning signal and related information in real time on the monitoring platform. As a possible implementation method, the visualization interface includes the display of the event type, position information, and early warning level. Specifically: The event type can be displayed through a text label (such as "track loosening" or "illegal construction"); The position information can be intuitively displayed through a marked point on the track plan; The warning levels can be distinguished by color markings (e.g., red for high-level warnings, yellow for medium-level warnings, and green for low-level warnings).

[0101] It can be understood that, in order to improve the monitoring efficiency, the information visualization unit can also support the function of simultaneously displaying multiple events. In some embodiments, the system can form a heat map of track anomaly events by superimposing the track position markings of different events, thereby helping the monitoring personnel quickly identify the areas with concentrated risks.

[0102] The data storage unit is used to record all warning information, including event type, location information, warning level, and timestamp, etc. This information is stored in the database in the form of logs, supporting subsequent queries and analyses. For example, all high-level warning records within a certain time period can be queried through the retrieval function to analyze the safety status of track operation.

[0103] As an extended solution, the warning management module can also be docked with an external response system. In some embodiments, when a high-level warning is triggered, the system can automatically send an alarm message to the track maintenance personnel, along with the event type and positioning data. In addition, the system can also link with the train operation dispatching system to temporarily limit the train operation speed in high-risk areas to reduce the probability of safety accidents.

[0104] In a possible implementation manner, in order to improve the reliability of warning signals, the warning management module can also introduce multi-signal fusion technology. For example, the intensity information of vibration signals and the confidence score of event classification can be combined to calculate a comprehensive warning index :

[0105] Where: represents the intensity of the vibration signal (such as the maximum value of the envelope); represents the confidence score of event classification; and are weight coefficients, which can be adjusted according to actual needs.

[0106] Generally speaking, the present invention combines the phase optical time domain reflectometry (Φ-OTDR) with deep learning algorithms to achieve high-precision real-time monitoring of vibration signals along the track and identification of abnormal events. The present invention can accurately capture abnormal signals in track vibrations, identify various abnormal events such as track loosening, illegal construction, and foreign object intrusion, and provide real-time warnings and positioning, providing reliable guarantees for the operation safety of rail transit, and having the characteristics of high precision, real-time performance, and wide adaptability.

[0107] Correspondingly, please refer to the appendixFigure 2 , the present invention also provides a distributed fiber optic acoustic wave sensing track safety monitoring method, including the following steps: S1. Collect vibration signals along the track using the phase optical time domain reflectometry technique; First, the method is based on the phase optical time domain reflectometry technique. The vibration signals of the optical fiber along the track are captured by the sensing acquisition module. The pulsed light source injects high-power pulsed light into the optical fiber arranged on the track. The vibration in the optical fiber will cause changes in the backscattered light signal. The optical signal is converted into an electrical signal by the photodetector, and the timing data corresponding to the vibration signal is generated.

[0108] S2. Suppress the noise of the collected vibration signals and extract abnormal vibration signals through modal decomposition; The collected vibration signals are subjected to noise suppression and decomposition by the signal processing module. Through the multi-resolution noise reduction algorithm, the background noise is removed and the abnormal vibration information is retained; subsequently, the modal decomposition method is used to decompose the signal into several modal components, and the vibration signals related to abnormal events are extracted.

[0109] S3. Extract the frequency band energy characteristics using wavelet packet decomposition and calculate the signal envelope based on the Hilbert transform; The processed abnormal vibration signals enter the feature extraction module to further extract time-frequency characteristics and envelope characteristics. Using the wavelet packet decomposition technique, the energy distribution of the signal in different frequency bands is calculated to generate frequency characteristics; at the same time, based on the envelope analysis technique, the time-domain characteristics of the signal are obtained to form a complete feature vector as the input data for the subsequent classification model.

[0110] S4. Input the extracted features into the deep learning model to classify and identify the types of track abnormal events; The feature vector is input into the deep learning classification model of the event recognition module to identify the types of track abnormal events. The classification model combines the convolutional neural network and the long short-term memory network to extract the spatial distribution pattern and time series pattern of the features respectively, and finally outputs the event type and confidence level. The classification result can distinguish various abnormal events such as track loosening, illegal construction, and foreign object intrusion.

[0111] S5. Calculate the specific location of the track event based on the optical signal delay difference; After classification, the event positioning module calculates the specific location of the event in the track using the propagation time delay difference of the optical signal. By mapping the optical fiber length coordinates to the geographic coordinate system of the track, the spatial positioning information of the event is generated, providing an accurate location basis for the subsequent warning signal.

[0112] S6. Generate a warning signal according to the event type and location; Finally, the early warning management module generates early warning signals based on the classified event types and location information, and displays the relevant information through the visualization platform. The hierarchical processing of early warning signals can intuitively reflect the degree of danger of events and support subsequent response decisions. At the same time, all early warning information is recorded in the database for subsequent queries and statistical analysis of the track safety status.

[0113] Through the collaborative work of data collection, processing, feature extraction, classification and recognition, location and early warning, this method constructs an efficient and accurate track safety monitoring process, which has the significant advantages of strong real-time performance and high robustness, providing reliable guarantee for track traffic safety.

[0114] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Distributed fiber optic acoustic wave sensing rail safety monitoring system, characterized in that: include: The sensing acquisition module is used to capture the vibration signal of the optical fiber along the track based on the phase optical time domain reflection technology and generate time series vibration data; A signal processing module, used for performing denoising on the vibration data and decomposing abnormal vibration signals; A feature extraction module, used to extract time-frequency features and envelope features from the abnormal vibration signal; An event recognition module is used to classify rail vibration events based on the extracted features and identify event types; An event location module, used to determine the specific location of the event on the track based on the time delay of the optical signal; The warning management module is used to generate warning signals and visualize information based on the identified event type and location information.

2. The distributed optical fiber acoustic wave sensing rail safety monitoring system according to claim 1 is characterized in that: The sensor acquisition module comprises: A pulse light source for injecting high-power light pulses along the optical fiber; A signal collection unit, used for collecting backscattered light signals in the optical fiber; The data conversion unit is used to convert the collected scattered light signals into orbital vibration data.

3. The distributed optical fiber acoustic wave sensing rail safety monitoring system according to claim 1 is characterized in that: The signal processing module comprises: A noise suppression unit for filtering out background noise in vibration data; The modal decomposition unit is used to decompose the vibration signal through variational mode decomposition and extract the abnormal vibration signal.

4. The distributed optical fiber acoustic wave sensing rail safety monitoring system according to claim 1 is characterized in that: The feature extraction module comprises: A frequency decomposition unit, used to perform multi-layer frequency band decomposition of abnormal vibration signals by using wavelet packet decomposition, and extract energy characteristics corresponding to each frequency band; An envelope analysis unit, used to calculate the instantaneous envelope of the abnormal vibration signal based on Hilbert transform and extract the dynamic change characteristics of the vibration amplitude; The feature integration unit is used to combine the extracted frequency band energy features and vibration amplitude dynamic change features into a feature vector and transmit it to the event recognition module.

5. The distributed optical fiber acoustic wave sensing rail safety monitoring system according to claim 4 is characterized in that: The frequency decomposition unit extracts the frequency band energy features by the following steps: The abnormal vibration signal is decomposed into multiple frequency band signals using wavelet packet decomposition technology. Each frequency band signal is calculated using the recursive formula: ， ， in, For the Layer, The decomposition signal of each node, and are the low-pass and high-pass coefficients of the decomposition filter, respectively. represents the discrete sampling points of the time series, A time variable representing a signal; The energy characteristics of the decomposed signal of each frequency band are calculated. The calculation formula of the energy characteristics is: ， in, For the The energy characteristics of the frequency band, Indicates Layer, The node decomposition signal is The value at the sampling point.

6. The distributed optical fiber acoustic wave sensing rail safety monitoring system according to claim 4 is characterized in that: The envelope analysis unit extracts the dynamic change characteristics of the vibration amplitude through the following steps: For abnormal modal signals Perform Hilbert transform and calculate the instantaneous signal: ， in, Represents the modal signal The Hilbert transform result is: The time domain representation of abnormal vibration signal, represents the integral variable, represents the time variable, is a mathematical constant; Use original signal And its Hilbert transform result Calculating the vibration envelope , the calculation formula of the vibration envelope is: ， in, Represents the instantaneous envelope of the vibration signal.

7. The distributed optical fiber acoustic wave sensing rail safety monitoring system according to claim 1 is characterized in that: The event recognition module comprises: A feature input unit, used to receive the frequency features and time domain features output by the feature extraction module, and combine them into a multi-dimensional feature vector; Deep learning classification unit, which classifies and recognizes multi-dimensional feature vectors based on convolutional neural networks and long short-term memory networks; The classification output unit is used to generate the type label and confidence score of the rail vibration event based on the deep learning classification results, and pass the results to the event location module.

8. The distributed optical fiber acoustic wave sensing rail safety monitoring system according to claim 1 is characterized in that: The event location module includes: A time delay calculation unit, used to calculate the time delay difference of optical signal propagation caused by an abnormal vibration event in the optical fiber; A position calculation unit, used to calculate the specific position of the abnormal vibration event on the track according to the time delay difference and the refractive index of the optical fiber; The event calibration unit is used to map the calculated position to the orbital geographic coordinate system and provide event positioning results for orbital points.

9. The distributed optical fiber acoustic wave sensing rail safety monitoring system according to claim 1 is characterized in that: The early warning management module includes: An alarm generating unit, used for generating a multi-level alarm signal according to the event type and location information; A visualization unit to display the type, location and severity of track events in real time; Data storage unit, used to record event history information and support query and analysis.

10. A distributed optical fiber acoustic wave sensing rail safety monitoring method, applied to a system as claimed in any one of claims 1 to 9, characterized in that: The following steps are involved: The vibration signals along the track are collected using phase optical time domain reflectometry technology; Suppress the noise of the collected vibration signals and extract abnormal vibration signals through modal decomposition; The frequency band energy characteristics are extracted by wavelet packet decomposition, and the signal envelope is calculated based on Hilbert transform; The extracted features are input into the deep learning model to classify and identify the types of track anomaly events; Calculate the specific position of the orbital event based on the delay difference of the optical signal; Generate early warning signals based on event type and location.

Images