Real-time monitoring method and sensing system for catenary cable state based on distributed optical fiber sensing

Abstract

The application discloses a real-time monitoring method and a sensing system for a catenary cable state based on distributed optical fiber sensing, and the Mach-Zehnder interference and phi-OTDR fusion type distributed optical fiber sensing system comprises a first light source, a second light source, a Mach-Zehnder interferometer, a phi-OTDR and a data acquisition and processing module; Raman amplifiers are arranged at both ends of the Mach-Zehnder interferometer, so that the sensing distance is prolonged; the optical fiber sensing system extracts the phase-modulated interference light signal and the back Rayleigh scattering light signal corresponding to the fault event when the fault event occurs; then, the extracted back Rayleigh scattering light signal is processed to obtain the intensity change information of the fault event, and the position of the fault event is demodulated; finally, the two-dimensional optical signal is sent into a pre-trained Swin Transformer network model to extract the time domain feature, the frequency domain feature and the spatial domain feature, and the type of the fault event is identified through a classifier.

Description

Technical Field

[0001] This invention belongs to the field of sensing and intelligent detection, and specifically relates to a method based on Mach-Zehnder interferometry and... A real-time monitoring and sensing system for the status of contact network cables using integrated distributed fiber optic sensing. Background Technology

[0002] Currently, the monitoring of the overhead contact system's operational status mainly relies on regular and irregular manual inspections, vehicle inspections, and testing vehicle checks. This approach cannot achieve continuous recording and monitoring of operational status and technical parameters, nor can it detect and eliminate problems in the overhead contact system's operational status in real time. The existing 6C system of the railway also lacks real-time online monitoring and detection capabilities, primarily relying on high-definition imaging from cameras. Later, it requires a large number of experienced engineers to manually interpret the images, resulting in a huge workload, low efficiency, and relatively limited data on the overhead contact system's operational status, thus hindering improvements in the safety, reliability, and maintenance efficiency of the overhead contact system.

[0003] In recent years, cases of overhead contact line faults affecting train operation have occurred frequently. In particular, auxiliary conductors such as lightning protection wires and overhead ground wires are directly connected to the integrated grounding system. Faults such as wire breaks and detachments are difficult for traditional protection systems to identify and locate. Delayed fault diagnosis and handling can expand the scope of the fault and even affect train safety, significantly impacting train operation efficiency, safety, and emergency repair. Cable faults are characterized by difficulty in early detection, high randomness, and difficulty in fault location, directly affecting the reliability of the overhead contact line system. Real-time monitoring of the auxiliary conductors' operating status, timely identification of fault information, and implementation of appropriate maintenance strategies based on fault type are crucial for improving the overall operational safety of the railway overhead contact line and preventing escalation of accidents. Currently, research on intelligent monitoring or fault location of transmission lines using optical fiber technology has begun in China. Tong Tangli et al. (Tong Tangli, Gao Yan, Wang Pengfei et al. A method for monitoring the galloping of overhead lines based on distributed optical fiber sensing technology combined with Wiener filtering [J]. Power Grid and Clean Energy, 2022, 38(09):17-24.) proposed a long-distance method for monitoring the galloping of overhead lines based on distributed optical fiber sensing technology combined with Wiener filtering, which is resistant to electromagnetic interference. Wang Yougang (Wang Yougang. Research on distributed stress monitoring model of optical fiber composite overhead phase line based on LSTM [J]. Power Big Data, 2023, 26(11):33-40.DOI:10.19317 / j.cnki.1008-083x.2023.11.004.) proposed a distributed stress detection method for optical fiber composite overhead phase line based on LSTM algorithm. Zhang Qi (Zhang Qi. Fault Location and Analysis of Power Overhead Conductors and Optical Cables Based on Brillouin Fiber Sensing [D]. Harbin University of Science and Technology, 2022. DOI:10.27063 / d.cnki.ghlgu.2022.000677.) Addressing the technical bottlenecks and limitations in the current maintenance of power overhead conductors, this paper proposes a fault location and analysis method for power overhead conductors and optical cables based on Brillouin fiber optic sensing.

[0004] Although research has begun on intelligent monitoring of transmission lines, its application scenarios, functional positioning, and hierarchical architecture differ significantly from those of overhead contact lines, making it difficult to meet the complex operating conditions of railways. Therefore, it is necessary to develop and design a distributed fiber optic sensing cable condition monitoring system for railway overhead contact lines in challenging and complex environments. This system should feature a simple structure, real-time monitoring capabilities, high identification accuracy, and high positioning precision. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a real-time monitoring and sensing system for contact network cable status based on integrated distributed optical fiber sensing. This system aims to address the monitoring, identification, and location of potential issues such as wire breakage, loosening, lightning strikes, and abnormal temperature rises in contact network conductors under complex environments and operating conditions. In long-distance, large-area contact network cable status monitoring, there are problems with rapid and accurate identification, precise location, and poor stability. This invention is based on Mach-Zehnder interferometry and phase-sensitive optical time-domain reflectometry (hereinafter referred to as...). This integrated distributed fiber optic sensing system continuously monitors the status of contact network cables using Mach-Zehnder interferometry sensing, exhibiting a wide frequency response range. When a fault is detected, it simultaneously utilizes Mach-Zehnder interferometry and... Identify the light signals and then... The location information can be demodulated, and multi-point positioning can be achieved, thus realizing the complementary advantages of the two sensing systems mentioned above.

[0006] A Machzonde interference and A converged distributed fiber optic sensing system includes a first light source, a second light source, a Mach-Zehnder interferometer, and... and data acquisition and processing module;

[0007] The first light source is connected in sequence to the first fiber optic attenuator, the first fiber optic coupler, and the input end of the Mach-Zehnder interferometer. The output end of the Mach-Zehnder interferometer is connected in sequence to the first fiber optic coupler, the first photodetector, and the data acquisition and processing module.

[0008] The second light source is connected in sequence to the second fiber optic attenuator and the semiconductor optical amplifier. and a data acquisition and processing module; the output of the semiconductor optical amplifier is respectively connected to the data acquisition and processing module and The pulsed electrical signal used for transmission to the data acquisition and processing module serves as a trigger signal, and the pulsed light is also used for transmission to... The It includes a pulsed light amplifier, a fiber optic circulator, a Mach-Zehnder interferometer, and a second optical filter connected in sequence. The second optical filter is connected in sequence to a second photodetector and a data acquisition and processing module.

[0009] The laser from the second light source enters the Mach-Zehnder interferometer via an optical fiber circulator and a dense wavelength division multiplexer, and is coupled together with a laser from the first light source into the sensing fiber.

[0010] Furthermore, the input end of the Mach-Zehnder interferometer is equipped with a first Raman amplifier, and the output end of the Mach-Zehnder interferometer is equipped with a second Raman amplifier, thereby extending the sensing distance.

[0011] The data acquisition and processing module includes a signal extraction and preprocessing submodule, a positioning submodule, and a fault identification submodule;

[0012] The signal extraction and preprocessing submodule is used to extract the Mach-Zehnder interferometry and... The phase-modulated fault event corresponding interference optical signal and backscattered Rayleigh optical signal of the integrated distributed optical fiber sensing subsystem are obtained; then the extracted optical signals are preprocessed, namely, the interference optical signal is subjected to endpoint detection and the backscattered Rayleigh optical signal is processed by the moving average algorithm.

[0013] The positioning submodule is used to perform interpolation on the backscattered Rayleigh light signal to demodulate the location of the fault event.

[0014] The fault identification submodule is used to convert the interference light signal and the backscattered Rayleigh light signal preprocessed by the signal extraction and preprocessing submodule into a two-dimensional time spectrum through wavelet transform. Then, the two-dimensional time spectrum is fed into the pre-trained Swin Transformer network model to extract the time domain features, frequency domain features and spatial domain features of the light signal, and the type of fault event is identified by a classifier.

[0015] Furthermore, the splitting ratio of the first fiber coupler and the second fiber coupler is 50:50, the wavelength of the first fiber coupler is 1550nm, and the wavelength of the second fiber coupler is 1455nm.

[0016] Furthermore, the output light of the first Raman amplifier and the second Raman amplifier is in the 1455nm band.

[0017] Furthermore, the center wavelength of the first optical filter is the center wavelength of the first light source λ1±0.2nm, and only the laser emitted by the first light source can pass through, which is used to filter out the pulsed light from the second light source.

[0018] Furthermore, the first and second light sources are narrowband continuous light lasers with a working wavelength of 1550nm and a maximum output power of 10mW.

[0019] Furthermore, the pre-set fault events include disconnection, loosening, lightning strike, abnormal temperature rise, and no fault.

[0020] A method for real-time monitoring of contact network cable status based on distributed optical fiber sensing, the method utilizing the Mach-Zehnder interferometry and... The integrated distributed fiber optic sensing system includes:

[0021] Step 1: Real-time detection of interference light signals. When an abnormal condition occurs in the contact network cable, the sensing light signal changes, and the phase-modulated interference light signal and backscattered Rayleigh light signal are acquired.

[0022] Step 2: Remove the DC component from the interference optical signal obtained from the fault event, perform endpoint detection, and obtain the interference optical signal corresponding to the fault event;

[0023] The backscattered Rayleigh light signal obtained from the fault event is subjected to a moving average algorithm and difference calculation to obtain the obvious intensity change information caused by the fault event, and the location of the fault event is demodulated at the same time.

[0024] Step 3: Perform wavelet transform on the interference light signal and the backscattered Rayleigh light signal processed in Step 2 to convert them into a two-dimensional time spectrum. Then, feed the two-dimensional time spectrum into the pre-trained Swin Transformer network model to extract the time domain features, frequency domain features and spatial domain features of the light signal. The type of fault event is identified by a classifier.

[0025] Step one specifically includes:

[0026] When an abnormal condition occurs in the contact network cable, it will cause a change in the sensor optical signal; the Mach-Zehnder interferometer and In the integrated distributed optical fiber sensing system, the signal from the first light source passes through the first optical fiber attenuator, the first optical fiber coupler, the Mach-Zehnder interferometer, and the first optical fiber coupler in sequence to obtain the phase-modulated interference light signal. After passing through the first photodetector, it is sent to the data acquisition and processing module to realize the feature extraction of the modulated light signal.

[0027] When an abnormality occurs in the contact wire cable, it is simultaneously detected. The backscattered Rayleigh light signal from the second light source passes sequentially through a second fiber attenuator, a semiconductor optical amplifier, a pulsed optical amplifier, a fiber circulator, and a dense wavelength division multiplexer before entering the Mach-Zehnder interferometer and being transmitted within the sensing fiber. The resulting continuous backscattered Rayleigh light signal then passes through a fiber circulator and a second optical filter to acquire the backscattered Rayleigh light signal. After passing through a second photodetector, it is sent to the data acquisition and processing module to achieve feature extraction of the backscattered Rayleigh light signal.

[0028] Step two, the processing of the backscattered Rayleigh light signal, includes:

[0029] A fixed backscattering Rayleigh curve is selected as the reference curve for the interpolation operation, while the Mach-Zehnder interferometry and... The backscattered Rayleigh signal curves acquired by the fusion distributed fiber optic sensing system at other times are successively compared with the reference curve to obtain the obvious intensity change information caused by the fault event. The location information of the fault event is obtained by detecting the change in echo intensity and its delay time.

[0030] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0031] This invention employs bidirectional distributed Raman amplification to uniformly amplify the transmitted light along the optical fiber, effectively extending the sensing distance and improving the signal-to-noise ratio of the far-end sensing optical signal.

[0032] right The sensor optical signals are processed by algorithms to improve the location accuracy of fault events;

[0033] This invention integrates Mach-Zehnder interferometer and The system extracts features from the sensor's optical signals and converts the optical signals into two-dimensional time-frequency signals to extract features from fault events. This effectively enhances the deep learning model's ability to extract the essential spatiotemporal features of fault events, thereby achieving better event recognition and detection performance, increasing pattern recognition accuracy, and further reducing the system's false alarm rate. Moreover, the Swing Transformer network model used can maintain a fast recognition speed while ensuring a high recognition rate, meeting the needs of actual detection. Attached Figure Description

[0034] Figure 1 The flowchart of the real-time monitoring method for contact network cable status based on distributed optical fiber sensing of the present invention is shown.

[0035] Figure 2 This includes Machzender interference and Schematic diagram of a converged distributed fiber optic sensing system;

[0036] Figure 3 This is a flowchart of the network model structure of the Swing Transformer used in this invention.

[0037] In the picture:

[0038] 1: First light source; 2: Second light source; 3: First fiber optic attenuator

[0039] 4, 13: First fiber coupler; 5: First Raman amplifier; 6: Second Raman amplifier

[0040] 7, 8: Second fiber optic coupler; 9: First wavelength division multiplexer

[0041] 10: Second Wavelength Division Multiplexer 11: Third Wavelength Division Multiplexer 12: Fourth Wavelength Division Multiplexer

[0042] 14: Second fiber optic attenuator; 15: Semiconductor optical amplifier

[0043] 16: Pulsed optical amplifier; 17: Fiber optic circulator; 18: Dense wavelength division multiplexer.

[0044] 19: First optical filter; 20: Second optical filter; 21: First photodetector

[0045] 22: Second photodetector; 23: Data acquisition card; 24: Industrial control computer Detailed Implementation

[0046] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The specific embodiments described are only for explanation and illustration of the present invention and are not intended to limit the present invention.

[0047] like Figure 1 As shown, a real-time monitoring method for the status of contact network cables based on distributed optical fiber sensing is disclosed. This real-time monitoring method utilizes a Mach-Zehnder interferometer and... The integrated distributed fiber optic sensing system, wherein the Mach-Zehnder interferometer and The integrated distributed fiber optic sensing system is pre-programmed with five fault events: wire breakage, loosening, lightning strike, abnormal temperature rise, and no fault. It can issue alarms based on the severity of the fault events, reminding relevant personnel to intervene in a timely manner and ensuring the safety of railway operations.

[0048] like Figure 2 As shown, the Mach-Zehnder interferometry and phase-sensitive optical time-domain reflectometry (hereinafter referred to as...) The integrated distributed fiber optic sensing system includes a first light source 1, a second light source 2, a Mach-Zehnder interferometer, and... And data acquisition and processing module.

[0049] Figure 2 The schematic diagram of the integrated distributed optical fiber sensing subsystem is shown, including a first light source 1, a second light source 2, a first optical fiber attenuator 3, first optical fiber couplers 4 and 13, a first Raman amplifier 5, a second Raman amplifier 6, second optical fiber couplers 7 and 8, a first wavelength division multiplexer 9, a second wavelength division multiplexer 10, a third wavelength division multiplexer 11, a fourth wavelength division multiplexer 12, a second optical fiber attenuator 14, a semiconductor optical amplifier 15, a pulsed optical amplifier 16, an optical fiber circulator 17, a dense wavelength division multiplexer 18, a first optical filter 19, a second optical filter 20, a first photodetector 21, a second photodetector 22, and a data acquisition and processing module. The data acquisition and processing module includes a connected data acquisition card 23 and an industrial control computer 24. The first light source 1 and the second light source 2 are narrowband continuous light lasers with center wavelengths of λ1 (1550.12nm) and λ2 (1550.92nm), respectively, and a maximum output power of 10mW. The data acquisition card 23 is used to acquire the sensing electrical signals transmitted back from the first and second photodetectors 21 and 22. The industrial control computer 24 is used to analyze and demodulate the received signals to obtain the location of the fault and identify the fault event. The wavelength of the first fiber coupler 4 and 13 is 1550nm, and the wavelength of the second fiber coupler 7 and 8 is 1455nm.

[0050] All fiber optic couplers mentioned in this article have a splitting ratio of 50:50.

[0051] The first fiber optic attenuator 3 has an input end and an output end. Its input end is connected to the first light source 1, and its output end is connected to the first fiber optic coupler 4.

[0052] The first fiber coupler 4 has one input end and two output ends. Its input end is connected to the output end of the first fiber attenuator 3. The attenuated optical signal is split into two optical signals with the same power by the first fiber coupler 4 and transmitted to the first wavelength division multiplexer 9 and the dense wavelength division multiplexer 18.

[0053] The first fiber optic coupler 13 has two input terminals and one output terminal. Its output terminal is connected to the first photodetector 21, and its two input terminals are connected to the third wavelength division multiplexer 11 and the first optical filter 19, respectively. The center wavelength of the first optical filter 19 is λ1±0.2nm, and only the laser light emitted by the first light source 1 can pass through, which is used to filter out the pulse light from the second light source.

[0054] The input end of the second fiber coupler 7 is connected to the first Raman amplifier 5, and the two output ends are connected to the first wavelength division multiplexer 9 and the second wavelength division multiplexer 10, respectively; the input end of the second fiber coupler 8 is connected to the second Raman amplifier 6, and the two output ends are connected to the third wavelength division multiplexer 11 and the fourth wavelength division multiplexer 12, respectively; and the output light of the first Raman amplifier 5 and the second Raman amplifier 6 is in the 1455nm band.

[0055] The first wavelength division multiplexer 9 is connected to the third wavelength division multiplexer 11 via a sensing fiber, the second wavelength division multiplexer 10 is connected to the fourth wavelength division multiplexer 12 via a sensing fiber; and the second wavelength division multiplexer 10 is also connected to the dense wavelength division multiplexer 18, and the fourth wavelength division multiplexer 12 is connected to the first optical filter 19.

[0056] The second optical filter 20 is used to filter out noise.

[0057] The second light source 2 is connected to the input of the second fiber optic attenuator 14. The output of the second fiber optic attenuator 14 is connected to the input of a semiconductor optical amplifier 15. The output of the semiconductor optical amplifier 15 is connected to a data acquisition card 23 and a pulsed optical amplifier 16, respectively, for transmitting pulsed electrical signals to the data acquisition card 23 and pulsed light to the pulsed optical amplifier 16. The output of the pulsed optical amplifier 16 is connected to a fiber optic circulator 17, which is a three-port circulator connected to the output of the pulsed optical amplifier 16, the input of the dense wavelength division multiplexer 18, and the input of the second optical filter 20, respectively. The output of the second optical filter 20 is connected to a second photodetector 22, and the output of the second photodetector 22 is connected to the data acquisition card 23, which is connected to an industrial control computer.

[0058] The method for real-time monitoring of contact network cable status based on distributed optical fiber sensing includes:

[0059] Step 1: Real-time detection of interference light signals

[0060] S101: When an abnormal condition occurs in the contact network cable, it will cause a change in the sensor optical signal; the Mach-Zehnder interferometer and... In the integrated distributed optical fiber sensing system, the signal from the first light source 1 passes through the first optical fiber attenuator 3, the first optical fiber coupler 4, the Mach-Zehnder interferometer, and the first optical fiber coupler 13 in sequence to obtain the phase-modulated interference light signal. The data is then collected by the first photodetector 21 and the data acquisition card 23 and sent to the industrial control computer 24 to realize the feature extraction of the modulated light signal.

[0061] Specifically:

[0062] The laser emitted by the first light source 1 is transmitted to the first fiber coupler 4 via the first fiber attenuator 3. The first fiber coupler 4 splits the signal light into two beams, which then enter the Mach-Zehnder interferometer for interference. After passing through the sensing fiber, interference occurs again at the second fiber coupler 13. To extend the sensing distance, the Mach-Zehnder interferometer incorporates bidirectional distributed Raman amplification. The output light of the first Raman amplifier 5 and the second Raman amplifier 6 is in the 1455nm band. It is split into two paths by the second fiber couplers 7 and 8, and then enters the sensing fiber through the first wavelength division multiplexer 9, the second wavelength division multiplexer 10, the third wavelength division multiplexer 11, and the fourth wavelength division multiplexer 12, respectively, to amplify the transmitted light. The interfered signal light is transmitted to the first photodetector 21 and then enters the industrial control computer 24 via the data acquisition card 23.

[0063] S102: Simultaneously detect when an abnormal condition occurs in the contact wire cable. The backscattered Rayleigh light signal is processed, and the sensing light signal is processed; that is, the signal from the second light source 2 in the Mach-Zehnder interferometer and φ-OTDR fused distributed fiber optic sensing system passes sequentially through the second fiber attenuator 14, the semiconductor optical amplifier 15, and then enters the system. The light is transmitted through a pulsed light amplifier 16, an optical fiber circulator 17, and a dense wavelength division multiplexer 18, and then enters a Mach-Zehnder interferometer and is transmitted within the sensing optical fiber. The resulting continuous backscattered Rayleigh light signal is then obtained through the optical fiber circulator 17 and the second optical filter 20. The data is then collected by the second photodetector 22 and the data acquisition card 23 and sent to the industrial control computer 24 to realize the feature extraction of the backscattered Rayleigh light signal.

[0064] The laser emitted by the second light source 2 is transmitted to the semiconductor optical amplifier 15 via the second fiber optic attenuator 14. The pulsed electrical signal serves as the trigger signal for data acquisition by the data acquisition card 23. The pulsed light enters the pulsed optical amplifier 16 for amplification, passes through the fiber optic circulator 17, and enters one end of the dense wavelength division multiplexer 18. It is coupled together with one laser from the first light source 1 into the sensing fiber. The center wavelength of the first optical filter 19 is λ1±0.2nm, and only the laser emitted by the first light source 1 can pass through. When the pulsed light signal from the second light source (center wavelength λ2) is transmitted in the sensing fiber, it will generate a continuous backscattered Rayleigh light signal. Then, it passes through the fiber optic circulator 17, enters the second optical filter 20 to filter out noise, is received by the second photodetector 22, and enters the industrial control computer 24 via the data acquisition card 23.

[0065] Step 2: Preprocess the interference light signal and backscattered Rayleigh light signal obtained from the fault event to obtain phase change information.

[0066] S201: Extract the modulated interference optical signal of the fault event obtained in step one, perform DC removal and endpoint detection preprocessing, and construct a one-dimensional time sequence signal of the fault event.

[0067] When an abnormal condition occurs in the contact network cable (i.e., changes in external vibration, strain, temperature, etc., which cause changes in optical parameters such as intensity, phase, frequency, and polarization state of the transmitted light), it will cause changes in the sensing optical signal. Different fault events have different phase modulations on the optical fiber, and the corresponding interference optical signal characteristics are also different. Therefore, step one analyzes and processes external events by detecting the changes in the amplitude of the optical signal caused by the phase change.

[0068] The industrial control computer 24 expresses the interference intensity I after removing the DC component from the obtained modulated optical signal and detecting the endpoint as follows:

[0069]

[0070] A0 represents the amplitude of the continuous laser signal, φ(t) represents the phase change information introduced by the external event, and τ represents the time delay of the optical signal arriving at the first photodetector 21. Since different external fault events result in different phase modulations φ(t), the optical signals received by the industrial control computer 24 also exhibit different characteristics.

[0071] S202: Extract the modulated backscattered Rayleigh light signal of the fault event obtained in step one, and use the moving average algorithm to preprocess the obtained backscattered Rayleigh light signal to improve its signal-to-noise ratio and construct a one-dimensional time-series signal of the fault event; at the same time, perform difference operation on the backscattered Rayleigh light signal to demodulate the location of the fault event.

[0072] To obtain the intensity change information caused by the fault event after moving average calculation, a fixed backscattering Rayleigh curve is selected as the reference curve for the interpolation calculation, while the Mach-Zehnder interferometry and... By successively subtracting the backscattered Rayleigh scattering signal curves acquired by the fusion-based distributed fiber optic sensing system at other times from the reference curve, the significant intensity changes caused by the fault event can be obtained. The location of the fault event can be determined by detecting the changes in echo intensity and their delay time.

[0073] When an abnormality occurs in the contact network cable, the change in the phase of the backscattered Rayleigh light is proportional to the amplitude of the fault event acting on the sensing fiber of the Mach-Zehnder interferometer, causing a change in the intensity of the backscattered Rayleigh light at the corresponding location. Meanwhile, the intensity of the backscattered Rayleigh light at other locations unaffected by the fault event remains essentially unchanged. During long-distance sensing and monitoring, due to the transmission loss of the sensing fiber, the backscattered Rayleigh light is easily affected by external environmental noise and circuit-related noise.

[0074] The light field of the Rayleigh scattering light signal acquired by the industrial control computer 24 can be expressed as E. R Assume a sensing fiber of length L can be divided into N segments, each with a length ΔL = L / N, and the length of a segment is consistent with the width of the pulse light. These N segments can be represented as N mirrors with low reflectivity that is dependent on the incident light wavelength. These mirrors can be considered as the vector superposition of scatterers from each segment within length ΔL. Assuming there are R randomly and uniformly distributed Rayleigh scatterers within each segment of the sensing fiber, and these scatterers maintain a consistent polarization state and are independent of each other, then the vector field sum of these scatterers within the k-th segment is:

[0075]

[0076] Where A k θ k ai and A is considered a random variable. k Let θ be the sum of the amplitude vectors of the R Rayleigh scatterers in the k-th smallest segment of the sensing fiber. k Let a be the sum of the phase vectors of the R Rayleigh scatterers in the k-th segment of the sensing fiber. i Let be the amplitude of the i-th Rayleigh scatterer in the sensing fiber within a length ΔL. Let be the phase value of the i-th Rayleigh scatterer in the sensing fiber within the length ΔL range.

[0077] The multiple Rayleigh backscattered beams generated at the k-th segment of the sensing fiber will interfere due to their strong coherence. The optical field after coherent superposition can be expressed as:

[0078]

[0079] Where E0 is the electric field of the incident light field, and α is the loss coefficient of the sensing fiber.

[0080] Equation (3) shows that the coherent result of multiple backscattered Rayleigh beams in a certain segment of sensing fiber can be represented by the complex superposition of the amplitude and phase of the scatterer in that segment of sensing fiber. By analyzing the amplitude change of the backscattered Rayleigh curve, the phase change caused by the change of external physical quantity can be detected.

[0081] Among them, Raman pumping, Mach-Zehnder interferometry transmission light and The evolution of the probe optical pulse power along the optical fiber can be simplified as follows:

[0082]

[0083] The boundary conditions are: P R + (0)=P R,in + P R - (L)=P R,in - P S + (0)=P S,in + P P + (0)=P P,in + The superscripts + and - represent forward and backward transmission signals, respectively, P R + (z) and P R - (z) represents the Raman pump light propagating forward and backward along the optical fiber, P S +(z) is the forward propagation optical signal of the Mach-Zehnder interferometer, P P + (z) is The detection light pulse, α i ω represents the loss coefficient. i Let i = R, S, P represent the angular frequency of the optical signal, corresponding to optical signals with center wavelengths of 1455 nm and 1550.12 nm and 1550.92 nm, respectively. R This represents the Raman gain coefficient of a 1455nm optical signal compared to a 1550nm optical signal. The power distribution of the Rayleigh scattering curve is obtained through α BS P p (0)G(z)2Δz is obtained, where Δz is obtained from the probe light pulse, and α BS Pp(z) is the Rayleigh scattering coefficient, and G(z) is the network gain, which can be calculated from Pp(z) / Pp(0).

[0084] Step 3: The interference light signal and the backscattered Rayleigh light signal processed in Step 2 are fed into the pre-trained multi-class feature extraction deep learning network structure.

[0085] Pre-trained deep learning networks

[0086] Six thousand samples were pre-collected and processed to validate the feasibility and effectiveness of the proposed model. The dataset collected using the Mach-Zehnder interferometer consisted of 3000 samples. The collected dataset size is 3000. Five types of fault events (wire breakage, loosening, lightning strike, abnormal temperature rise, and no fault) were pre-selected, with 1200 sets of optical signals for each. Training, validation, and test data were randomly set and selected in an 8:1:1 ratio, with no overlap between data sets. The Mach-Zehnder interferometry and... Both the interference and backscattered Rayleigh light signals obtained by the fusion-based distributed fiber optic sensing system are trained using the Swin Transformer algorithm model. The Swin Transformer model maintains a high recognition rate while also achieving a fast recognition speed, meeting the needs of practical detection. During the training phase, the weight parameters provided by the official Swin Transformer platform are used as the model's initialization parameters, and the parameters are fine-tuned using a self-made dataset to achieve optimal detection performance. Training the labeled training and validation sets of images on the parameter-tuned Swin Transformer network model for 150 epochs generates the corresponding best training result weight file, best.pt.

[0087] Figure 3The flowchart illustrates the structure of a multi-class feature extraction deep learning network. The two-dimensional time-series spectrum of the fault event is fed into the Swing Transformer network model. After Patch Partitioning, Linear Embedding, and multi-level Patch Merging and two consecutive Swing Blocks based on sliding window self-attention, the signal is then classified using a Multilayer Perceptron (MLP) system, finally outputting the recognition result. In this embodiment, the model parameters are adjusted according to the type and characteristics of the fault event to train the neural network model, resulting in high recognition accuracy and classification stability. The trained Swing Transformer algorithm model is saved as a multi-class feature extraction deep learning network. The one-dimensional time-series signals obtained in steps S201 and S202 are trained together in this multi-class feature extraction deep learning network structure for joint recognition.

[0088] Processing optical signals

[0089] The one-dimensional time-series sensor signal obtained in steps S201 and S202 of step two is converted into a two-dimensional time spectrum using wavelet transform. This two-dimensional time spectrum is then fed into a pre-trained Swin Transformer algorithm model to adaptively extract time-domain, frequency-domain, and spatial-domain features of the optical signal. Finally, a classifier is used to identify and detect specific fault events. This feature enhancement method, which converts one-dimensional time-series signals into two dimensions, not only expands the dimensionality of the original sensor feature information but also effectively enhances the ability of deep learning models to adaptively extract typical and effective features of sensor signals.

[0090] Wavelet transform introduces a variable-width time-frequency window, enabling better capture of various signal features. The core idea is to create a mother wavelet function of finite length or rapidly decaying, generate multiple sub-wavelet functions through different scaling and translations, and finally match them with the input signal to analyze the time-frequency information of the signal at different scales and time locations, thus obtaining the time-frequency representation of the signal.

[0091] The specific implementation process is as follows: If the Mach-Zehnder fiber interferometer and The obtained sensor optical signal can be represented as x(t), then the time spectrum W(τ, f) obtained by wavelet transform of x(t) can be expressed as:

[0092]

[0093] Where, ψ a,bLet be the mother wavelet function, 'a' be the transform scaling factor, and 'b' be the translation factor. Since 'a' and 'b' are continuous values, there can be an infinite number of sub-wavelets, which are then matched with the original signal. This gives wavelet transform a better ability to adapt to changes in time-frequency resolution compared to short-time Fourier transform, and it can better highlight the local time-frequency features of the signal.

[0094] S303: For the two-dimensional time-frequency signal recognition in S302, precision, recall and F1 score are used as evaluation indicators. The event with the largest mAP (mean Average Precision) is taken as the final recognized event, that is, the intrusion event to which the current disturbance belongs is determined.

[0095] Finally, the fault event type, location, and time are simultaneously displayed and recorded on the industrial control computer. For five pre-set fault events—wire breakage, loosening, lightning strike, abnormal temperature rise, and no fault—alarms are triggered according to the severity of the fault event, alerting relevant personnel to intervene promptly and ensuring railway operational safety.

[0096] Although preferred embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims, and all of these are within the scope of protection of the present invention.

Claims

1. A Mach-Zehnder interferometer and φ-OTDR fusion type distributed optical fiber sensing system, characterized in that: The system comprises a first light source (1), a second light source (2), a Mach-Zehnder interferometer, a φ-OTDR, and a data acquisition and processing module. The first light source (1) is connected with a first optical fiber attenuator (3), a first optical fiber coupler (4) and an input end of the Mach-Zehnder interferometer in sequence. The second light source (2) is connected with a second optical fiber attenuator (14), a semiconductor optical amplifier (15) φ-OTDR and the data acquisition and processing module in sequence. The laser from the second light source (2) enters the Mach-Zehnder interferometer through the optical fiber circulator (17) and the dense wavelength division multiplexer (18) and is coupled into the sensing optical fiber together with the laser from the first light source (1). The input end of the Mach-Zehnder interferometer is provided with a first Raman amplifier (5) and the output end of the Mach-Zehnder interferometer is provided with a second Raman amplifier (6), thereby extending the sensing distance. The input end of the third optical fiber coupler (7) is connected with the first Raman amplifier (5) and the input end of the fourth optical fiber coupler (8) is connected with the second Raman amplifier (6). The data acquisition and processing module comprises a signal extraction and preprocessing submodule, a positioning submodule and a fault identification submodule. The signal extraction and preprocessing submodule is used for extracting the phase-modulated interference light signal and the backscattered Rayleigh light signal corresponding to the fault event of the Mach-Zehnder interferometer and φ-OTDR fusion type distributed optical fiber sensing subsystem. The positioning submodule is used for performing difference operation on the backscattered Rayleigh light signal to demodulate the position of the fault event. The fault identification submodule is used for converting the interference light signal and the backscattered Rayleigh light signal preprocessed by the signal extraction and preprocessing submodule into a two-dimensional time-frequency spectrum through wavelet transform, and then sending the two-dimensional time-frequency spectrum into a pre-trained Swin Transformer network model to extract the time domain feature, the frequency domain feature and the spatial domain feature of the light signal, and identifying the type of the fault event through a classifier. 2.The Mach-Zehnder interferometer and φ-OTDR fusion type distributed optical fiber sensing system according to claim 1, characterized in that, The splitting ratios of the first fiber coupler (4), the second fiber coupler (13), the third fiber coupler (7) and the fourth fiber coupler (8) are all 50:50, the wavelengths of the first fiber coupler (4) and the second fiber coupler (13) are 1550 nm, and the wavelengths of the third fiber coupler (7) and the fourth fiber coupler (8) are 1455 nm. 3.The Mach-Zehnder interferometer and φ-OTDR fusion type distributed optical fiber sensing system according to claim 1, characterized in that, The output light of the first Raman amplifier (5) and the second Raman amplifier (6) is in the 1455 nm wave band. 4.The Mach-Zehnder interferometer and φ-OTDR fusion type distributed optical fiber sensing system according to claim 1, characterized in that, The central wavelength of the first optical filter (19) is the central wavelength λ1±0.2 nm of the first light source, and only the laser emitted by the first light source (1) can pass through, which is used to filter out the pulsed light from the second light source (2). 5.The Mach-Zehnder interferometer and φ-OTDR fusion type distributed optical fiber sensing system according to claim 1, characterized in that, The first light source (1) and the second light source (2) are narrow-band continuous light lasers, and the working wavelength is in the 1550 nm wave band, and the maximum output power is 10 mW. 6.The Mach-Zehnder interferometer and φ-OTDR fusion type distributed optical fiber sensing system according to claim 1, characterized in that, The pre-set fault events include broken line, loose, lightning strike, abnormal temperature rise and no fault.

7. A real-time monitoring method for the state of a catenary cable based on distributed optical fiber sensing, the method using the Mach-Zehnder interference and φ-OTDR fusion type distributed optical fiber sensing system according to any one of claims 1-5, comprising: Step one: real-time detection of interference light signals, when the catenary cable has an abnormal condition, the sensing light signal changes, and the phase-modulated interference light signal and the backscattered Rayleigh light signal are obtained; Step two: removing the direct current component of the modulated light signal obtained by the fault event, performing endpoint detection, and obtaining the interference light signal corresponding to the fault event; The backscattered Rayleigh light signal obtained by the fault event is subjected to moving average algorithm and difference operation to obtain the obvious intensity change information caused by the fault event, and the position of the fault event is demodulated; Step three: performing wavelet transform on the interference light signal and the backscattered Rayleigh light signal processed in step two to convert them into two-dimensional time-frequency spectrum, then inputting the two-dimensional time-frequency spectrum into the pre-trained Swin Transformer network model to extract the time domain features, frequency domain features and spatial domain features of the light signal, and identifying the type of the fault event through the classifier.

8. The real-time monitoring method of claim 7, wherein Step one specifically includes: When the catenary cable has an abnormal condition, it will cause changes in the sensing light signal; the signal from the first light source (1) in the Mach-Zehnder interference and φ-OTDR fusion type distributed optical fiber sensing system passes through the first fiber attenuator (3), the first fiber coupler (4), the Mach-Zehnder interferometer and the second fiber coupler (13) in turn, and the phase-modulated interference light signal is obtained, which is sent to the data acquisition and processing module through the first photodetector (21) to realize feature extraction of the modulated light signal; When the abnormal condition occurs in the overhead line cable, the back Rayleigh scattering light signal of the φ-OTDR is detected, the signal from the second light source (2) passes through the second optical fiber attenuator (14), the semiconductor optical amplifier (15), the pulse optical amplifier (16), the optical fiber circulator (17), and the dense wavelength division multiplexer (18) in turn, enters the Mach-Zehnder interferometer, and is transmitted in the sensing optical fiber, the continuous back Rayleigh scattering light signal generated by the Mach-Zehnder interferometer passes through the optical fiber circulator (17) and the second optical filter (20) to obtain the back Rayleigh scattering light signal, and then passes through the second photoelectric detector (22) to send to the data acquisition and processing module to realize the feature extraction of the back Rayleigh scattering light signal.

9. The real-time monitoring method of claim 7, wherein The processing of the back Rayleigh scattering light signal in step two includes: A fixed back Rayleigh scattering curve is selected as a reference curve for difference operation, and the back Rayleigh scattering signal curves collected by the Mach-Zehnder interferometer and the φ-OTDR fusion type distributed optical fiber sensing system at other times are sequentially subjected to difference operation with the reference curve to obtain obvious intensity change information caused by the fault event, and the position information of the fault event is obtained by detecting the echo intensity change and the delay time.

Images