Optical fiber line fault detection method and system

Abstract

The present application discloses a method and system for detecting optical fiber line faults, the method comprising: using a broadband light source at both ends of an optical fiber in an optical fiber network, and frequency encoding the signal in the broadband light source to obtain an optical signal; collecting optical fiber sensing data; performing multimodal acquisition of the optical signal to obtain multimodal optical signal data; preprocessing the multimodal optical signal data and the optical fiber sensing data to obtain clear signal data; analyzing and fusing the clear signal data to obtain fused data; inputting the fused data into a pre-built fault identification model to output the fault type. The present invention effectively solves the problem of insufficient complex fault identification capability and environmental interference weakening the stability of the model through technologies such as multimodal data fusion, high-frequency signal extraction, physical model constraints and feature compression, and significantly improves the accuracy, robustness and generalization of optical fiber line fault detection.

Description

Technical Field

[0001] The present application relates to the field of optical communication technology, and in particular to a method and system for detecting optical fiber line faults. Background Art

[0002] As a key infrastructure for modern communications and sensing, optical fiber lines are widely used in communication networks, industrial monitoring, smart cities, and other fields. However, due to the structural characteristics of optical fibers and complex environmental conditions, their lines are susceptible to a variety of faults, including breakage, bending, contamination, abnormal temperature, external force impact, etc. These faults may lead to a decrease in signal transmission performance or even cause communication interruption.

[0003] Existing optical fiber line fault detection, for example, a Chinese patent with announcement number CN111740777B discloses an optical fiber line fault detection system and a detection method, wherein the optical fiber line fault detection system connects the transmitting optical fiber connected to the optical module to be tested and the active end of the receiving light together to form a complete optical circuit, in which the transmitting optical power of the transmitting optical fiber of the optical module to be tested and the receiving optical power of the receiving light can be simultaneously obtained at the end of the optical module to be tested through the network management unit of the optical module to be tested itself, without the need to set up detection personnel and detection equipment such as optical power meters at both ends of the optical fiber line, which is convenient and accurate, and reduces the personnel and equipment costs of optical fiber line fault detection while improving the efficiency of optical fiber line fault detection.

[0004] However, in the process of implementing relevant technical solutions, at least the following technical problems were found:

[0005] Relying on single-mode data (such as OTDR or DTS), it is difficult to accurately identify complex faults (such as multiple fault superposition or tiny cracks). In actual operation, the interference of environmental noise, temperature fluctuations and other factors on signal quality has not been fully compensated, further weakening the stability and robustness of detection. These problems make the generalization ability of fault identification in complex scenarios insufficient and difficult to adapt to various practical application requirements. Summary of the invention

[0006] In order to solve the above problems, an embodiment of the present invention provides a method for detecting optical fiber line faults, the method comprising:

[0007] A broadband light source is used at both ends of the optical fiber in the optical fiber network, and the signal in the broadband light source is frequency-encoded, and the frequency encoding includes frequency modulation and phase modulation; a directly modulated laser is used to perform frequency modulation on the optical fiber network; the optical fiber network is divided into N optical fiber lines, and phase modulation is performed in each optical fiber line to obtain an optical signal;

[0008] Collect fiber optic sensing data;

[0009] Perform multimodal acquisition on the optical signal to obtain multimodal optical signal data; the multimodal acquisition includes super-resolution spectral reconstruction model acquisition and reconstruction model acquisition; the reconstruction model acquisition method includes: inputting the reflected echo data into a pre-built time domain reconstruction model to output a high-resolution reconstruction signal; the super-resolution spectral reconstruction model acquisition method includes: inputting the phase information data into the super-resolution spectral reconstruction model to output a high-resolution spectrum signal;

[0010] The super-resolution spectrum reconstruction model includes a spectrum construction model and a consistency physical model;

[0011] The training methods for the spectrum construction model include:

[0012] S101: inputting phase information data into a spectrum construction model and outputting a high-resolution spectrum signal;

[0013] S102: inputting the high-resolution spectrum signal into the consistency physical model to check whether the high-resolution spectrum signal satisfies the physical constraint condition;

[0014] S103: If the physical constraint condition is not met, the spectrum is corrected by feedback of the joint loss function to build the model, and S101 to S103 are repeated until the joint loss function converges, and the training is completed;

[0015] Pre-process multi-modal optical signal data and optical fiber sensing data to obtain clear signal data;

[0016] Analyze and fuse the clear signal data to obtain fused data;

[0017] The fused data is input into the pre-built fault identification model and the fault type is output.

[0018] Furthermore, the training method of the time domain reconstruction model includes:

[0019] All reflection echo data are used as input of a time domain reconstruction model; the time domain reconstruction model takes the high-resolution reconstruction signal corresponding to each group of reflection echo data as output, takes the actual high-resolution reconstruction signal corresponding to each group of reflection echo data as a prediction target, and takes minimizing the sum of the first prediction accuracies of all predicted high-resolution reconstruction signals as a training target; the time domain reconstruction model is trained until the sum of the first prediction accuracies reaches convergence and the training is stopped; the time domain reconstruction model is a convolutional neural network model.

[0020] Furthermore, the super-resolution spectrum reconstruction model includes a spectrum construction model and a consistency physical model;

[0021] The training methods for the spectrum construction model include:

[0022] S101: inputting phase information data into a spectrum construction model and outputting a high-resolution spectrum signal;

[0023] S102: inputting the high-resolution spectrum signal into the consistency physical model to check whether the high-resolution spectrum signal satisfies the physical constraint condition;

[0024] S103: If the physical constraint condition is not met, the spectrum is corrected by feedback of the joint loss function to build the model, and S101 to S103 are repeated until the joint loss function converges, and the training is completed.

[0025] Furthermore, the consistency physical model includes:

[0026] ;

[0027] ;

[0028] ;

[0029] ;

[0030] In the formula, is the frequency domain; The amplitude of the frequency domain signal; is the time domain; is the angular frequency in radians per second; is the phase of the frequency domain signal; is the refractive index of the optical fiber; is the wavelength of light; is the path length of the optical fiber signal propagation.

[0031] Furthermore, the training method of the fault identification model includes:

[0032] All fused data are used as input of a fault identification model. The fault identification model uses the fault type corresponding to each set of fused data as output, the actual fault type corresponding to each set of fused data as a prediction target, and minimizing the sum of the second prediction accuracies of all predicted fault types as a training target. The fault identification model is trained until the sum of the second prediction accuracies reaches convergence, and the training is stopped. The fault identification model is a convolutional neural network model.

[0033] On the other hand, the present application also provides an optical fiber line fault detection system, comprising:

[0034] The detection signal transmission module uses a broadband light source at both ends of the optical fiber in the optical fiber network, and performs frequency encoding on the signal in the broadband light source, and the frequency encoding includes frequency modulation and phase modulation; a directly modulated laser is used to perform frequency modulation on the optical fiber network; the optical fiber network is divided into N optical fiber lines, and phase modulation is performed in each optical fiber line to obtain an optical signal;

[0035] A collection module, wherein the collection module is used to collect optical fiber sensing data;

[0036] A multimodal acquisition module, wherein the multimodal acquisition module is used to perform multimodal acquisition on optical signals to obtain multimodal optical signal data; the multimodal acquisition includes super-resolution spectral reconstruction model acquisition and reconstruction model acquisition; the reconstruction model acquisition method includes: inputting reflection echo data into a pre-built time domain reconstruction model to output a high-resolution reconstruction signal; the super-resolution spectral reconstruction model acquisition method includes: inputting phase information data into a super-resolution spectral reconstruction model to output a high-resolution spectrum signal; the super-resolution spectral reconstruction model includes a spectrum construction model and a consistency physical model;

[0037] The training methods for the spectrum construction model include:

[0038] S101: inputting phase information data into a spectrum construction model and outputting a high-resolution spectrum signal;

[0039] S102: inputting the high-resolution spectrum signal into the consistency physical model to check whether the high-resolution spectrum signal satisfies the physical constraint condition;

[0040] S103: If the physical constraint condition is not met, the spectrum is corrected by feedback of the joint loss function to build the model, and S101 to S103 are repeated until the joint loss function converges, and the training is completed;

[0041] A preprocessing module, which is used to preprocess the multimodal optical signal data and the optical fiber sensing data to obtain clear signal data;

[0042] A data fusion module, wherein the data fusion module is used to analyze and fuse the clear signal data to obtain fused data;

[0043] The fault analysis module is used to input the fusion data into the pre-built fault identification model and output the fault type.

[0044] The technical effects and advantages of the optical fiber line fault detection method and system provided by the present invention are as follows:

[0045] The present invention effectively solves the problems of insufficient complex fault identification capability caused by reliance on single modal data and weakening model stability due to environmental interference through technologies such as multimodal data fusion, high-frequency signal extraction, physical model constraints and feature compression, significantly improves the accuracy, robustness and generalization capability of optical fiber line fault detection, and provides reliable technical support for optical fiber network operation and maintenance in complex scenarios. The present invention adopts a broadband light source and introduces frequency coding to obtain an optical signal with stronger high-frequency characteristics, making the characteristics of small disturbances more obvious. The echo intensity of OTDR and the phase spectrum information of OFDR are integrated, and the temperature and vibration data of DTS and DAS are combined to comprehensively characterize the fault characteristics through multimodal data fusion, which significantly improves the detection capability of complex faults (such as multiple fault superposition and small cracks). In the preprocessing stage, wavelet decomposition is used to reduce the noise of multimodal signals, retain low-frequency components and remove high-frequency noise. Through baseline drift correction and temperature compensation, the interference of equipment drift and temperature change on signal characteristics is eliminated, the stability and consistency of signal quality are enhanced, and a high fault detection accuracy is maintained in complex environments. In the super-resolution spectral reconstruction model, a consistency physical model is introduced. Through frequency consistency, phase consistency and time-frequency consistency constraints, it is ensured that the spectrum reconstruction result conforms to the physical laws of the optical fiber propagation path, and the joint loss function optimizes the training process of the deep learning model, which can effectively avoid spectrum distortion. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of a method for detecting optical fiber line faults in Embodiment 1;

[0047] Figure 2 This is a flow chart of the spectrum construction model training method in Example 1;

[0048] Figure 3 This is a connection diagram of an optical fiber line fault detection system in Embodiment 2. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] Embodiment 1:

[0051] See also Figure 1 As shown, the optical fiber line fault detection method described in this embodiment includes:

[0052] A broadband light source is used at both ends of an optical fiber in a fiber optic network, and a signal in the broadband light source is frequency-encoded to obtain an optical signal;

[0053] Collect fiber optic sensing data;

[0054] Perform multi-modal acquisition on optical signals to obtain multi-modal optical signal data;

[0055] Pre-process multi-modal optical signal data and optical fiber sensing data to obtain clear signal data;

[0056] Analyze and fuse the clear signal data to obtain fused data;

[0057] The fused data is input into the pre-built fault identification model and the fault type is output.

[0058] Optical fiber networks include optical fibers, optical signals, and node devices (such as light sources (lasers), optical detectors (photodiodes), optical amplifiers (EDFAs), and optical switches); broadband light sources, such as pulsed lasers and other frequency-intensive light sources, can better detect any tiny disturbances in optical fiber transmission (such as bends, cracks, or external force impacts);

[0059] Frequency coding includes frequency modulation and phase modulation. Directly modulated laser frequency modulation is used on the optical fiber network to divide the optical fiber network into N optical fiber lines. Phase modulation is performed in each optical fiber line to obtain an optical signal.

[0060] light signal The methods for obtaining include:

[0061] ;

[0062] In the formula, is the initial amplitude of the optical electric field (constant), which determines the intensity of the optical signal; is a constant ; is an imaginary unit; is the carrier angular frequency of the light wave, in radians per second; is the time variable, the unit is second; is the modulation depth, which is dimensionless and determines the strength of the phase modulation and affects the degree of splitting of the optical signal spectrum. When <<1, the optical signal is mainly concentrated at the center frequency and the spectrum is narrow. When >1, the spectrum is split into multiple side bands, and the modulation effect is significant; is the angular frequency of the modulation signal, measured in radians per second, which determines the frequency of the modulation signal, that is, the periodicity of the phase change; The sine function representing the modulation signal over time defines how the modulation signal affects the phase of the light wave. The modulation amplitude is given by Decide.

[0063] Introducing a broadband light source and frequency encoding the optical signal can give the signal in the optical fiber high-frequency characteristics, thereby enhancing the detectability of tiny disturbances through frequency domain characteristic extraction (such as the spectral distribution of the signal waveform).

[0064] Fiber optic sensing data includes temperature data and strain data of the optical fiber, which are collected using distributed fiber optic sensors (such as DTS and DAS); multimodal acquisition includes super-resolution spectral reconstruction model acquisition and reconstruction model acquisition.

[0065] The optical signal includes reflected echo data and phase information data. The reflected echo data is measured using an optical time domain reflectometer (OTDR) to locate the approximate fault point, and then a broadband light source is used at both ends of the optical fiber to complete bidirectional signal transmission and obtain the reflection anomaly point. If the approximate fault point coincides with the reflection anomaly point, the fault point can be accurately located. The phase information data is extracted using an optical frequency domain reflectometer (OFDR) to identify tiny physical deformations.

[0066] The reconstruction model acquisition method includes: inputting the reflected echo data into a pre-built time domain reconstruction model, and outputting a high-resolution reconstruction signal.

[0067] The training methods of the time domain reconstruction model include:

[0068] All reflected echo data are used as the input of the time domain reconstruction model. The time domain reconstruction model predicts the corresponding high-resolution reconstructed signal for each group of reflected echo data as output, takes the actual high-resolution reconstructed signal corresponding to each group of reflected echo data as the prediction target, and takes minimizing the sum of the first prediction accuracies of all predicted high-resolution reconstructed signals as the training goal.

[0069] Among them, the calculation formula for the first prediction accuracy is: ,in, is the number of each set of reflection echo data, is the first prediction accuracy, For the The predicted high-resolution reconstructed signal corresponding to the reflected echo data, For the The actual high-resolution reconstructed signal corresponding to the group of reflected echo data is obtained; the time domain reconstruction model is trained until the sum of the first prediction accuracies reaches convergence and the training is stopped; the time domain reconstruction model is a convolutional neural network model.

[0070] By performing nonlinear feature extraction on low-resolution signals through the time domain reconstruction model (CNN), it is possible to restore hidden detail information from the original signal, significantly improve the time domain resolution, and increase the adaptability in complex environments (such as high-noise scenarios or multiple fault point distributions).

[0071] The super-resolution spectrum reconstruction model acquisition method includes: inputting phase information data into the super-resolution spectrum reconstruction model and outputting a high-resolution spectrum signal.

[0072] The super-resolution spectrum reconstruction model includes a spectrum construction model and a consistency physical model;

[0073] like Figure 2 As shown, the training method of the spectrum construction model includes:

[0074] S101: inputting phase information data into a spectrum construction model and outputting a high-resolution spectrum signal;

[0075] S102: inputting the high-resolution spectrum signal into the consistency physical model to check whether the high-resolution spectrum signal satisfies the physical constraint condition;

[0076] S103: If the physical constraint condition is not met, the spectrum is corrected by feedback of the joint loss function to build the model, and S101 to S103 are repeated until the joint loss function converges, and the training is completed;

[0077] The consistency physical model includes:

[0078] ;

[0079] ;

[0080] ;

[0081] Optical fiber propagation path formula:

[0082] ;

[0083] In the formula, is the frequency domain, which represents the spectrum distribution of the optical signal; The amplitude of the frequency domain signal indicates the strength of the reflected signal; is the time domain, which means the optical signal in time changes in the reflection intensity or phase; is the angular frequency in radians per second; It is the phase of the frequency domain signal, reflecting the delay or deformation of the optical signal propagation path; is the refractive index of the optical fiber; is the wavelength of light; is the path length of the optical fiber signal propagation.

[0084] Physical constraints include frequency consistency, phase consistency, and time-frequency consistency;

[0085] Frequency consistency loss includes the amplitude and phase of the high-resolution spectrum signal being consistent with the results of the consistent physical model calculation; the physical model calculation results include , , and .

[0086] Phase coherence includes the phase of the high-resolution spectrum signal Satisfy the fiber propagation path formula in the consistent physical model.

[0087] Time-frequency consistency includes that the frequency domain and time domain of the high-resolution spectrum signal are consistent with the calculation results of the consistency physical model;

[0088] Joint loss function include:

[0089] ;

[0090] ;

[0091] ;

[0092] ;

[0093] In the formula, , and It is a preset weight coefficient used to balance the importance of different loss items; is the frequency consistency loss term; is the phase consistency loss term; is the time-frequency consistency loss term; is the spectrum of the high-resolution spectrum signal; is the phase of the high-resolution spectrum signal; It is the time domain of high-resolution spectrum signal.

[0094] By combining the spectrum construction model with the consistent physical model, high-resolution spectrum reconstruction of the optical fiber signal is achieved, and the physical consistency of the reconstruction result is ensured by physical constraints. The high-frequency details of the signal can be restored without being restricted by the sampling rate and instrument bandwidth. The reason for using the consistent physical model is that although the deep learning model has powerful nonlinear fitting capabilities, it may generate spectra that do not conform to physical laws (such as distorted frequency amplitudes or unreasonable phase distributions). The consistent physical model can correct this problem well.

[0095] The multimodal optical signal data includes high-resolution spectrum signals and high-resolution reconstruction signals;

[0096] Preprocessing methods include:

[0097] Perform wavelet decomposition on multi-modal optical signal data and optical fiber sensor data to separate low-frequency components and high-frequency components;

[0098] The low-frequency components (signal main components) are retained, and the high-frequency components are threshold processed to obtain clear multi-modal optical signal data and clear optical fiber sensing data;

[0099] The clear multimodal optical signal data and the clear optical fiber sensing data are aligned on the same time and space axis and then normalized. After the data is normalized, baseline drift correction and temperature compensation are performed.

[0100] Baseline drift correction and temperature compensation code example:

[0101] import numpy as np

[0102] import matplotlib.pyplot as plt

[0103] # analog signal

[0104] x = np.linspace(0, 100, 1000) # Fiber length

[0105] baseline_drift = 0.05 * x # simulate baseline drift

[0106] temperature_change = 0.1 * np.sin(0.1 * x) # simulate temperature change

[0107] true_signal = np.sin(2 * np.pi * 0.01 * x) # simulate the real signal

[0108] measured_signal = true_signal + baseline_drift + temperature_change #Measurement signal

[0109] # Baseline drift correction

[0110] baseline_mean = np.mean(baseline_drift) # Calculate the baseline mean

[0111] signal_no_baseline = measured_signal - baseline_mean # Correct baseline drift

[0112] # Temperature compensation

[0113] temperature_coeff = 0.1 # Temperature coefficient (assumed)

[0114] temperature_ref = np.mean(temperature_change) # Reference temperature

[0115] temperature_deviation = temperature_change - temperature_ref # Temperature deviation

[0116] signal_final = signal_no_baseline - temperature_coeff * temperature_deviation # Compensate for temperature effects

[0117] # Drawing comparison

[0118] plt.figure(figsize=(10, 6))

[0119] plt.plot(x, measured_signal, label="Measured Signal")

[0120] plt.plot(x, true_signal, label="True Signal")

[0121] plt.plot(x, signal_final, label="Corrected Signal")

[0122] plt.legend()

[0123] plt.xlabel("Position (x)")

[0124] plt.ylabel("Signal")

[0125] plt.title("Baseline Drift and Temperature Compensation")

[0126] plt.show().

[0127] The analysis fusion adopts the weighted fusion method, that is, according to the importance of each modal feature, weights are assigned and then fused; partial fusion has been performed in the preprocessing stage, and the analysis fusion is to enable the subsequent machine learning model to quickly identify the fault type.

[0128] It should be noted that in machine learning, the fewer features, the faster the recognition. In the previous steps, the multimodal optical signal data is continuously reduced, and will eventually be recognized with one feature or a few features, but the information is not actually reduced. It is a way of compressing information. For example, the features in the multimodal optical signal data and the optical fiber sensor data include: a reflection point where the echo signal drops sharply, and there is no subsequent reflection signal; a significant jump in the phase and amplitude of the spectrum; an interruption of the temperature or vibration signal; and the final fused data is 1. In the machine learning model, if the fused data is 1 and is within the numerical range representing the fault type of optical fiber break, the final output is optical fiber break.

[0129] Fault types include fiber breakage, fiber bending, fusion loss, fiber contamination, fiber microcracks, temperature anomalies, external force impact, fiber stretching and compression, fiber aging and multiple fault superposition.

[0130] The characteristic recognition logic of the fault type is as follows:

[0131] Fiber breakage:

[0132] describe:

[0133] The optical fiber is physically disconnected, resulting in the complete inability to transmit signals.

[0134] feature:

[0135] OTDR: The echo signal shows a reflection point where it drops sharply, and there is no subsequent reflection signal.

[0136] OFDR: Significant jumps in phase and amplitude occur in the spectrum.

[0137] DTS / DAS: Temperature or vibration signal interrupted.

[0138] Influence:

[0139] Communications on the entire fiber optic line were interrupted.

[0140] Fiber Bending:

[0141] describe:

[0142] Excessive bending of the optical fiber causes partial loss of the optical signal or increased reflection.

[0143] feature:

[0144] OTDR: A slight drop or abnormal reflection peak appears in the reflected signal.

[0145] OFDR: The spectrum amplitude fluctuates significantly in a specific frequency band.

[0146] DTS: The temperature gradient in the bend area may be abnormal.

[0147] DAS: Abnormal vibration signal strength.

[0148] Influence:

[0149] The optical fiber signal attenuates and the communication quality decreases.

[0150] Fiber splicing loss:

[0151] describe:

[0152] The fiber fusion point is poorly connected, resulting in partial reflection and loss of the optical signal.

[0153] feature:

[0154] OTDR: The echo signal shows a slight drop point.

[0155] OFDR: The spectrum phase changes slightly, but the amplitude fluctuates slightly.

[0156] DTS / DAS: Generally no significant changes.

[0157] Influence:

[0158] Signal transmission efficiency is reduced.

[0159] Fiber contamination or contamination points:

[0160] describe:

[0161] There is localized contamination (such as dust, moisture, or oil) in the optical fiber, causing signal attenuation.

[0162] feature:

[0163] OTDR: Slow attenuation characteristics appear in the echo signal.

[0164] OFDR: The phase changes in the spectrum show a slow trend.

[0165] DTS: Abnormal temperature distribution in the polluted area.

[0166] DAS: No obvious features.

[0167] Influence:

[0168] Long-term communication performance degradation.

[0169] Optical fiber microcracks:

[0170] describe:

[0171] Tiny cracks on the surface of the optical fiber sheath or core may cause signal quality degradation during long-term use.

[0172] feature:

[0173] OTDR: A small drop point appears in the reflected signal, and the subsequent signal quality deteriorates.

[0174] OFDR: Slight amplitude and phase fluctuations appear in the spectrum.

[0175] DTS / DAS: May not have significant features.

[0176] Influence:

[0177] Optical fiber reliability decreases and may gradually develop into a break.

[0178] Abnormal temperature:

[0179] describe:

[0180] The external environment temperature rises or drops abnormally, affecting the performance of the optical fiber.

[0181] feature:

[0182] DTS: A significant abnormal gradient appears in the temperature distribution.

[0183] DAS: Temperature changes may induce changes in strain or vibration signals.

[0184] OTDR / OFDR: No direct change in signal characteristics, but correlated with temperature sensing data.

[0185] Influence:

[0186] Fiber optic performance is affected, especially excessive temperature may cause damage.

[0187] External impact:

[0188] describe:

[0189] The optical fiber is subjected to external impact or interference, resulting in instantaneous signal fluctuations.

[0190] feature:

[0191] DAS: Abnormal high-frequency peaks appear in the vibration signal.

[0192] DTS: The temperature in the impact area may fluctuate.

[0193] OTDR: The signal is momentarily abnormal and recovers over time.

[0194] OFDR: No significant changes are likely.

[0195] Influence:

[0196] Short-term disturbances usually do not cause long-term damage.

[0197] Fiber stretching or compression:

[0198] describe:

[0199] The optical fiber is subjected to tension or compression, causing the signal characteristics to change.

[0200] feature:

[0201] OTDR: The echo signal strength may decrease slowly.

[0202] OFDR: The spectrum phase changes continuously.

[0203] DAS: Abnormalities in a specific pattern occur in the strain signal.

[0204] DTS: The temperature distribution in the stretching area may change.

[0205] Influence:

[0206] The optical fiber performance changes and long-term stretching may cause breakage.

[0207] Fiber aging:

[0208] describe:

[0209] Long-term use of optical fibers leads to performance degradation (such as signal attenuation and strain accumulation).

[0210] feature:

[0211] OTDR: The overall signal strength decreases, and the curve shows a gentle attenuation trend.

[0212] OFDR: The spectrum amplitude shows uniform attenuation.

[0213] DTS / DAS: No obvious features.

[0214] Influence:

[0215] Optical fiber reliability decreases and requires regular maintenance or replacement.

[0216] Multiple faults superposition:

[0217] describe:

[0218] Optical fibers can experience multiple types of faults simultaneously (such as breaks and bends, contamination and stretching).

[0219] feature:

[0220] OTDR: Various types of reflection anomalies appear in the echo signal.

[0221] OFDR: The spectrum amplitude and phase changes are complex.

[0222] DTS / DAS: Feature superposition of multiple fault areas.

[0223] Influence:

[0224] The fault types are complex and difficult to identify.

[0225] The training methods of the fault identification model include:

[0226] All fused data are used as the input of the fault identification model. The fault identification model predicts the corresponding fault type for each set of fused data as output, takes the actual fault type corresponding to each set of fused data as the prediction target, and takes minimizing the sum of the second prediction accuracies of all predicted fault types as the training target.

[0227] Among them, the calculation formula for the second prediction accuracy is: ,in, is the number of each group of fusion data, is the second prediction accuracy, For the The predicted fault type corresponding to the fused data, For the The actual fault type corresponding to the group fusion data; the fault identification model is trained until the sum of the second prediction accuracies reaches convergence and the training is stopped; the fault identification model is a convolutional neural network model.

[0228] Embodiment 2:

[0229] like Figure 3 As shown, based on the same inventive concept as the optical fiber line fault detection method in the aforementioned embodiment, the present application provides an optical fiber line fault detection system, and the method and system embodiments in the embodiments of the present application are based on the same inventive concept. The system includes:

[0230] The detection signal transmission module uses a broadband light source at both ends of the optical fiber in the optical fiber network, and performs frequency encoding on the signal in the broadband light source, and the frequency encoding includes frequency modulation and phase modulation; a directly modulated laser is used to perform frequency modulation on the optical fiber network; the optical fiber network is divided into N optical fiber lines, and phase modulation is performed in each optical fiber line to obtain an optical signal;

[0231] A collection module, wherein the collection module is used to collect optical fiber sensing data;

[0232] A multimodal acquisition module, wherein the multimodal acquisition module is used to perform multimodal acquisition on optical signals to obtain multimodal optical signal data; the multimodal acquisition includes super-resolution spectral reconstruction model acquisition and reconstruction model acquisition; the reconstruction model acquisition method includes: inputting reflection echo data into a pre-built time domain reconstruction model to output a high-resolution reconstruction signal; the super-resolution spectral reconstruction model acquisition method includes: inputting phase information data into a super-resolution spectral reconstruction model to output a high-resolution spectrum signal; the super-resolution spectral reconstruction model includes a spectrum construction model and a consistency physical model;

[0233] The training methods for the spectrum construction model include:

[0234] S101: inputting phase information data into a spectrum construction model and outputting a high-resolution spectrum signal;

[0235] S102: inputting the high-resolution spectrum signal into the consistency physical model to check whether the high-resolution spectrum signal satisfies the physical constraint condition;

[0236] S103: If the physical constraint condition is not met, the spectrum is corrected by feedback of the joint loss function to build the model, and S101 to S103 are repeated until the joint loss function converges, and the training is completed;

[0237] A preprocessing module, which is used to preprocess the multimodal optical signal data and the optical fiber sensing data to obtain clear signal data;

[0238] A data fusion module, wherein the data fusion module is used to analyze and fuse the clear signal data to obtain fused data;

[0239] The fault analysis module is used to input the fusion data into the pre-built fault identification model and output the fault type.

[0240] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0241] What has been described above is only a preferred specific implementation manner of the embodiments of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can make equivalent substitutions or changes according to the technical scheme and concept of the present application within the technical scope disclosed in the present application, which should be covered by the protection scope of the present application.

Claims

1. A method for detecting optical fiber line faults, characterized in that: Methods include: A broadband light source is used at both ends of an optical fiber in an optical fiber network, and a signal in the broadband light source is frequency-encoded, wherein the frequency encoding includes frequency modulation and phase modulation; Frequency modulation using directly modulated lasers over fiber optic networks; The optical fiber network is divided into N optical fiber lines, and phase modulation is performed in each optical fiber line to obtain an optical signal; Collect fiber optic sensing data; Perform multi-modal acquisition on optical signals to obtain multi-modal optical signal data; Multimodal acquisition includes super-resolution spectral reconstruction model acquisition and reconstruction model acquisition; The reconstruction model acquisition method includes: inputting the reflected echo data into a pre-built time domain reconstruction model, and outputting a high-resolution reconstruction signal; the super-resolution spectrum reconstruction model acquisition method includes: inputting the phase information data into the super-resolution spectrum reconstruction model, and outputting a high-resolution spectrum signal; The super-resolution spectrum reconstruction model includes a spectrum construction model and a consistency physical model; The training methods for the spectrum construction model include: S101: inputting phase information data into a spectrum construction model and outputting a high-resolution spectrum signal; S102: inputting the high-resolution spectrum signal into the consistency physical model to check whether the high-resolution spectrum signal satisfies the physical constraint condition; S103: If the physical constraint condition is not met, the spectrum is corrected by feedback of the joint loss function to build the model, and S101 to S103 are repeated until the joint loss function converges, and the training is completed; Pre-process multi-modal optical signal data and optical fiber sensing data to obtain clear signal data; Analyze and fuse the clear signal data to obtain fused data; The fused data is input into the pre-built fault identification model and the fault type is output.

2. A method for detecting optical fiber line faults according to claim 1, characterized in that: The training method of the time domain reconstruction model includes: All reflection echo data are used as input of a time domain reconstruction model; the time domain reconstruction model takes the high-resolution reconstruction signal corresponding to each group of reflection echo data as output, takes the actual high-resolution reconstruction signal corresponding to each group of reflection echo data as a prediction target, and takes minimizing the sum of the first prediction accuracies of all predicted high-resolution reconstruction signals as a training target; the time domain reconstruction model is trained until the sum of the first prediction accuracies reaches convergence and the training is stopped; the time domain reconstruction model is a convolutional neural network model.

3. The optical fiber line fault detection method according to claim 1, characterized in that: The consistency physical model includes: ； ； ； ； In the formula, is the frequency domain; The amplitude of the frequency domain signal; is the time domain; is the angular frequency in radians per second; is the phase of the frequency domain signal; is the refractive index of the optical fiber; is the wavelength of light; is the path length of the optical fiber signal propagation.

4. A method for detecting optical fiber line faults according to claim 1, characterized in that: The training method of the fault identification model includes: All fused data are used as input of a fault identification model. The fault identification model uses the fault type corresponding to each set of fused data as output, the actual fault type corresponding to each set of fused data as a prediction target, and minimizing the sum of the second prediction accuracies of all predicted fault types as a training target. The fault identification model is trained until the sum of the second prediction accuracies reaches convergence, and the training is stopped. The fault identification model is a convolutional neural network model.

5. An optical fiber line fault detection system, characterized in that: include: The detection signal transmission module uses a broadband light source at both ends of the optical fiber in the optical fiber network, and performs frequency encoding on the signal in the broadband light source, wherein the frequency encoding includes frequency modulation and phase modulation; Frequency modulation using directly modulated lasers over fiber optic networks; The optical fiber network is divided into N optical fiber lines, and phase modulation is performed in each optical fiber line to obtain an optical signal; A collection module, wherein the collection module is used to collect optical fiber sensing data; A multimodal acquisition module, wherein the multimodal acquisition module is used to perform multimodal acquisition on the optical signal to obtain multimodal optical signal data; Multimodal acquisition includes super-resolution spectral reconstruction model acquisition and reconstruction model acquisition; The reconstruction model acquisition method includes: inputting the reflected echo data into a pre-built time domain reconstruction model, and outputting a high-resolution reconstruction signal; the super-resolution spectrum reconstruction model acquisition method includes: inputting the phase information data into the super-resolution spectrum reconstruction model, and outputting a high-resolution spectrum signal; the super-resolution spectrum reconstruction model includes a spectrum construction model and a consistency physical model; The training methods for the spectrum construction model include: S101: inputting phase information data into a spectrum construction model and outputting a high-resolution spectrum signal; S102: inputting the high-resolution spectrum signal into the consistency physical model to check whether the high-resolution spectrum signal satisfies the physical constraint condition; S103: If the physical constraint condition is not met, the spectrum is corrected by feedback of the joint loss function to build the model, and S101 to S103 are repeated until the joint loss function converges, and the training is completed; A preprocessing module, which is used to preprocess the multimodal optical signal data and the optical fiber sensing data to obtain clear signal data; A data fusion module, wherein the data fusion module is used to analyze and fuse the clear signal data to obtain fused data; The fault analysis module is used to input the fusion data into the pre-built fault identification model and output the fault type.

Images