A method and apparatus for distributed fiber optic acoustic anomaly detection

Abstract

The present application relates to the field of distributed optical fiber sensing, in particular to a method and device for detecting abnormal sound waves in distributed optical fiber. Mainly including: obtaining the X-state electrical signal and Y-state electrical signal after mixing the RBS optical signal from a chirp optical pulse with reference light, and obtaining the intensity electrical signal of the RBS optical signal; obtaining the phase curve of the distributed optical fiber along the line according to the X-state electrical signal and Y-state electrical signal, and obtaining the intensity curve according to the intensity electrical signal; inputting the phase curve into the convolutional neural network to obtain the vibration feature, inputting the intensity curve into the multi-layer perceptron module to obtain the fading feature, and inputting the vibration feature and the fading feature into the decision module after fusion to output the abnormal detection result. The present application can eliminate the influence of interference fading according to the fading feature.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of distributed optical fiber sensing, and in particular to a method and device for detecting acoustic anomalies in distributed optical fiber sensing. BACKGROUND

[0002] Distributed optical fiber sensing technology based on phase sensitive optical time-domain reflection (Φ-OTDR) has many advantages such as high measurement accuracy, fast response speed, long monitoring distance, and anti-electromagnetic interference, and has been widely applied in perimeter security, rail transit, and oil and gas pipeline monitoring. The principle is that the disturbance signal applied to the optical fiber causes changes in the refractive index of the optical fiber, which in turn causes changes in the optical path and the phase of the probe light. According to the different demodulation methods, Φ-OTDR is divided into intensity demodulation type and phase demodulation type.

[0003] The typical architecture of intensity demodulation type Φ-OTDR is shown in Figure 1 A narrow line laser (NLL) outputs light that is modulated into a pulsed light signal by an acoustic optical modulator (AOM), amplified by an Erbium Doped Fiber Application Amplifier (EDFA), and injected into an optical fiber. The Rayleigh Backscattering (RBS) signal reflected back is detected by a Photo-Diode (PD), collected by a Data Acquisition (DAQ) card, and input to a data processing unit for processing. Since the intensity of the scattered signal does not change linearly with the disturbance signal, intensity demodulation type Φ-OTDR is usually only used for qualitative measurement, and the detection accuracy is relatively low.

[0004] Phase demodulation type Φ-OTDR can realize quantitative restoration of disturbance signal because the phase signal is linearly related to the disturbance signal. At present, most of the phase demodulation type Φ-OTDR systems adopt coherent detection technology. The basic principle is that a narrow linewidth light source is modulated into a pulse signal and injected into a sensing optical fiber. The reflected Rayleigh backscattering light signal is mixed with the intrinsic signal for coherent detection. The phase information of RBS is obtained by using a suitable demodulation algorithm. The phase information obtained after demodulation is input into a signal processing unit for pattern recognition and event detection. The traditional pattern recognition method is mainly based on the feature extraction method of the signal. However, the feature extraction process is often very tedious and the calculation amount is large. In recent years, researchers have tried many signal recognition methods based on deep learning, such as using convolutional neural network (CNN) and recurrent neural network (RNN) model for pattern recognition. However, the existing deep learning pattern recognition algorithm mostly uses single phase information input.

[0005] A typical convolutional neural network-based pattern recognition method is shown in FIG. 1. The preprocessed phase signal is input into the model as the input. After passing through multiple convolutional layers and pooling layers, the input is input into a decision module composed of a fully connected layer and a soft-max to obtain the output of event detection. Figure 2

[0006] In the phase demodulation type Φ-OTDR system, interference fading effect inevitably occurs, which greatly deteriorates the signal-to-noise ratio of the fading point and causes serious distortion of the sensing information. Specifically, there are inevitably minimum points in the Rayleigh scattering intensity. After superimposing the detector intensity noise and after the demodulation algorithm, the phase result will have abnormal values. It is usually difficult to distinguish the phase jump caused by interference fading from the jump caused by actual disturbance. Therefore, the Φ-OTDR system based on single phase signal detection has problems such as event false alarm caused by interference fading.

[0007] For the coherent detection Φ-OTDR system, the main methods to eliminate interference fading are:

[0008] (1) The phase shift double pulse method is used to eliminate interference fading. Two pulses with a phase difference of π are sequentially injected into the optical fiber. The complementary characteristics of the interference pattern are used to obtain signals with different positions of interference fading points. After superimposing the two signals, most of the interference fading points can be eliminated.

[0009] Technical defects: The modulation technology is relatively complex, and accurate pulse synchronization is required.

[0010] (2) Chirped pulse Φ-OTDR technology is used to eliminate interference fading by using multiple Rayleigh scattering signals with different frequencies.​

[0011] Technical defects: the hardware is relatively complex, and a swept light source device and a frequency domain demodulation algorithm are needed.

[0012] Therefore, how to overcome the defects of the prior art and solve the problem of interference fading in the existing Ф-OTDR system is a problem to be solved in the technical field. SUMMARY

[0013] In view of the above defects or improvement needs of the prior art, the present application solves the problem of interference fading in the existing Ф-OTDR system.

[0014] The embodiment of the present application adopts the following technical solutions:

[0015] In a first aspect, the present application provides a method for distributed optical fiber acoustic anomaly detection, specifically: obtaining X-state and Y-state electrical signals after mixing RBS optical signals from a chirp light pulse with reference light, and obtaining intensity electrical signals of the RBS optical signals; obtaining a phase curve along the line of the distributed optical fiber according to the X-state and Y-state electrical signals, and obtaining an intensity curve according to the intensity electrical signals; inputting the phase curve into a convolutional neural network to obtain vibration features, inputting the intensity curve into a multilayer perceptron module to obtain fading features, and inputting the vibration features and the fading features into a decision module after fusion to output an anomaly detection result.

[0016] Preferably, the X-state and Y-state electrical signals of the RBS optical signals from the chirp light pulse specifically include: mixing the RBS signals from the chirp light pulse with reference light according to two orthogonal X-polarization states and Y-polarization states respectively for balanced reception, obtaining beat frequency electrical signals of the X-polarization state as the X-state electrical signals, and obtaining beat frequency electrical signals of the Y-polarization state as the Y-state electrical signals.

[0017] Preferably, the phase curve along the line of the distributed optical fiber obtained according to the X-state and Y-state electrical signals specifically includes: sorting all X-state and Y-state electrical signals according to time for one-to-one correspondence; obtaining corresponding phase signals respectively by using a Hilbert phase demodulation algorithm for the sorted X-state and Y-state electrical signals, and performing average superposition to obtain a demodulated phase signal sequence; and performing normalization processing on the phase sequence to form a phase curve.

[0018] Preferably, the Hilbert phase demodulation algorithm is used to obtain the corresponding phase signals respectively, and the demodulated phase signal sequence is obtained by averaging and superimposing, specifically including: calculating the Hilbert transform of the X-state electric signal and the Y-state electric signal point by point, respectively obtaining the Hilbert transform sequence of the X-state and the Hilbert transform sequence of the Y-state, respectively calculating the arctangent of the Hilbert transform sequence of the X-state and the Hilbert transform sequence of the Y-state; for the arctangent of the X-state and the arctangent of the Y-state, the corresponding phase signal of each point is obtained by calculating the mean value point by point, and the phase signal sequence is obtained by sorting the phase signal frame by frame.

[0019] Preferably, the intensity curve is obtained according to the intensity electric signal, specifically including: the intensity electric signal obtained by sampling is sorted frame by frame; the intensity electric signal is converted into signal-to-noise ratio data and logarithm is taken to normalize, forming the intensity curve.

[0020] Preferably, the vibration feature and the fading feature are fused and input into the decision module for output of the abnormal detection result, specifically including: the phase signal sequence is input into the convolutional neural network module for vibration feature extraction, obtaining the vibration feature vector; the intensity signal sequence is input into the multilayer perceptron module for fading feature extraction, obtaining the fading feature vector; the vibration feature vector and the fading feature vector are input into the connection layer, and the pattern recognition result is obtained through the classifier module as the output of the abnormal detection result.

[0021] Preferably, the outputs on the convolutional neural network and the multilayer perceptron module are fused, and the fused features are input into the decision module for output of the abnormal detection result, specifically including: the output of the convolutional neural network module and the output of the multilayer perceptron module are fused through the full connection layer, and after two full connections and activation layers, the output is input into the decision module for output of the abnormal detection result.

[0022] In another aspect, the present application provides a device for distributed optical fiber acoustic anomaly detection, comprising: a circulator, a first erbium-doped fiber amplifier, a fiber coupler, a polarization diversity receiving module, a photodetector and a processing unit, specifically: one of the output ports of the circulator is connected with the input port of the fiber coupler, the first output port of the fiber coupler is connected with the polarization diversity receiving module, the second output port of the fiber coupler is connected with the photodetector, and the output port of the polarization diversity receiving module and the output port of the photodetector are both connected with the processing unit; wherein the RBS optical signal of the chirped optical pulse passes through the circulator and the first erbium-doped fiber amplifier, and is divided into two optical signals through the fiber coupler; the first optical signal is decomposed into two orthogonal X polarization states and Y polarization states through the polarization diversity receiving module, and is mixed with the reference optical signal for balanced reception to obtain X-state electrical signal and Y-state electrical signal; the second optical signal is converted into an intensity electrical signal through the photodetector; the processing unit is used for obtaining the phase curve of the distributed optical fiber along the line according to the X-state electrical signal and the Y-state electrical signal, and obtaining the intensity curve according to the intensity electrical signal; and is further used for inputting the phase curve into a convolutional neural network to obtain vibration features, inputting the intensity curve into a multilayer perceptron module to obtain fading features, and inputting the vibration features and the fading features into a decision module after fusion to output the anomaly detection result.

[0023] Preferably, the polarization diversity receiving module specifically comprises: a polarization diversity coherent receiver, a first amplification circuit and a second amplification circuit, specifically; the first output port of the fiber coupler is connected with the first input port of the polarization diversity coherent receiver, the second input port of the polarization diversity coherent receiver is used for inputting the reference light; the first output port of the polarization diversity coherent receiver is connected with the first amplification circuit, used for amplifying and outputting the X-state electrical signal; and the second output port of the polarization diversity coherent receiver is connected with the second amplification circuit, used for amplifying and outputting the Y-state electrical signal.

[0024] Preferably, the device further comprises a laser, a polarization maintaining fiber coupler and a second erbium-doped fiber amplifier, the laser is connected with the input end of the polarization maintaining fiber coupler, one of the output ends of the polarization maintaining fiber coupler is connected with the second erbium-doped fiber amplifier, and the other output end is connected with one of the input ends of the polarization diversity receiving module; and the output end of the second erbium-doped fiber amplifier is connected with the circulator

[0025] Compared with the prior art, the embodiment of the present application has the beneficial effects that: the RBS optical signal is decomposed into X-polarization state electrical signal and Y-polarization state electrical signal to identify vibration features, and the fading features are extracted through the intensity signal, and then the vibration features and the fading features are fused for judgment and detection, and the influence of interference fading is eliminated according to the fading features. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments of the present application will be briefly introduced. Obviously, the drawings described below are only some of the embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0027] Figure 1 It is a schematic diagram of a typical architecture of the existing Φ-OTDR;

[0028] Figure 2 It is a schematic diagram of the existing neural network structure;

[0029] Figure 3 It is a flow chart of a distributed optical fiber acoustic anomaly detection method provided by the embodiments of the present application;

[0030] Figure 4 It is a flow chart of another distributed optical fiber acoustic anomaly detection method provided by the embodiments of the present application;

[0031] Figure 5 It is a flow chart of another distributed optical fiber acoustic anomaly detection method provided by the embodiments of the present application;

[0032] Figure 6 It is a schematic diagram of the neural network structure used by the embodiments of the present application;

[0033] Figure 7 It is a schematic diagram of a distributed optical fiber acoustic anomaly detection device structure provided by the embodiments of the present application;

[0034] Figure 8 It is a schematic diagram of another distributed optical fiber acoustic anomaly detection device structure provided by the embodiments of the present application. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application, and do not limit the present application.

[0036] The present application is a system architecture of a specific function system, so in the specific embodiments, the functional logical relationship of each structure module is mainly described, and the specific software and hardware implementation is not limited.

[0037] In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as there is no conflict between them. The present application will be described in detail below in combination with the drawings and embodiments.

[0038] Embodiment 1:

[0039] The embodiment aims at the influence of interference fading on the existing coherent detection based phi-OTDR system, and proposes an abnormality detection method for mixed detection of different polarization state signals and intensity signals, so as to eliminate the influence of interference fading.

[0040] As shown in Figure 3 The specific steps of the distributed optical fiber acoustic anomaly detection method provided by the embodiment are as follows:

[0041] Step 101: Obtain the X-state electric signal and Y-state electric signal after mixing of the RBS optical signal from a chirp optical pulse and reference light, and obtain the intensity electric signal of the RBS optical signal.

[0042] In order to eliminate the interference fading effect, in the embodiment, a polarization diversity coherent receiving module is used to obtain the X-state electric signal and Y-state electric signal after mixing of the RBS optical signal and the reference light. An intensity detection module is used to obtain the intensity electric signal of the RBS optical signal.

[0043] Step 102: Obtain the phase curve of the distributed optical fiber along the line according to the X-state electric signal and Y-state electric signal, and obtain the intensity curve according to the intensity electric signal.

[0044] Since the RBS optical signal is the return optical signal of different positions along the distributed optical fiber, the return time point of the optical signal corresponds to the return position, so the phase electric signal and the intensity electric signal arranged in time sequence can correspond to the positions along the distributed optical fiber one by one. Therefore, after obtaining the X-state electric signal, the Y-state electric signal and the intensity signal, the X-state electric signal and the Y-state electric signal need to be sorted according to time, and demodulated into phase signals, and finally the phase curve sorted according to time is obtained, and the intensity curve sorted according to time is corresponded point by point.

[0045] Step 103: input the phase curve into a convolutional neural network to obtain vibration features, input the intensity curve into a multilayer perceptron module to obtain fading features, and input the vibration features and the fading features after fusion into a decision module to output the abnormality detection result.

[0046] After obtaining the phase curve and the intensity curve, the neural network can be used for corresponding feature extraction, and the vibration features and the fading features are fused. Since the intensity signal is obtained by a non-coherent intensity detection module, it is not affected by interference fading. When interference fading occurs, the phase curve will produce abnormal jitter, while the intensity signal will not produce abnormal jitter, so the fading features can be extracted. Through the correspondence of the vibration features and the fading features, the decision module can eliminate the fading interference in the vibration features through the fading features, and obtain accurate abnormality detection results.

[0047] After the steps 101-103 provided in the embodiment, the influence of the fading interference can be effectively eliminated, and an accurate detection result not affected by the influence can be obtained.

[0048] Based on the above technical solution, the specific implementation process of the method is described below in combination with actual data in a certain specific scene. In actual implementation, the following specific implementation can be adjusted according to the principle of the distributed optical fiber acoustic anomaly detection method in the embodiment according to actual conditions.

[0049] The laser in the Φ-OTDR device emits a chirp light pulse as probe light, and the RBS of the chirp light pulse is collected by the signal collection unit. The RBS contains vibration feature information of different positions along the distributed optical fiber.

[0050] The RBS signal from the chirp light pulse is mixed with reference light in two orthogonal X polarization states and Y polarization states respectively for balanced reception, and the beat frequency electrical signal of the X polarization state is obtained as the X state electrical signal, and the beat frequency electrical signal of the Y polarization state is obtained as the Y state electrical signal.

[0051] In the scenario provided in the embodiment, the X state electrical signal obtained by the RBS through coherent reception is represented as {x(k); k=1,....125000}, the Y state electrical signal is represented as {y(k); k=1,....125000}, and the intensity electrical signal directly detected by the received signal is represented as {s(k); k=1,....125000}. Wherein, k represents the number of sampling points, the sampling rate of the signal collection unit is 250MSPS, the unit length of each sampling point is about 0.4m, and the total sampling length is 50km.

[0052] After obtaining the X state electrical signal and the Y state electrical signal, all the X state electrical signals and the Y state electrical signals need to be sorted according to time for one-to-one correspondence, so that the phase characteristics and attenuation information at the same time point can be corresponded in subsequent processing.

[0053] Further, in order to facilitate data processing, the sorted data also needs to be normalized.

[0054] Firstly, the sorted X state electrical signal and Y state electrical signal are respectively obtained by using the Hilbert phase demodulation algorithm to obtain the corresponding phase signal, and the demodulated phase signal sequence is obtained by averaging and superimposing.

[0055] As shown in Figure 4 , the phase signal sequence can be obtained by the following steps.

[0056] Step 201: Calculate the Hilbert transform of the X-state electric signal and the Y-state electric signal point by point, respectively, to obtain the Hilbert transform sequence of the X-state and the Hilbert transform sequence of the Y-state, and calculate the arctangent of the Hilbert transform sequence of the X-state and the Hilbert transform sequence of the Y-state respectively.

[0057] In the scenario of the embodiment, the Hilbert transform of the X-state electric signal and the Y-state electric signal is calculated point by point to obtain {H(x(k)); k=1,....125000} and {H(y(k)); k=1,....125000}. The arctangent of H(x) and x, H(y) and y can be calculated using the following formula 1 and formula 2 respectively.

[0058]

[0059]

[0060] Step 202: For the arctangent of the X-state and the arctangent of the Y-state, calculate the mean value point by point to obtain the phase signal corresponding to each point, and sort the phase signals frame by frame to obtain a phase signal sequence.

[0061] The phase signal sequence obtained in step 201 is and The mean value of each point is calculated to obtain the phase signal value (k) corresponding to each point. Then, all phase signal values (k) are sorted frame by frame to obtain a phase signal sequence (n, k), where k represents different distances on the optical fiber, and n represents different RBS pulses. The repetition period of the pulse is 1 KHz, so in 1 second, n=1, 2, 3,..., 1000.

[0062] After the steps 201-202 provided in the embodiment, the phase signal sequence required in the step can be obtained.

[0063] Then, the phase sequence is normalized to form a phase curve. In the scenario of the embodiment, the phase curve { (n, k); n=1,....1000} within each second can be normalized according to each k. For example, the phase curve can be normalized according to the maximum and minimum using formula 1.

[0064]

[0065] On the other hand, as shown in Figure 5 , an intensity curve also needs to be obtained according to the intensity electric signal.

[0066] Step 301: The intensity electric signal obtained by sampling is sorted frame by frame. ​

[0067] In the scenario of the embodiment, the intensity signal s(k) obtained by sampling is sorted to obtain an intensity signal sequence s(n, k), which corresponds to a phase signal sequence, k represents different distances on the optical fiber, and n represents different RBS pulses. The repetition period of the pulses is 1 KHz, so n = 1, 2, 3,..., 1000 in 1 second.

[0068] Step 302: Convert the intensity electrical signal into signal-to-noise ratio data and take the logarithm to normalize, to form an intensity curve.

[0069] According to each k, the intensity signal sequence {s(n, k); n = 1,..., 1000} in 1 second is normalized to obtain an intensity curve. Specifically, the intensity curve can be normalized by converting the intensity curve into signal-to-noise ratio data and taking the logarithm using formula 4.

[0070]

[0071] After steps 301-302 provided in the embodiment, the intensity curve corresponding to the phase curve can be obtained.

[0072] After the above process, the original data is obtained and preprocessed, and the phase curve and the intensity curve are obtained. Then, the phase curve and the intensity curve can be fused and processed using a neural network, and features therein are recognized to achieve anomaly detection.

[0073] When performing pattern recognition, the vibration features in the phase curve need to be recognized to obtain the abnormal features contained therein, and the fading features in the intensity curve need to be recognized to eliminate the interference fading in the phase curve at the corresponding time point. Specifically, the phase signal sequence is input into a convolutional neural network module to extract vibration features, to obtain a vibration feature vector; the intensity signal sequence is input into a multilayer perceptron module to extract fading features, to obtain a fading feature vector. The pattern recognition result obtained by the classifier module is output as an anomaly detection result.

[0074] After the vibration features and the fading features are extracted by pattern recognition, the outputs of the convolutional neural network module and the multilayer perceptron module are fused through a fully connected layer, and after two fully connected and activation layers, the outputs are input into a decision module to output an anomaly detection result.

[0075] In a specific scenario, the neural network can be used as Figure 6As shown in the neural network model, the convolutional neural network module comprises a plurality of convolutional layers, pooling layers and activation layer networks; the multilayer perceptron module comprises a plurality of fully connected layers and activation layer networks. Specifically, the normalized phase curve is input into the convolutional neural network module for feature extraction, and the convolutional neural network module comprises 12 layers of convolutional layers, pooling layers and activation layer networks arranged repeatedly. The normalized intensity curve is input into the multilayer perceptron module for feature extraction, and the multilayer perceptron module comprises 6 layers of fully connected layers and activation layer networks arranged repeatedly. Then, the output of the convolutional neural network module and the output of the multilayer perceptron module are fused through a fully connected layer, and after passing through two fully connected and activation layers, they are input into the decision module for output of the abnormality detection result. In other implementation scenarios, appropriate neural network models and appropriate parameters can also be selected as needed.

[0076] The method for distributed optical fiber acoustic wave anomaly detection provided in this embodiment extracts features from phase signals and intensity signals respectively, obtains corresponding vibration features and fading features, and then fuses them, so as to eliminate the interference fading of the vibration features at the corresponding time point by using the fading features, and simply and effectively eliminate the influence of interference fading on the distributed optical fiber acoustic wave anomaly detection.

[0077] Embodiment 2

[0078] Based on the method for distributed optical fiber acoustic wave provided in embodiment 1, this embodiment further provides a device for distributed optical fiber acoustic wave anomaly detection, which is used to obtain the X-state electrical signal, the Y-state electrical signal and the intensity electrical signal in step 101.

[0079] As shown in Figure 7 In order to decompose the RBS optical signal into two polarization state electrical signals, the device provided in this embodiment uses a polarization diversity receiving module, and on the basis of Figure 1 As shown in the Ф-OTDR, the device provided in this embodiment comprises a circulator, a first erbium-doped fiber amplifier, a fiber coupler, a polarization diversity receiving module, a photodetector, a first amplification circuit and a processing unit. Specifically, one output port of the circulator is connected with the input port of the fiber coupler through the first erbium-doped fiber amplifier, the first output port of the fiber coupler is connected with the polarization diversity receiving module, the second output port of the fiber coupler is connected with the photodetector, the output port of the polarization diversity receiving module is connected with the processing unit, and the output port of the photodetector is connected with the processing unit through the first amplification circuit.

[0080] The RBS optical signal of the chirped optical pulse is split into two optical signals through a circulator and a first erbium-doped fiber amplifier via a fiber coupler. After balanced reception by mixing the first optical signal with a reference optical signal through a polarization diversity receiving module, the first optical signal is decomposed into two orthogonal X polarization states and Y polarization states to obtain X-state electrical signals and Y-state electrical signals. The second optical signal is converted into an intensity electrical signal through a photoelectric detector and a first amplification circuit. The processing unit is configured to obtain a phase curve of the distributed optical fiber along the line according to the X-state electrical signals and the Y-state electrical signals, and obtain an intensity curve according to the intensity electrical signal; and is further configured to input the phase curve into a convolutional neural network to obtain vibration features, input the intensity curve into a multilayer perceptron module to obtain fading features, and input the vibration features and the fading features into a decision module after fusion to output an abnormality detection result.

[0081] In a specific implementation scenario, as shown in Figure 8 The polarization diversity receiving module specifically includes a polarization diversity coherent receiver, a second amplification circuit and a third amplification circuit. Specifically, the first output port of the fiber coupler is connected to the first input port of the polarization diversity coherent receiver, and the second input port of the polarization diversity coherent receiver is configured to input a reference light; the first output port of the polarization diversity coherent receiver is connected to the second amplification circuit, and the second amplification circuit is configured to amplify and output the X-state electrical signals. The second output port of the polarization diversity coherent receiver is connected to the third amplification circuit, and the third amplification circuit is configured to amplify and output the Y-state electrical signals. Through this optical path structure, the X-state electrical signals and the Y-state electrical signals in the RBS optical signal can be extracted.

[0082] In a specific implementation scenario, the device provided by the embodiment further includes a laser, a polarization maintaining fiber coupler, an acousto-optic modulator and a second erbium-doped fiber amplifier. The laser is connected to the input end of the polarization maintaining fiber coupler, one of the output ends of the polarization maintaining fiber coupler is connected to the second erbium-doped fiber amplifier, the other output end is connected to one of the input ends of the polarization diversity receiving module, and the output end of the second erbium-doped fiber amplifier is connected to the circulator. The laser is a narrow linewidth laser, which can output a higher quality optical signal to meet the needs of long distance detection. The laser outputs a coherent light source for detection to the first port of the polarization maintaining fiber coupler, the second port of the polarization maintaining fiber coupler outputs the detection light to the acousto-optic modulator to generate a chirped optical pulse signal, and the third port of the polarization maintaining fiber coupler outputs a reference light to the polarization diversity coherent receiving device. The pulse optical signal output by the acousto-optic modulator is amplified by the erbium-doped fiber amplifier and then input to the first port of the circulator, and the second port of the circulator injects the pulse optical signal into the sensing optical fiber.

[0083] The chirped optical pulse signal generates RBS on the sensing optical fiber. When the sensing optical fiber vibrates, the phase signal of the RBS also has corresponding vibration features. Meanwhile, the RBS also has corresponding fading features when it returns on the sensing optical fiber.

[0084] The RBS signal generated by the sensing fiber passes through the second port of the circulator, is input into the erbium-doped fiber amplifier from the third port for signal amplification, the first port of the coupler is connected to the output port of the erbium-doped fiber amplifier 7, and the second port of the coupler inputs the amplified and split RBS signal into the polarization diversity coherent receiving device.

[0085] The polarization diversity coherent receiving device respectively mixes and receives the RBS signal from the chirped optical pulse with reference light according to two orthogonal X polarization states and Y polarization states to obtain beat frequency electrical signals of the two polarization states, respectively outputs one X state electrical signal and one Y state electrical signal, and after passing through a first amplification circuit and a second amplification circuit, the X state electrical signal and the Y state electrical signal are received, collected and demodulated by a signal processing unit.

[0086] The third port of the coupler inputs the amplified and split RBS signal into a photodetector for receiving, the photodetector converts the received RBS optical signal into an electrical signal, and after passing through an amplification circuit, the electrical signal is received, collected and demodulated by a signal processing unit to obtain an intensity signal sequence.

[0087] The signal collection processing unit receives the X and Y two-way electrical signals from the polarization diversity coherent receiving device and the intensity electrical signal, and according to the method provided in Embodiment 1, finally obtains distributed optical fiber acoustic wave anomaly detection data.

[0088] Corresponding to the actual implementation scene in Embodiment 1, the device provided in the present embodiment can select the following devices and parameters: the photodetector 12 is an avalanche photodetector, converts the received RBS optical signal into an electrical signal, and after passing through an amplification circuit, the intensity signal sequence is received by the signal collection processing unit. The sensing fiber is a non-polarization maintaining single-mode fiber with a length of 50 kilometers. The signal processing unit can also control the acousto-optic modulator to generate a pulse sequence to pulse modulate the continuous light signal output by the coherent light source. The typical pulse width is 200 ns, and the pulse repetition frequency is 1 kHz. The working frequency of the acousto-optic modulator is 200M. The amplification gain of each amplification circuit in the device is adjustable, and the range is 0-30dB. The sampling rate of the signal collection processing unit is 250MSPS, and the quantization resolution is 14 bits. In specific implementation, other devices and parameters can also be selected as needed to meet the functional needs of the devices in the above technical solutions

[0089] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for distributed optical fiber acoustic anomaly detection, characterized in that, Specifically, it includes: The X-state electrical signal and Y-state electrical signal of the RBS optical signal from a chirped light pulse are obtained after mixing with the reference light, and the intensity electrical signal of the RBS optical signal is obtained through a photodetector. The phase curves along the distributed optical fiber are obtained based on the X-state electrical signals and the Y-state electrical signals, and the intensity curves are obtained based on the intensity electrical signals. The process involves inputting the phase curve into a convolutional neural network to obtain vibration features, and inputting the intensity curve into a multilayer perceptron module to obtain fading features. The vibration and fading features are then fused and input into a decision module for anomaly detection output. This includes: inputting the phase signal sequence into the convolutional neural network module for vibration feature extraction to obtain a vibration feature vector; inputting the intensity signal sequence into the multilayer perceptron module for fading feature extraction to obtain a fading feature vector; inputting the vibration and fading feature vectors into a connection layer, and then passing them through a classifier module to obtain a pattern recognition result, which is output as the anomaly detection result; fusing the outputs of the convolutional neural network module and the multilayer perceptron module through a fully connected layer, and then inputting them into the decision module for anomaly detection output after passing through two fully connected and activation layers.

2. The method for distributed optical fiber acoustic anomaly detection according to claim 1, characterized in that, The acquisition of the X-state electrical signal and Y-state electrical signal after mixing the RBS optical signal from a chirped optical pulse with the reference light specifically includes: The RBS signal from the chirped light pulse is mixed and balanced with the reference light according to two orthogonal X-polarization states and Y-polarization states, respectively. The beat frequency electrical signal of the X-polarization state is used as the X-state electrical signal, and the beat frequency electrical signal of the Y-polarization state is used as the Y-state electrical signal.

3. The method for distributed optical fiber acoustic anomaly detection according to claim 1, characterized in that, The process of obtaining the phase curve along the distributed optical fiber based on the X-state electrical signal and the Y-state electrical signal specifically includes: All X-state and Y-state electrical signals are sorted by time and matched one-to-one. The sorted X-state and Y-state electrical signals are used to obtain the corresponding phase signals by the Hilbert phase demodulation algorithm, and then averaged and superimposed to obtain the demodulated phase signal sequence. The phase sequence is then normalized to form a phase curve.

4. The method for distributed optical fiber acoustic anomaly detection according to claim 3, characterized in that, The process of obtaining the corresponding phase signals using the Hilbert phase demodulation algorithm and then averaging and superimposing them to obtain the demodulated phase signal sequence specifically includes: Calculate the Hilbert transform of the X-state and Y-state electrical signals point by point to obtain the Hilbert transform sequences of the X-state and Y-state respectively, and calculate the arctangent of the Hilbert transform sequences of the X-state and Y-state respectively. For the arctangent of the X state and the arctangent of the Y state, the mean value is calculated point by point to obtain the phase signal corresponding to each point, and the phase signals are sorted frame by frame to obtain the phase signal sequence.

5. The method for distributed optical fiber acoustic anomaly detection according to claim 1, characterized in that, The process of obtaining the intensity curve based on the intensity electrical signal specifically includes: The sampled intensity electrical signals are sorted frame by frame. The intensity electrical signal is converted into signal-to-noise ratio data and its logarithm is taken for normalization to form an intensity curve.

6. A device for distributed optical fiber acoustic anomaly detection, characterized in that, include: The components include a circulator, a first erbium-doped fiber amplifier, a fiber coupler, a polarization diversity receiver module, a photodetector, a first amplifier circuit, and a processing unit. Specifically: One of the output ports of the circulator is connected to the input port of the fiber coupler via a first erbium-doped fiber amplifier. The first output port of the fiber coupler is connected to the polarization diversity receiving module. The second output port of the fiber coupler is connected to the photodetector. The output port of the polarization diversity receiving module is connected to the processing unit. The output port of the photodetector is connected to the processing unit via a first amplification circuit. Among them, the RBS optical signal of the chirped optical pulse passes through a circulator and the first erbium-doped fiber amplifier, and is split into two optical signals by an optical fiber coupler; After the first optical signal is mixed and balanced with the reference optical signal by the polarization diversity receiving module, it is decomposed into two orthogonal X-polarization states and Y-polarization states, resulting in X-state electrical signals and Y-state electrical signals. The second optical signal is converted into an intensity electrical signal by a photodetector and the first amplifier circuit; The processing unit is used to obtain the phase curve along the distributed optical fiber based on the X-state electrical signal and the Y-state electrical signal, and to obtain the intensity curve based on the intensity electrical signal; it is also used to input the phase curve into a convolutional neural network to obtain vibration features, input the intensity curve into a multilayer perceptron module to obtain fading features, and fuse the vibration features and fading features before inputting them into a decision module for anomaly detection result output; the processing unit is used to input the phase signal sequence into the convolutional neural network module for vibration feature extraction to obtain a vibration feature vector; input the intensity signal sequence into the multilayer perceptron module for fading feature extraction to obtain a fading feature vector; input the vibration feature vector and the fading feature vector into a connection layer, and after passing through a classifier module to obtain a pattern recognition result as an anomaly detection result for output; fuse the output of the convolutional neural network module and the output of the multilayer perceptron module through a fully connected layer, and after passing through two fully connected and activation layers, input them into the decision module for anomaly detection result output.

7. The apparatus for distributed optical fiber acoustic anomaly detection as described in claim 6, characterized in that, The polarization diversity receiving module specifically includes: a polarization diversity coherent receiver, a second amplifier circuit, and a third amplifier circuit; specifically; The first output port of the fiber optic coupler is connected to the first input port of the polarization diversity coherent receiver, and the second input port of the polarization diversity coherent receiver is used to input reference light; The first output port of the polarization diversity coherent receiver is connected to the second amplifier circuit, which amplifies the X-state electrical signal and outputs it. The second output port of the polarization diversity coherent receiver is connected to a third amplifier circuit, which amplifies the Y-state electrical signal before outputting it.

8. The apparatus for distributed optical fiber acoustic anomaly detection as described in claim 6, characterized in that, The device further includes a laser, a polarization-maintaining fiber coupler, an acousto-optic modulator, and a second erbium-doped fiber amplifier. The laser is connected to the input end of the polarization-maintaining fiber coupler, one output end of the polarization-maintaining fiber coupler is connected to the second erbium-doped fiber amplifier, and the other output end is connected to one input end of the polarization diversity receiving module. The output of the second erbium-doped fiber amplifier is connected to the circulator.

Images