Method and device for eliminating crosstalk in sensor channels in fiber optic interferometer sensor network

Abstract

The present invention proposes a method and device for eliminating crosstalk in a sensor channel of a fiber interferometer sensor network. By constructing a deep learning neural network, a loss function of the neural network is constructed by coupling physical detuning with autocorrelation of time domain signals, and the coefficient relationship between the main interference term and the parasitic interference term in the fiber optic sensor network is trained and learned. The processing model is distilled and solidified into the signal processing front end of the fiber optic interferometer sensor network. The purpose is to significantly improve the anti-interference ability of the fiber optic interferometer network without increasing the hardware cost and complexity of the system, so as to provide support for improving its sensing performance.

Description

Technical Field

[0001] The present invention mainly relates to the field of optical fiber sensing technology, in particular to a method and a device for eliminating crosstalk in a sensing channel of an optical fiber interferometer sensing network. Background Art

[0002] In the underwater and space fields, various types of fiber optic interferometers are used as basic sensing units, and sensor networks are constructed through time division, wavelength division and other multiplexing methods. This sensing structure has the advantages of high sensitivity, anti-electromagnetic interference, and easy large-scale networking. It is especially suitable for high-precision weak signal sensing in relatively harsh environments, such as the acquisition of weak underwater acoustic signals. This sensor network usually uses time division, wavelength division and other multiplexing technologies to form a large-scale array to enhance the detection capability of weak signals.

[0003] In the fiber interferometer sensor network, inter-channel interference is a key indicator that affects the system's sensing performance. Parasitic interference is the root cause of interference between channels in the fiber interferometer multiplexing network, which has been widely verified by experiments. At present, the parasitic interference suppression methods can be summarized into two categories: one is to control the coherence length by adding an optical modulation module, so that the original parasitic interference term no longer has coherence characteristics. This method can be used to query multiple sensor network channels at the same time, but the cost is that the hardware cost of the system increases, the modulation and demodulation parameters increase, and the technical complexity of the system increases. One is to form an algorithm to eliminate parasitic interference terms by establishing a theoretical model of parasitic interference. The advantage is that the hardware cost and the complexity of the modulation system will not increase when the software algorithm is used to solve the problem. The cost is that the physical mechanism model can only solve the crosstalk problem of a small number of channels, and its effect will decrease as the scale of the sensor network expands. Summary of the invention

[0004] In view of the technical problems existing in the prior art, the present invention proposes a method and a device for eliminating crosstalk in a sensing channel of a fiber optic interferometer sensing network.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] In one aspect, the present invention provides a method for eliminating crosstalk in a sensing channel of a fiber interferometer sensing network, comprising:

[0007] Applying single-frequency sinusoidal signals of different frequencies to the link transmission optical fibers and each sensing optical fiber in the optical fiber interferometer sensing network to obtain the original interference signals of all sensing channels in the optical fiber interferometer sensing network;

[0008] The original interference result of each sensing channel is synthesized in the complex domain by a phase-generated carrier wave, so as to obtain the complex domain expression of the original interference result of each sensing channel, that is, to obtain the complex interference signal of each sensing channel;

[0009] Demodulate the complex interference signal of each sensing channel to obtain the time domain and frequency domain results of the demodulated signal of each sensing channel;

[0010] Compare the real frequency actually applied to each sensing channel, suppress the line spectrum of the non-real frequency in the frequency domain result of the demodulated signal corresponding to each sensing channel to the background noise level, and obtain the target frequency domain result of the demodulated signal corresponding to each sensing channel;

[0011] Perform inverse Fourier transform on the suppressed signal frequency results corresponding to each sensor channel to obtain the demodulated signal target time domain results corresponding to each sensor channel;

[0012] The complex interference signal of each sensor channel is input as training data into the pre-built neural network model to predict the demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel; the predicted demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel are autocorrelated with the target time domain result and demodulation signal target frequency domain result corresponding to each sensor channel, and the absolute value of the autocorrelation coefficient is used as the loss function. The network parameters are iteratively learned according to the gradient descent method until the model converges, the iteration is stopped, the current network parameters are saved, and a distillation model is formed;

[0013] The trained distillation model is used to reconstruct the original interference signals of all interference channels in the fiber optic interferometer sensor network and eliminate the interference of the fiber optic interferometer sensor network.

[0014] Preferably, the single-frequency sinusoidal signals of different frequencies applied to the transmission optical fiber and each sensing optical fiber in the optical fiber interferometer sensing network are within the response frequency band of the optical fiber interferometer sensing network, and the interval between any two single-frequency sinusoidal signals is not less than 10 Hz.

[0015] Preferably, the fiber interferometer sensing network has (N-1) sensing channels, the neural network model contains an input layer, (N-1) hidden layers and an output layer, the number of neurons in the output layer is (N-1), the number of neurons in each hidden layer is: [(N-1)+K], K is the serial number of the hidden layer, wherein the [(N-1)+K]th neuron in the Kth hidden layer is equal to the complex conjugate of the Kth neuron in the (K-1)th hidden layer;

[0016] The complex interference signal of each sensor channel is used as the input of the input layer, and the output layer re-outputs (N-1) neurons. Each output neuron represents the reconstructed interference signal. The output interference signal is demodulated using the random phase fading suppression demodulation method to obtain the predicted time domain results and frequency domain results of the demodulated signal corresponding to each sensor channel.

[0017] Preferably, by setting a user threshold , when the absolute value of the loss function is less than When , the iteration stops, the current network parameters are saved, and the distillation model is formed.

[0018] Furthermore, the present invention also includes evaluating the interference suppression effect of the trained distillation model based on the test data, including:

[0019] Applying single-frequency sinusoidal signals of different frequencies to the transmission optical fiber and each sensing optical fiber in the optical fiber interferometer sensor network, wherein the signal frequency of each single-frequency sinusoidal signal is different from the signal frequency corresponding to the training data, to obtain the original interference signals of all sensing channels in the optical fiber interferometer sensor network;

[0020] The original interference results of each sensing channel are synthesized in the complex domain by a phase-generated carrier wave, so as to obtain the complex domain expression of the original interference results of each sensing channel, that is, to obtain the complex interference signals of each sensing channel as test data;

[0021] The test data is input into the distillation model, and the distillation model outputs the reconstructed interference signal, which is demodulated using the random phase fading suppression demodulation method. If the interference signal between the sensor channels after output is lower than -40dB, it is considered that the trained distillation model is effective in suppressing interference between sensor channels.

[0022] On the other hand, a device for eliminating crosstalk of sensing channels in a fiber optic interferometer sensing network comprises:

[0023] The first module is used to apply single-frequency sinusoidal signals of different frequencies to the link transmission optical fiber and each sensing optical fiber in the optical fiber interferometer sensing network, and simultaneously obtain the original interference signals of all sensing channels in the optical fiber interferometer sensing network;

[0024] The second module is used to use a complex domain synthesis method based on phase-generated carriers to obtain the complex domain expression of the original interference results of each sensing channel, that is, to obtain the complex interference signals of each sensing channel;

[0025] The third module is used to demodulate the complex interference signal of each sensing channel to obtain the time domain and frequency domain results of the demodulated signal of each sensing channel;

[0026] The fourth module is used to compare the real frequency actually applied to each sensing channel, suppress the line spectrum of the non-real frequency in the frequency domain result of the demodulated signal corresponding to each sensing channel to the background noise level, and obtain the target frequency domain result of the demodulated signal corresponding to each sensing channel;

[0027] The fifth module is used to perform inverse Fourier transform on the suppressed signal frequency results corresponding to each sensing channel to obtain the demodulated signal target time domain results corresponding to each sensing channel;

[0028] The sixth module is used to input the complex interference signal of each sensor channel as training data into the pre-built neural network model, and predict the demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel; autocorrelate the predicted demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel with the target time domain result and demodulation signal target frequency domain result corresponding to each sensor channel, use the absolute value of the autocorrelation coefficient as the loss function, perform iterative learning of network parameters according to the gradient descent method, until the model converges, stop iteration, save the current network parameters, and form a distillation model;

[0029] The seventh module is used to reconstruct the original interference signals of all interference channels in the fiber optic interferometer sensor network using the trained distillation model, thereby eliminating the interference of the fiber optic interferometer sensor network.

[0030] Further, the fiber optic interferometer sensor network includes a sensor network modulation and demodulation module, a transmission link unit and a fiber optic interferometer sensor network unit;

[0031] The sensor network modulation and demodulation module includes a narrow line width laser, an acousto-optic modulator, a matching interferometer, a polarization modulator, a photoelectric conversion and acquisition module and a main control module;

[0032] The narrow linewidth laser is connected to the input port of the acousto-optic modulator through an optical fiber; the output port of the acousto-optic modulator is connected to the input port of the matching interferometer through an optical fiber; the output port of the matching interferometer is connected to the input port of the polarization modulator through an optical fiber; the photoelectric conversion and acquisition module is electrically connected to the main control module; the main control module is connected to the acousto-optic modulator, the matching interferometer and the polarization modulator through electric wires, and is connected to the host computer through an Ethernet cable, wherein the narrow linewidth laser is used to generate continuous laser, the acousto-optic modulator is used to modulate the laser into a pulse laser, the polarization modulator is used to adjust the polarization state of each pulse in the pulse pair, the photoelectric conversion and acquisition module is used to perform photoelectric conversion and sampling on the optical pulse signal returned by the fiber interferometer sensor network unit, the main control module is used to generate modulation signals for the acousto-optic modulator, the matching interferometer and the polarization modulator, and the main control module is used to generate a synchronous trigger signal for the photoelectric conversion and acquisition module;

[0033] A transmission link unit, used to connect the sensor network modulation and demodulation module and the fiber optic interferometer sensor network unit;

[0034] The fiber interferometer sensing network unit is used to sense external signals, and includes a first node, a first sensing fiber, a second node, a second sensing fiber, a third node, ..., the (N-1)th node, the (N-1)th sensing fiber, and the Nth node, wherein the first node, the first sensing fiber, and the second node constitute a fiber interferometer, the second node, the second sensing fiber, and the third node constitute a fiber interferometer, and so on, the (N-1)th node, the (N-1)th sensing fiber, and the Nth node constitute a fiber interferometer. The fiber interferometer sensing network unit includes a total of (N-1) fiber interferometers, that is, it has (N-1) sensing channels, wherein the first node, the second node, the third node, ..., the (N-1)th node, and the Nth node are fiber gratings or fiber couplers and Faraday rotators.

[0035] Furthermore, the transmission link unit includes a fiber circulator and a link transmission optical fiber, the fiber circulator has three ports, namely an input port, a first output port, and a second output port; the output port of the polarization modulator is connected to the input port of the fiber circulator through an optical fiber; the first output port of the fiber circulator is connected to the link transmission optical fiber, and the link transmission optical fiber is a single-mode optical fiber; the second output port of the fiber circulator is connected to the photoelectric conversion and acquisition module.

[0036] Further, the matching interferometer includes an input polarization-maintaining fiber coupler, an output polarization-maintaining fiber coupler, an interferometer long-arm polarization-maintaining fiber and an interferometer short-arm polarization-maintaining fiber, wherein the interferometer long-arm polarization-maintaining fiber is partially wound around a piezoelectric ceramic to form a PZT fiber phase modulator;

[0037] The matching interferometer is used to convert a pulse into a pair of pulses, and the pulse time interval is exactly the same as the time it takes for light to travel back and forth between two adjacent nodes in the fiber interferometer sensor network unit, and a phase-generated carrier modulation signal is introduced through a PZT fiber phase modulator. ,in is the modulation amplitude, is the modulation frequency.

[0038] The present invention proposes a method and device for eliminating inter-channel interference by reconstructing interference terms using a neural network. Compared with the prior art, the technical effects of the present invention are:

[0039] (1) The present invention uses a neural network to reconstruct interference terms, thereby eliminating parasitic interference terms caused by Rayleigh scattering and reflected light from other stray points in the fiber interferometer network, including link interference introduced by the transmission optical fiber and interference transmitted from front to back in the fiber interferometer network, thereby improving the anti-interference ability of the fiber interferometer sensor network and providing node output with better signal-to-noise ratio for subsequent possible network signal fusion processing.

[0040] (2) The present invention adopts the method of reconstructing interference terms to achieve the purpose of eliminating interference. There is no special restriction on the frequency and amplitude of the interference signal, and it has a suppressing effect on both single-frequency interference signals and broadband interference signals.

[0041] (3) The present invention relies on a neural network algorithm to obtain the law and coefficients of parasitic interference terms in the fiber interferometer network through large-scale data learning, avoiding the traditional method of introducing a large number of fiber modulators to control the coherence length of the extracavity laser and reduce parasitic interference. The cost, volume, power consumption and reliability of the system can be greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0043] Figure 1 A schematic diagram of the structure of a fiber optic interferometer sensor network provided by an embodiment;

[0044] Figure 2 A schematic diagram of the structure of a matching interferometer provided by an embodiment;

[0045] Figure 3 A schematic diagram of the structure of an optical fiber circulator provided by an embodiment;

[0046] Figure 4 is a flow chart of a method for eliminating crosstalk in a sensor channel in a fiber interferometer sensor network provided by an embodiment;

[0047] Figure 5 A neural network framework structure diagram suitable for a fiber optic interferometer sensor network provided in one embodiment.

[0048] Numbers in the figure:

[0049] 1 is a sensor network modulation and demodulation module, 101 is a narrow line width light source, 102 is an acousto-optic modulator, 103 is a matching interferometer, 104 is a polarization modulator, 105 is a photoelectric conversion and acquisition module, 106 is a main control module, and 107 is a host computer;

[0050] 2 is a transmission link unit, 201 is an optical fiber circulator, and 202 is a link transmission optical fiber;

[0051] 3 is a fiber interferometer sensor network, 301 is the first node, 302 is the first sensing fiber, 303 is the second node, 304 is the second sensing fiber, 305 is the third node, 306 is the (N-1)th node, 307 is the (N-1)th sensing fiber 307, and 308 is the Nth node;

[0052] 103A is an input polarization-maintaining fiber coupler, 103B is an output polarization-maintaining fiber coupler, 103C is a matching interferometer long-arm polarization-maintaining fiber, 103D is a matching interferometer short-arm polarization-maintaining fiber, and 103F is a PZT fiber phase modulator;

[0053] 201A, input port, 201B, first output port, 201C, second output port. DETAILED DESCRIPTION

[0054] 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.

[0055] In one embodiment of the present invention, referring to Figure 1 , a fiber optic interferometer sensor network is provided, including a sensor network modulation and demodulation module 1, a transmission link unit 2 and a fiber optic interferometer sensor network unit 3.

[0056] The sensor network modulation and demodulation module 1 is used to generate an inquiry laser with specific modulation information and demodulate the return light of the sensor network to extract the sensor signal. Specifically, the sensor network modulation and demodulation module 1 includes a narrow linewidth laser 101, an acousto-optic modulator 102, a matching interferometer 103, a polarization modulator 104, an optoelectronic conversion and acquisition module 105 and a main control module 106.

[0057] The narrow linewidth laser 101 is connected to the input port of the acousto-optic modulator 102 through an optical fiber; the output port of the acousto-optic modulator 102 is connected to the input port of the matching interferometer 103 through an optical fiber; the output port of the matching interferometer 103 is connected to the input port of the polarization modulator 104 through an optical fiber; the photoelectric conversion and acquisition module 105 is electrically connected to the main control module 106; the main control module 106 is connected to the acousto-optic modulator 102, the matching interferometer 103 and the polarization modulator 104 through electric wires, and the main control module 106 is connected to the host computer 107 through an Ethernet cable. Among them, the narrow line width laser 101 is used to generate continuous laser, the acousto-optic modulator 102 is used to modulate the laser into pulse laser, the polarization modulator 104 is used to adjust the polarization state of each pulse in the pulse pair, the photoelectric conversion and acquisition module 105 is used to perform photoelectric conversion and sampling on the optical pulse signal returned by the fiber interferometer sensor network unit 3, the main control module 106 is used to generate the modulation signal of the acousto-optic modulator 102, the matching interferometer 103 and the polarization modulator 104, and the main control module 106 is used to generate the synchronization trigger signal of the photoelectric conversion and acquisition module 105. The host computer 107 controls the fiber interferometer sensor network by controlling the main control module 106, and is also used to collect and store the training set and test set data and run the neural network algorithm.

[0058] The transmission link unit 2 is used to connect the sensor network modulation and demodulation module 1 and the fiber interferometer sensor network unit 3. The transmission link unit 2 includes a fiber circulator 201 and a link transmission optical fiber 202. Figure 3 The schematic diagram of the structure of an optical fiber circulator provided in an embodiment is shown in FIG. 201. The optical fiber circulator 201 has three ports, namely, an input port 201A, a first output port 201B, and a second output port 201C. The output port of the polarization modulator 104 is connected to the input port of the optical fiber circulator 201 through an optical fiber; the first output port of the optical fiber circulator 201 is connected to the link transmission optical fiber 202, which is a single-mode optical fiber; the second output port of the optical fiber circulator 201 is connected to the photoelectric conversion and acquisition module 105.

[0059] The fiber optic interferometer sensing network unit 3 is used to sense external signals; it includes a first node 301, a first sensing fiber 302, a second node 303, a second sensing fiber 304, a third node 305, ..., an (N-1)th node 306, an (N-1)th sensing fiber 307, and an Nth node 308, wherein the first node 301, the first sensing fiber 302 and the second node 303 constitute a fiber optic interferometer, the second node 303, the second sensing fiber 304 and the third node 305 constitute a fiber optic interferometer, and so on, the (N-1)th node 306, the (N-1)th sensing fiber 307 and the Nth node 308 constitute a fiber optic interferometer. The fiber interferometer sensing network unit 3 comprises a total of (N-1) fiber interferometers, i.e., has (N-1) sensing channels, wherein the first node 301, the second node 303, the third node 305, ..., the (N-1)th node 306, and the Nth node 308 are a combination of a fiber grating or a fiber coupler and a Faraday rotator.

[0060] Reference Figure 2 In one embodiment, the matching interferometer 103 includes an input polarization-maintaining fiber coupler 103A, an output polarization-maintaining fiber coupler 103B, an interferometer long-arm polarization-maintaining fiber 103C, and an interferometer short-arm polarization-maintaining fiber 103D, wherein a portion of the interferometer long-arm polarization-maintaining fiber 103C is wound around a piezoelectric ceramic (PZT) to form a PZT fiber phase modulator 103F.

[0061] In one embodiment, the matching interferometer 103 is used to convert a pulse into a pair of pulses, and the pulse time interval is exactly the same as the time it takes for light to travel back and forth between two adjacent nodes in the fiber interferometer sensor network unit 3, and a phase-generated carrier modulation signal is introduced through the PZT fiber phase modulator 103F. ,in is the modulation amplitude, The carrier modulation signal introduced by the PZT fiber phase modulator 103F can also be a heterodyne signal.

[0062] In one embodiment, the laser light wavelength generated by the narrow linewidth laser 101 is in the 1550 nm band.

[0063] For the fiber interferometer sensor network provided in the above embodiment, the host computer includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method for eliminating the crosstalk of the sensor channel in the fiber interferometer sensor network when executing the computer program. Figure 4 , a method for eliminating crosstalk in a sensor channel in a fiber interferometer sensor network is provided, comprising the following steps:

[0064] Applying single-frequency sinusoidal signals of different frequencies to the link transmission optical fibers and each sensing optical fiber in the optical fiber interferometer sensing network to obtain the original interference signals of all sensing channels in the optical fiber interferometer sensing network;

[0065] The original interference result of each sensing channel is synthesized in the complex domain by a phase-generated carrier wave, so as to obtain the complex domain expression of the original interference result of each sensing channel, that is, to obtain the complex interference signal of each sensing channel;

[0066] Demodulate the complex interference signal of each sensing channel to obtain the time domain and frequency domain results of the demodulated signal of each sensing channel;

[0067] Compare the real frequency actually applied to each sensing channel, suppress the line spectrum of the non-real frequency in the frequency domain result of the demodulated signal corresponding to each sensing channel to the background noise level, and obtain the target frequency domain result of the demodulated signal corresponding to each sensing channel;

[0068] Perform inverse Fourier transform on the suppressed signal frequency results corresponding to each sensor channel to obtain the demodulated signal target time domain results corresponding to each sensor channel;

[0069] The complex interference signal of each sensor channel is input as training data into the pre-built neural network model to predict the demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel; the predicted demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel are autocorrelated with the target time domain result and demodulation signal target frequency domain result corresponding to each sensor channel, and the absolute value of the autocorrelation coefficient is used as the loss function. The network parameters are iteratively learned according to the gradient descent method until the model converges, the iteration is stopped, the current network parameters are saved, and a distillation model is formed;

[0070] The trained distillation model is used to reconstruct the original interference signals of all interference channels in the fiber optic interferometer sensor network and eliminate the interference of the fiber optic interferometer sensor network.

[0071] Preferably, the single-frequency sinusoidal signals of different frequencies applied to the transmission optical fiber and each sensing optical fiber in the optical fiber interferometer sensing network are within the response frequency band of the optical fiber interferometer sensing network, and the interval between any two single-frequency sinusoidal signals is not less than 10 Hz.

[0072] Reference Figure 5An embodiment provides a neural network framework structure diagram suitable for a fiber interferometer sensor network, wherein the fiber interferometer sensor network has (N-1) sensor channels, and the neural network model contains an input layer, (N-1) hidden layers, and an output layer. The number of neurons in the output layer is (N-1), and the number of neurons in each hidden layer is: [(N-1)+K], where K is the sequence number of the hidden layer, wherein the [(N-1)+K]th neuron in the Kth hidden layer is equal to the complex conjugate of the Kth neuron in the (K-1)th hidden layer. The input layer can also be marked as the 0th layer, and the output layer is marked as the Nth layer.

[0073] The complex interference signal of each sensor channel is used as the input of the input layer, and the output layer re-outputs (N-1) neurons. Each output neuron represents the reconstructed interference signal. The output interference signal is demodulated using the random phase fading suppression demodulation method to obtain the predicted time domain results and frequency domain results of the demodulated signal corresponding to each sensor channel.

[0074] Preferably, by setting a user threshold , when the absolute value of the loss function is less than When , the iteration stops, the current network parameters are saved, and the distillation model is formed.

[0075] Furthermore, the present invention also includes evaluating the interference suppression effect of the trained distillation model based on the test data, including:

[0076] Applying single-frequency sinusoidal signals of different frequencies to the transmission optical fiber and each sensing optical fiber in the optical fiber interferometer sensor network, wherein the signal frequency of each single-frequency sinusoidal signal is different from the signal frequency corresponding to the training data, to obtain the original interference signals of all sensing channels in the optical fiber interferometer sensor network;

[0077] The original interference results of each sensing channel are synthesized in the complex domain by a phase-generated carrier wave, so as to obtain the complex domain expression of the original interference results of each sensing channel, that is, to obtain the complex interference signals of each sensing channel as test data;

[0078] The test data is input into the distillation model, and the distillation model outputs the reconstructed interference signal, which is demodulated using the random phase fading suppression demodulation method. If the interference signal between the sensor channels after output is lower than -40dB, it is considered that the trained distillation model is effective in suppressing interference between sensor channels.

[0079] The present invention adopts a deep learning neural network and constructs the loss function of the neural network by coupling physical detuning with time domain signal autocorrelation. The coefficient relationship between the main interference term and the parasitic interference term in the optical fiber sensor network can be obtained by learning a large amount of data, and the processing model is further distilled out and solidified into the signal processing front end of the optical fiber interferometer sensor network. Without increasing the hardware cost and complexity of the system, the original interference signal of the large-scale optical fiber interferometer network is reconstructed and the interference between channels is eliminated.

[0080] Without loss of generality, in one embodiment, based on Figure 1 The optical fiber interferometer sensor network of the structure shown provides a method for eliminating crosstalk in the sensor channel of the optical fiber interferometer sensor network, comprising the following steps:

[0081] The frequencies applied to the first sensing optical fiber 302, the second sensing optical fiber 304, ..., and the (N-1)th sensing optical fiber 307 are respectively , , …, A single-frequency sinusoidal signal with a frequency of A single-frequency sinusoidal signal, where Response frequency band of optical fiber sensor network Within, and any two The interval between them should not be less than 10Hz.

[0082] The host computer 107 outputs the interference results of all the sensing channels in the fiber interferometer sensing network, which includes (N-1) sensing channels in total. Each sensing channel contains 4 polarization state interference results, which are recorded as , , and ,in A total of M (M>1000) groups were collected, that is, a total of 4M original interference results were contained.

[0083] The 4M original interference results are synthesized in the complex domain using a phase-generated carrier-based method to obtain a complex domain expression, which is recorded as , , and ,in . A total of 4M original interference results are expressed in the complex domain.

[0084] The complex interference signal of each sensor channel is demodulated using a phase carrier and polarization synthesis hybrid demodulation algorithm to obtain the demodulation signal time domain of each sensor channel in the fiber interferometer sensor network. And frequency domain results ,in , represents the time variable, Represents a frequency variable.

[0085] Compare the actual frequency applied to each sensor channel and convert the demodulated signal frequency domain results corresponding to each sensor channel into The amplitude value of the spectrum signal of the non-real frequency is replaced by noise to generate the target frequency domain result of the demodulated signal , the demodulated signal target frequency domain result Perform inverse Fourier transform to obtain the target time domain result of the demodulated signal .

[0086] Construct a neural network model with an input layer, (N-1) hidden layers and an output layer. The number of neurons in the output layer is (N-1), and the number of neurons in each hidden layer is [(N-1)+K] K is the serial number of the hidden layer. Among them, the [(N-1)+K]th neuron in the Kth layer is equal to the complex conjugate of the Kth neuron in the (K-1)th layer, the number of output neurons is (N-1), the input layer can also be marked as the 0th layer, and the output layer is marked as the Nth layer.

[0087] Complex interference signal of each sensor channel ( ) is used as training data to input the pre-built neural network model, and the output layer outputs (N-1) neurons, denoted as Under the current neural network parameters, the complex interference signal of each sensor channel is , and ( ) is input into the neural network, and each input can also obtain a new output, which is recorded as , and . will output , , and Polarization synthesis algorithm is used to obtain the predicted demodulation signal time domain results of each sensor channel in the fiber interferometer sensor network. And frequency domain results The predicted demodulation signal time domain results and demodulation signal frequency domain results corresponding to each sensor channel are autocorrelated with the target demodulation signal time domain results and demodulation signal frequency domain results corresponding to each sensor channel. The absolute value of the autocorrelation coefficient -1 is used as the loss function, and the network parameters are iteratively learned according to the gradient descent method. Given the user threshold , when the absolute value of the loss function is less than When , the iteration stops, the current network parameters are saved, and the distillation model is formed.

[0088] Next, the interference suppression effect of the trained distillation model is evaluated based on the test data, including:

[0089] Test data collection:

[0090] The frequencies applied to the first sensing optical fiber 302, the second sensing optical fiber 304, ..., and the (N-1)th sensing optical fiber 307 are respectively , , …, A single-frequency sinusoidal signal with a frequency of A single-frequency sinusoidal signal, where Response frequency band of optical fiber sensor network Within, and any two The interval between them is not less than 10Hz, and .

[0091] The host computer 107 outputs the interference results of all channels in the fiber interferometer sensor network, which includes (N-1) sensor channels in total. Each sensor channel contains 4 polarization state interference results, which are recorded as , , and ,in A total of M (M>1000) groups were collected, that is, a total of 4M original interference results were contained.

[0092] The 4M original interference results are synthesized in the complex domain using a phase-generated carrier-based method to obtain a complex domain expression, which is recorded as , , and ,in . A total of 4M original interference results are expressed in the complex domain.

[0093] Evaluate the interference suppression effect of the trained distillation model based on the test data:

[0094] The , , and Input distillation model, the output of distillation model is recorded as , , and , the output , , and Polarization synthesis algorithm is used to obtain the predicted demodulation signal time domain results of each sensor channel in the fiber interferometer sensor network. And frequency domain results When the following conditions are met, the trained distillation model is judged to be qualified;

[0095] ;

[0096] It can be seen that this method can simultaneously eliminate the interference caused by the sensing channel and transmission link in the fiber interferometer sensor network without adding additional optical devices. Without increasing the power consumption and volume of the system, the performance of the fiber interferometer sensor network is effectively improved.

[0097] On the other hand, an embodiment of the present invention further provides a device for eliminating crosstalk in a sensing channel of a fiber interferometer sensing network, comprising:

[0098] The first module is used to apply single-frequency sinusoidal signals of different frequencies to the link transmission optical fiber and each sensing optical fiber in the optical fiber interferometer sensing network, and simultaneously obtain the original interference signals of all sensing channels in the optical fiber interferometer sensing network;

[0099] The second module is used to use a complex domain synthesis method based on phase-generated carriers to obtain the complex domain expression of the original interference results of each sensing channel, that is, to obtain the complex interference signals of each sensing channel;

[0100] The third module is used to demodulate the complex interference signal of each sensing channel to obtain the time domain and frequency domain results of the demodulated signal of each sensing channel;

[0101] The fourth module is used to compare the real frequency actually applied to each sensing channel, suppress the line spectrum of the non-real frequency in the frequency domain result of the demodulated signal corresponding to each sensing channel to the background noise level, and obtain the target frequency domain result of the demodulated signal corresponding to each sensing channel;

[0102] The fifth module is used to perform inverse Fourier transform on the suppressed signal frequency results corresponding to each sensing channel to obtain the demodulated signal target time domain results corresponding to each sensing channel;

[0103] The sixth module is used to input the complex interference signal of each sensor channel as training data into the pre-built neural network model, and predict the demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel; autocorrelate the predicted demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel with the target time domain result and demodulation signal target frequency domain result corresponding to each sensor channel, use the absolute value of the autocorrelation coefficient as the loss function, perform iterative learning of network parameters according to the gradient descent method, until the model converges, stop iteration, save the current network parameters, and form a distillation model;

[0104] The seventh module is used to reconstruct the original interference signals of all interference channels in the fiber optic interferometer sensor network using the trained distillation model, thereby eliminating the interference of the fiber optic interferometer sensor network.

[0105] The implementation methods of the above modules and the construction of the model can adopt the methods described in any of the above embodiments, which will not be repeated here.

[0106] Matters not covered by the present invention are known technologies.

[0107] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0108] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

[0109] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for eliminating crosstalk in a sensor channel in a fiber interferometer sensor network, characterized in that: include: Applying single-frequency sinusoidal signals of different frequencies to the link transmission optical fibers and each sensing optical fiber in the optical fiber interferometer sensing network to obtain the original interference signals of all sensing channels in the optical fiber interferometer sensing network; The original interference result of each sensing channel is synthesized in the complex domain by a phase-generated carrier wave, so as to obtain the complex domain expression of the original interference result of each sensing channel, that is, to obtain the complex interference signal of each sensing channel; Demodulate the complex interference signal of each sensing channel to obtain the time domain and frequency domain results of the demodulated signal of each sensing channel; Compare the real frequency actually applied to each sensing channel, suppress the line spectrum of the non-real frequency in the frequency domain result of the demodulated signal corresponding to each sensing channel to the background noise level, and obtain the target frequency domain result of the demodulated signal corresponding to each sensing channel; Perform inverse Fourier transform on the suppressed signal frequency results corresponding to each sensor channel to obtain the demodulated signal target time domain results corresponding to each sensor channel; The complex interference signal of each sensor channel is input as training data into the pre-built neural network model to predict the demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel; the predicted demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel are autocorrelated with the target time domain result and demodulation signal target frequency domain result corresponding to each sensor channel, and the absolute value of the autocorrelation coefficient is used as the loss function. The network parameters are iteratively learned according to the gradient descent method until the model converges, the iteration is stopped, the current network parameters are saved, and a distillation model is formed; The trained distillation model is used to reconstruct the original interference signals of all interference channels in the fiber optic interferometer sensor network and eliminate the interference of the fiber optic interferometer sensor network.

2. The method for eliminating crosstalk in a fiber optic interferometer sensor network according to claim 1, characterized in that: The single-frequency sinusoidal signals of different frequencies applied to the transmission optical fiber and each sensing optical fiber in the optical fiber interferometer sensor network are all within the response frequency band of the optical fiber interferometer sensor network, and the interval between any two single-frequency sinusoidal signals is not less than 10 Hz.

3. The method for eliminating crosstalk in a sensor channel in a fiber interferometer sensor network according to claim 1 or 2, characterized in that: The fiber interferometer sensing network has (N-1) sensing channels, and the neural network model contains an input layer, (N-1) hidden layers and an output layer. The number of neurons in the output layer is (N-1), and the number of neurons in each hidden layer is: [(N-1)+K], where K is the serial number of the hidden layer, wherein the [(N-1)+K]th neuron in the Kth hidden layer is equal to the complex conjugate of the Kth neuron in the (K-1)th hidden layer; The complex interference signal of each sensor channel is used as the input of the input layer, and the output layer re-outputs (N-1) neurons. Each output neuron represents the reconstructed interference signal. The output interference signal is demodulated using the random phase fading suppression demodulation method to obtain the predicted time domain results and frequency domain results of the demodulated signal corresponding to each sensor channel.

4. The method for eliminating crosstalk in a sensor channel in a fiber interferometer sensor network according to claim 3, characterized in that: Setting User Thresholds , when the absolute value of the loss function is less than When , the iteration stops, the current network parameters are saved, and the distillation model is formed.

5. The method for eliminating crosstalk in a sensor channel in a fiber interferometer sensor network according to claim 3, characterized in that: It also includes evaluating the interference suppression effect of the trained distillation model based on test data, including: Applying single-frequency sinusoidal signals of different frequencies to the transmission optical fiber and each sensing optical fiber in the optical fiber interferometer sensor network, wherein the signal frequency of each single-frequency sinusoidal signal is different from the signal frequency corresponding to the training data, to obtain the original interference signals of all sensing channels in the optical fiber interferometer sensor network; The original interference results of each sensing channel are synthesized in the complex domain by a phase-generated carrier wave, so as to obtain the complex domain expression of the original interference results of each sensing channel, that is, to obtain the complex interference signals of each sensing channel as test data; The test data is input into the distillation model, and the distillation model outputs the reconstructed interference signal, which is demodulated using the random phase fading suppression demodulation method. If the interference signal between the sensor channels after output is lower than -40dB, it is considered that the trained distillation model is effective in suppressing interference between sensor channels.

6. A device for eliminating crosstalk in a sensor channel in a fiber optic interferometer sensor network, characterized in that: include: The first module is used to apply single-frequency sinusoidal signals of different frequencies to the link transmission optical fiber and each sensing optical fiber in the optical fiber interferometer sensing network, and simultaneously obtain the original interference signals of all sensing channels in the optical fiber interferometer sensing network; The second module is used to use a complex domain synthesis method based on phase-generated carriers to obtain the complex domain expression of the original interference results of each sensing channel, that is, to obtain the complex interference signals of each sensing channel; The third module is used to demodulate the complex interference signal of each sensing channel to obtain the time domain and frequency domain results of the demodulated signal of each sensing channel; The fourth module is used to compare the real frequency actually applied to each sensing channel, suppress the line spectrum of the non-real frequency in the frequency domain result of the demodulated signal corresponding to each sensing channel to the background noise level, and obtain the target frequency domain result of the demodulated signal corresponding to each sensing channel; The fifth module is used to perform inverse Fourier transform on the suppressed signal frequency results corresponding to each sensing channel to obtain the demodulated signal target time domain results corresponding to each sensing channel; The sixth module is used to input the complex interference signal of each sensor channel as training data into the pre-built neural network model, and predict the demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel; autocorrelate the predicted demodulation signal time domain result and demodulation signal frequency domain result corresponding to each sensor channel with the target time domain result and demodulation signal target frequency domain result corresponding to each sensor channel, use the absolute value of the autocorrelation coefficient as the loss function, perform iterative learning of network parameters according to the gradient descent method, until the model converges, stop iteration, save the current network parameters, and form a distillation model; The seventh module is used to reconstruct the original interference signals of all interference channels in the fiber optic interferometer sensor network using the trained distillation model, thereby eliminating the interference of the fiber optic interferometer sensor network.

7. The device for eliminating crosstalk of sensing channels in a fiber optic interferometer sensing network according to claim 6, characterized in that: It also includes an eighth module, which is used to evaluate the interference suppression effect of the trained distillation model based on test data, including: obtaining test data, by applying single-frequency sinusoidal signals of different frequencies to the transmission optical fiber and each sensor optical fiber in the optical fiber interferometer sensor network, and the signal frequency of each single-frequency sinusoidal signal is different from the signal frequency corresponding to the training data, to obtain the original interference signals of all sensor channels in the optical fiber interferometer sensor network; using a complex domain synthesis method based on phase-generated carrier for the original interference results of each sensor channel, to obtain the complex domain expression of the original interference results of each sensor channel, that is, to obtain the complex interference signals of each sensor channel as test data; The test data is input into the distillation model, and the distillation model outputs the reconstructed interference signal, which is demodulated using the random phase fading suppression demodulation method. If the interference signal between the sensor channels after output is lower than -40dB, it is considered that the trained distillation model is effective in suppressing interference between sensor channels.

8. The device for eliminating crosstalk of sensing channels in a fiber optic interferometer sensing network according to claim 6, characterized in that: The fiber optic interferometer sensor network includes a sensor network modulation and demodulation module, a transmission link unit and a fiber optic interferometer sensor network unit; The sensor network modulation and demodulation module includes a narrow line width laser, an acousto-optic modulator, a matching interferometer, a polarization modulator, a photoelectric conversion and acquisition module and a main control module; The narrow linewidth laser is connected to the input port of the acousto-optic modulator through an optical fiber; the output port of the acousto-optic modulator is connected to the input port of the matching interferometer through an optical fiber; the output port of the matching interferometer is connected to the input port of the polarization modulator through an optical fiber; the photoelectric conversion and acquisition module is electrically connected to the main control module; the main control module is connected to the acousto-optic modulator, the matching interferometer and the polarization modulator through electric wires, and is connected to the host computer through an Ethernet cable, wherein the narrow linewidth laser is used to generate continuous laser, the acousto-optic modulator is used to modulate the laser into a pulse laser, the polarization modulator is used to adjust the polarization state of each pulse in the pulse pair, the photoelectric conversion and acquisition module is used to photoelectrically convert and sample the optical pulse signal returned by the fiber interferometer sensor network unit, the main control module is used to generate modulation signals for the acousto-optic modulator, the matching interferometer and the polarization modulator, and the main control module is used to generate a synchronous trigger signal for the photoelectric conversion and acquisition module; A transmission link unit, used to connect the sensor network modulation and demodulation module and the fiber optic interferometer sensor network unit; The fiber interferometer sensing network unit is used to sense external signals, and includes a first node, a first sensing fiber, a second node, a second sensing fiber, a third node, ..., the (N-1)th node, the (N-1)th sensing fiber, and the Nth node, wherein the first node, the first sensing fiber, and the second node constitute a fiber interferometer, the second node, the second sensing fiber, and the third node constitute a fiber interferometer, and so on, the (N-1)th node, the (N-1)th sensing fiber, and the Nth node constitute a fiber interferometer. The fiber interferometer sensing network unit includes a total of (N-1) fiber interferometers, that is, it has (N-1) sensing channels, wherein the first node, the second node, the third node, ..., the (N-1)th node, and the Nth node are fiber gratings or fiber couplers and Faraday rotators.

9. The device for eliminating crosstalk of sensing channels in a fiber optic interferometer sensing network according to claim 8, characterized in that: The transmission link unit includes an optical fiber circulator and a link transmission optical fiber, wherein the optical fiber circulator has three ports, namely an input port, a first output port, and a second output port; The output port of the polarization modulator is connected to the input port of the optical fiber circulator through an optical fiber; The first output port of the fiber optic circulator is connected to the link transmission optical fiber, which is a single-mode optical fiber; the second output port of the fiber optic circulator is connected to the photoelectric conversion and acquisition module.

10. The device for eliminating crosstalk of sensing channels in a fiber interferometer sensing network according to claim 8, characterized in that: The matching interferometer comprises an input polarization-maintaining fiber coupler, an output polarization-maintaining fiber coupler, an interferometer long-arm polarization-maintaining fiber and an interferometer short-arm polarization-maintaining fiber, wherein the interferometer long-arm polarization-maintaining fiber is partially wound around a piezoelectric ceramic to form a PZT fiber phase modulator; The matching interferometer is used to convert a pulse into a pair of pulses, and the pulse time interval is exactly the same as the time it takes for light to travel back and forth between two adjacent nodes in the fiber interferometer sensor network unit, and a phase-generated carrier modulation signal is introduced through a PZT fiber phase modulator. ,in is the modulation amplitude, is the modulation frequency.

Images