Distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning

Abstract

The invention discloses a distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning, and the device comprises a chaotic fiber laser, a linear Sagnac interferometer, and a computer. The chaotic fiber laser comprises a semiconductor pumping source, a wavelength division multiplexer, an erbium-doped fiber, a polarization controller, a fourth fiber coupler and an optical isolator which are connected in sequence, the optical isolator is connected with the wavelength division multiplexer, and the fiber coupler is connected with a linear Sagnac interferometer. According to the distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning, a chaotic fiber sensing system and a Sagnac interference light path are combined, and a convolutional neural network is introduced to extract weak vibration signals, so that an effective signal frequency range can be extracted under a relatively low signal-to-noise ratio; the sensitivity is higher and the detection range is wider.

Description

Technical Field

[0001] The present invention relates to the field of distributed chaotic fiber laser sensing technology, and in particular to a distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning. Background Art

[0002] Compared with traditional sensors, distributed fiber optic sensors have the characteristics of low cost, small size, resistance to electromagnetic interference, passivity, corrosion resistance, and adaptability to harsh environments. They are widely used in underground exploration, pipeline monitoring, border warning, etc., and have important applications in military, national defense, aerospace and other fields.

[0003] Common distributed fiber optic sensing systems mainly include three systems based on Rayleigh scattering, Brillouin scattering and Raman scattering. However, since they generally use narrow-linewidth pulse light sources with strong coherence, optical signals are prone to interference, which limits their measurement range. To meet the needs of practical applications, these systems need to achieve higher sensitivity and higher resolution. Compared with traditional sensing systems, chaotic laser light source sensing systems use chaotic lasers as light sources. Due to the nonlinear characteristics of chaotic signals themselves, they are more sensitive to weak signals, and the small changes loaded on them can be amplified so that weak signals can be further extracted later. Chaotic laser is a frequency-controllable, low-coherence light source with zero-order crosstalk characteristics. Combining chaotic laser with Sagnac interferometer can further improve the sensitivity of the sensor. At the same time, the central wavelength of the chaotic laser can be controlled within 5nm, so that this distributed chaotic laser fiber optic sensing system has a wide measurement range.

[0004] Machine learning has strong nonlinear modeling capabilities. With the extensive research of machine learning, a large number of machine learning models have been applied to various fields, and signal detection methods based on machine learning have gradually emerged. Traditional weak vibration laser sensor devices use interference, diffraction, self-mixing and other methods to detect sensor signals. These operations are relatively complex and have poor anti-interference capabilities. In contrast, the application of machine learning to sensor signal detection has the advantages of simple operation, strong anti-interference ability, and no need for prior knowledge, and has high application value. Summary of the invention

[0005] The purpose of the present invention is to provide a distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning, which combines the chaotic fiber sensing system with the Sagnac interference optical path to solve the problems of small measurement range, low sensitivity and poor anti-interference ability of the traditional sensing system. At the same time, a method for extracting weak vibration signals using a deep convolutional neural network is introduced to achieve the detection of weak vibration signals at higher sensitivity. Compared with the traditional laser sensing system, the detection of vibration signals has higher accuracy and anti-interference.

[0006] To achieve the above-mentioned purpose, the present invention provides a distributed chaotic fiber laser weak vibration signal sensing device based on machine learning, comprising a chaotic fiber laser, a linear Sagnac interferometer and a computer, wherein the chaotic fiber laser comprises a semiconductor pump source, a wavelength division multiplexer, an erbium-doped fiber, a polarization controller, a fourth fiber coupler and an optical isolator connected in sequence, the output end of the optical isolator is connected to the wavelength division multiplexer, and the fiber coupler is connected to the linear Sagnac interferometer.

[0007] Preferably, the linear Sagnac interferometer includes a coupling reflection component and a photoelectric conversion amplification analysis component, and the coupling reflection component includes a first fiber coupler, a delay fiber, a second fiber coupler, a sensing fiber, an armored fiber, and a third fiber coupler connected in sequence to a fourth fiber coupler;

[0008] The photoelectric conversion, amplification and analysis component comprises a photoelectric detector, a signal amplifier and a high-speed A / D acquisition card which are sequentially connected to the first optical fiber coupler, and the high-speed A / D acquisition card is connected to a computer.

[0009] Preferably, the chaotic fiber laser is a C-band chaotic fiber laser with controllable light intensity, Lyapunov index and center frequency, which utilizes a semiconductor pump source to regulate the nonlinear Kerr effect in the cavity to generate chaotic laser.

[0010] Preferably, the delay optical fiber, the sensing optical fiber and the armored optical fiber are all single-mode optical fibers, the first optical fiber coupler is a 3×3 optical fiber coupler, and the second optical fiber coupler and the third optical fiber coupler are both 1×2 optical fiber couplers.

[0011] A method for using a distributed chaotic fiber laser weak vibration signal sensing device based on machine learning, comprising the following steps:

[0012] Step S1, initializing the entire sensing device;

[0013] Step S2, the chaotic fiber laser is used as a chaotic light source to emit a chaotic signal, which is propagated through the interference optical path of a linear Sagnac interferometer, and a weak vibration signal is loaded at the sensing optical fiber;

[0014] Step S3, the chaotic optical signal loaded with the weak vibration signal is sequentially passed through a photoelectric detector, a signal amplifier, and a high-speed A / D acquisition card and then imported into a computer;

[0015] Step S4, processing the data imported into the computer through a convolutional neural network to determine the frequency range of the loaded weak vibration signal;

[0016] Step S5: Repeat step S4 to calculate the average value.

[0017] Therefore, the present invention adopts the above-mentioned distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning, and the beneficial effects are as follows:

[0018] (1) Compared with the traditional method of generating chaotic laser, the present invention uses the nonlinear Kerr effect to generate chaotic laser. The main mechanism is to use the self-phase modulation and mutual phase modulation of the light field, breaking through the lifetime limit of the upper energy level particles, so that the bandwidth of the generated chaotic optical signal far exceeds the chaotic laser generated by the traditional method, and the bandwidth can reach gigahertz.

[0019] (2) Compared with traditional fiber optic sensing technology, the present invention utilizes low-coherence chaotic optical signals to ensure that the sensing system can achieve a wider detection range under medium and high sensitivity requirements, thereby optimizing the shortcomings of narrow linewidth lasers with high sensitivity but low detection range.

[0020] (3) Compared with traditional vibration signal sensing detection, the present invention has stronger anti-disturbance and anti-noise capabilities, and the processing results are more intelligent and accurate. Specifically, when the signal processing device uses a deep convolutional neural network to model and process weak vibration signals, the signal can be detected at a low signal-to-noise ratio, and the signal-to-noise ratio can be as low as -22dB.

[0021] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a diagram of a distributed chaotic fiber laser weak vibration signal sensing device according to an embodiment of a distributed chaotic fiber laser weak vibration signal sensing device based on machine learning of the present invention;

[0023] Figure 2 It is a weak vibration signal loading device of a sensing optical fiber of an embodiment of a distributed chaotic optical fiber laser weak vibration signal sensing device based on machine learning of the present invention;

[0024] Figure 3 This is an output characteristic diagram of a chaotic fiber laser pump power of 41.5 mW in an embodiment of a method for using a distributed chaotic fiber laser weak vibration signal sensing device based on machine learning of the present invention, wherein (a) is a timing diagram and (b) is an autocorrelation diagram;

[0025] Figure 4 It is a flow chart of extracting weak vibration signals by using a deep convolutional neural network in a distributed chaotic fiber laser weak vibration signal sensing device and method embodiment based on machine learning of the present invention;

[0026] Figure 5It is a graph of experimental results of extracting weak vibration signals using a deep convolutional neural network in an embodiment of a distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning of the present invention, wherein (a) is a frequency response graph of c(t), (b) is a frequency response graph of y(t), (c) is an error frequency response graph of a Model trained using c(t), and (d) is a frequency response graph of y(t) after passing through the Model;

[0027] Figure 6 It is a comparison chart of the frequency response curves of a distributed chaotic fiber laser weak vibration signal sensing device and method embodiment based on machine learning of the present invention and a spontaneous emission laser (ASE) sensing device.

[0028] Reference numerals

[0029] 1. Semiconductor pump source; 2. Wavelength division multiplexer; 3. Erbium-doped optical fiber; 4. Polarization controller; 5. Fourth optical fiber coupler; 6. Optical isolator; 7. First optical fiber coupler; 8. Delay optical fiber; 9. Second optical fiber coupler; 10. Sensing optical fiber; 11. Third optical fiber coupler; 12. Photodetector; 13. Signal amplifier; 14. High-speed A / D acquisition card; 15. Computer. DETAILED DESCRIPTION

[0030] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0031] Unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0032] like Figure 1As shown, a distributed chaotic fiber laser weak vibration signal sensing device based on machine learning includes a chaotic fiber laser, a linear Sagnac interferometer and a computer 15. The chaotic fiber laser includes a 980nm semiconductor pump source 1, a wavelength division multiplexer 2 (WDM), an erbium-doped fiber 3 (EDF), a polarization controller 4 (PC), a fourth fiber coupler 5 (OC) and an optical isolator 6 (ISO) connected in sequence. The optical isolator 6 (ISO) is connected to the wavelength division multiplexer 2 (WDM), and each device in the chaotic fiber laser is connected with each other using a single-mode optical fiber.

[0033] The chaotic fiber laser in the present invention is a C-band (wavelength 1528-1567nm) chaotic fiber laser with controllable light intensity, Lyapunov index and center frequency. Specifically, a semiconductor pump source 1 and an erbium-doped fiber 3 (EDF) and other devices constitute an erbium-doped fiber ring laser (EDFRL). The semiconductor pump source 1 is used to adjust the nonlinear Gram effect in the cavity to generate chaotic laser, thereby obtaining a chaotic optical signal.

[0034] The fourth fiber coupler 5 (OC) is connected to a linear Sagnac interferometer, which includes a coupling reflection component and a photoelectric conversion amplification analysis component. The coupling reflection component includes a first fiber coupler 7 (OC1), a delay fiber 8, a second fiber coupler 9 (OC2), a sensing fiber 10, an armored fiber and a third fiber coupler 11 (OC3) which are sequentially connected to the fourth fiber coupler 5 (OC); the photoelectric conversion amplification analysis component includes a photodetector 12 (PD), a signal amplifier 13 and a high-speed A / D acquisition card 14 (ADC) which are sequentially connected to the first fiber coupler 7 (OC1), and the high-speed A / D acquisition card 14 (ADC) is connected to a computer 15.

[0035] Among them, the delay fiber 8, the sensing fiber 10 and the armored fiber are all single-mode fibers, the first fiber coupler 7 (OC1) is a 3×3 fiber coupler, the second fiber coupler 9 (OC2) and the third fiber coupler 11 (OC3) are all 1×2 fiber couplers, and the coupling ratios of the fourth fiber coupler 5 (OC), the first fiber coupler 7 (OC1), the second fiber coupler 9 (OC2) and the third fiber coupler 11 (OC3) are 1:9, 1:1, 1:1 and double-ended 1:1:1, respectively.

[0036] like Figure 2As shown, part of the sensing optical fiber 10 is evenly wound around the periphery of the titanium alloy steel tube to form an armored optical fiber, which is used to receive the tiny strain caused by the weak vibration signal on the sensing optical fiber 10. After the chaotic optical signal propagates through the structure, it is converted into an electrical signal by the photodetector 12 (PD), and then collected by a high-speed A / D acquisition card and transmitted to the computer 15. The computer 15 processes the data through a deep convolutional neural network to calculate and determine the frequency range of the weak vibration signal.

[0037] In this embodiment, the winding length of the armored optical fiber is 13m, and the dimensions of the titanium alloy steel pipe are 110mm in outer diameter, 3mm in thickness, and 120mm in height. The titanium alloy steel pipe can effectively receive weak vibrations and load vibration signals into the armored optical fiber.

[0038] Working principle: The back end of the photodetector 12 (PD) can be connected to a data acquisition device, i.e., a high-speed A / D acquisition card 14 (ADC), through a high-frequency coaxial cable. The data acquisition device is then connected to a data processing device through a data line. Finally, the data processing device is connected to a computer 15 through a data line to display data.

[0039] The chaotic light source is input into the first fiber coupler 7 (OC1), and the light source is divided into two paths by the first fiber coupler 7 (OC1). One path first passes through the delay fiber 8 and then enters the second fiber coupler 9 (OC2), and the other path is directly input into the second fiber coupler 9 (OC2). The two paths of light are output after merging, and enter the third fiber coupler 11 (OC3) after passing through the sensing fiber 10. The third fiber coupler 11 (OC3) acts as a reflector. After the reflected input light signal returns to the original path, it interferes and outputs at the first fiber coupler 7 (OC1), and then enters the photodetector 12 (PD) to be converted into an electrical signal. The electrical signal will be collected by the signal amplifier 13 and the high-speed A / D acquisition card 14 (ADC), and finally input into the computer 15 through the data line for analysis and processing.

[0040] A method for using a distributed chaotic fiber laser weak vibration signal sensing device based on machine learning, comprising the following steps:

[0041] Step S1, initializing the entire sensing device, including adjusting the polarization controller 4 (PC), initializing the A / D high-speed acquisition card 14 (ADC), and connecting all components.

[0042] Step S2, the chaotic fiber laser is used as a chaotic light source to emit a chaotic signal, which is propagated through the interference optical path of a linear Sagnac interferometer, and a weak vibration signal is loaded at the sensing optical fiber 10;

[0043] Step S3, the chaotic optical signal loaded with the weak vibration signal is sequentially passed through the photoelectric detector 12 (PD), the signal amplifier 13, and the high-speed A / D acquisition card 14 (ADC) and then imported into the computer 15;

[0044] Step S4, processing the data imported into the computer 15 through a convolutional neural network to determine the frequency range of the loaded weak vibration signal;

[0045] Specifically, the deep convolutional neural network has the feature extraction characteristics, and the chaotic system is dynamically reconstructed. After the chaotic light signal is loaded with a weak vibration signal, the obtained mixed signal has a frequency deviation from the original chaotic light signal. Using the deep convolutional neural network dynamic model reconstruction technology, the frequency range of the weak vibration signal can be determined without prior knowledge, achieving the purpose of remote state monitoring and signal extraction.

[0046] Step S5: Repeat step S4 to calculate the average value.

[0047] The specific implementation process is as follows Figure 3 As shown, the nonlinear effect in the cavity is regulated by adjusting the pump power of the semiconductor pump source 1, and the polarization controller 4 (PC) is adjusted to control the mode-locked state. The nonlinear Kerr effect is used to make the laser signal change to the periodic state, period-doubling state, and chaotic state in sequence. The dynamic equation of the pump modulation system is:

[0048]

[0049]

[0050] Among them, I P is the pump source intensity, I L is the normalized laser intensity, D is the normalized particle flip number, κ is the loss coefficient, and g is the gain coefficient. By adjusting the polarization controller 4 (PC), the light intensity in the annular cavity can be changed, and the pump power of the semiconductor pump source 1 can be changed to achieve the change of I P The purpose of changing the two parameters at the same time is equivalent to changing the coefficient of the normalized particle flip number. When a specific value is reached, the laser I L The output optical signal will transition from a periodic state to a period-doubling state, and finally from a period-doubling state to a chaotic state.

[0051] The chaotic light source is input to port 1 of the first optical fiber coupler 7 (OC1), and is divided into two optical paths through ports 4 and 6 and finally returns to the first optical fiber coupler 7 (OC1) to cause interference. The optical path is:

[0052] L1: 7-8-9-10-11-10-9-7;

[0053] L2: 7-9-10-11-10-9-8-7;

[0054] The interference light intensity received by the photodetector 12 (PD) is:

[0055]

[0056] Where I0 is the light intensity of the light source entering the coupler, is the phase difference generated by the weak vibration signal, and Δψ is the phase shift generated by the coupler. Since the interference light intensity I is a cosine function, when its phase shift is zero, the nearby signal detection sensitivity is the lowest, and the positive and negative phase shift cannot be identified, so phase bias is required. Affected by the random time delay of the chaotic laser, the phase shift of the static operating point of the system is no longer zero, and the bias of the static operating point is realized, thereby improving the sensitivity of the system.

[0057] like Figure 4 As shown, first obtain the chaotic light source time series c(t), and use c(t) to train the convolutional neural network model Model. Add a weak vibration signal s(t) at the sensing fiber 10, and take a 1kHz cosine vibration signal as an example. Using the model dynamic reconstruction principle, set the signal-to-noise ratio to -22dB, substitute the received mixed signal y(t) = c(t) + n(t) + s(t) into the Model for prediction, and finally export the prediction error result Error, and observe its frequency response diagram to determine the signal frequency range.

[0058] Depend on Figure 5 (a) and (b) show that the added vibration signal cannot be distinguished by the naked eye, which meets the condition of weak signal. Figure 5 (c) and (d) clearly show that the characteristic frequency error at 1 kHz is obvious, so the frequency range of the weak vibration signal detected is roughly around 1 kHz. The vibration signal can be demodulated using a filter later.

[0059] Figure 6 This is a frequency response comparison diagram of this embodiment and ASE (Amplified Spontaneous Emission) light source. Taking the audio vibration signal as an example, the sound volume is adjusted, and the sound pressure is measured by a sound pressure meter to be 98dBA. The optical power of the chaotic light source and the ASE light source are set to 15mW, and then the output frequency of the audio signal is changed to obtain the peak-to-peak value of the weak vibration signal received at different frequencies.

[0060] Depend on Figure 6It can be seen that under the same laser power, when the audio frequency vibration frequency is within 100Hz, the frequency response of the laser light source and the ASE light source in this embodiment are both around 30mV; when it exceeds 100Hz and is less than 1100Hz, the frequency response of the laser light source in this embodiment gradually increases to 340mV, and the ASE light source gradually increases to 50mV; when it is 1100Hz-3500Hz, the frequency response of the chaotic light source is stable at around 340mV, and the ASE light source is stable at around 50mV. It is not difficult to see from the figure that the chaotic fiber distributed sensing system proposed in the present invention is 7 times higher than the ASE light source in frequency response, which indirectly shows that the sensitivity of the chaotic fiber distributed sensing system proposed in this example is much higher than that of the ordinary ASE light source.

[0061] Therefore, the present invention adopts the above-mentioned distributed chaotic fiber laser weak vibration signal sensing device and method based on machine learning, utilizes the nonlinear Kerr effect of single-mode erbium-doped fiber to generate chaotic optical signals, extracts weak vibration signals through Sagnac interference optical path, and utilizes deep convolutional neural network to realize distributed sensing measurement of weak vibration signals, thereby improving the sensitivity of signal detection and signal-to-noise ratio.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A distributed chaotic fiber laser weak vibration signal sensing device based on machine learning, characterized in that: The invention comprises a chaotic fiber laser, a linear Sagnac interferometer and a computer. The chaotic fiber laser comprises a semiconductor pump source, a wavelength division multiplexer, an erbium-doped fiber, a polarization controller, a fourth fiber coupler and an optical isolator which are connected in sequence. The optical isolator is connected to the wavelength division multiplexer, and the fiber coupler is connected to the linear Sagnac interferometer.

2. According to a machine learning-based distributed chaotic fiber laser weak vibration signal sensing device according to claim 1, it is characterized in that: The linear Sagnac interferometer includes a coupling reflection component and a photoelectric conversion amplification analysis component, wherein the coupling reflection component includes a first fiber coupler, a delay fiber, a second fiber coupler, a sensing fiber, an armored fiber, and a third fiber coupler which are sequentially connected to a fourth fiber coupler; The photoelectric conversion amplification analysis component comprises a photoelectric detector, a signal amplifier and a high-speed A / D acquisition card which are sequentially connected to the first optical fiber coupler, and the high-speed A / D acquisition card is connected to a computer.

3. According to a machine learning-based distributed chaotic fiber laser weak vibration signal sensing device according to claim 2, it is characterized in that: The chaotic fiber laser is a C-band chaotic fiber laser with controllable light intensity, Lyapunov index and center frequency. It uses a semiconductor pump source to regulate the nonlinear Kerr effect in the cavity to generate chaotic laser.

4. According to the machine learning-based distributed chaotic fiber laser weak vibration signal sensing device of claim 3, it is characterized in that: The delay optical fiber, the sensing optical fiber and the armored optical fiber are all single-mode optical fibers, the first optical fiber coupler is a 3×3 optical fiber coupler, and the second optical fiber coupler and the third optical fiber coupler are both 1×2 optical fiber couplers.

5. A method for using the distributed chaotic fiber laser weak vibration signal sensing device based on machine learning as described in any one of claims 1 to 4, characterized in that: The following steps are involved: Step S1, initializing the entire sensing device; Step S2, the chaotic fiber laser is used as a chaotic light source to emit a chaotic signal, which is propagated through the interference optical path of a linear Sagnac interferometer, and a weak vibration signal is loaded at the sensing optical fiber; Step S3, the chaotic optical signal loaded with the weak vibration signal is sequentially passed through a photoelectric detector, a signal amplifier, and a high-speed A / D acquisition card and then imported into a computer; Step S4, processing the data imported into the computer through a convolutional neural network to determine the frequency range of the loaded weak vibration signal; Step S5: Repeat step S4 to calculate the average value.

Images