A vehicle vibration monitoring device and method based on dual-mode optical fiber

Abstract

The present invention discloses a vehicle vibration monitoring device and method based on a two-mode optical fiber. The continuous laser emitted by the laser is divided into two by an optical coupler. One path is modulated into a detection pulse by a pulse generator and sent into the two-mode sensing optical fiber, and the other path is used as the local oscillation light. After the backward Rayleigh scattered light returned by the two-mode sensing optical fiber is combined with the local oscillation light, the digital signal processing unit performs data processing, that is, the coefficient of variation method is first used to process the two-mode raw data to obtain a vibration peak with a higher signal-to-noise ratio, and then the time domain data of the vibration peak position is extracted. After the EMD feature is extracted, it is input into the trained neural network for recognition. Compared with the prior art, the present invention provides the speed and accuracy of vehicle vibration monitoring recognition.

Description

Technical Field

[0001] The present invention relates to the field of optical fiber sensing technology, and in particular to a vehicle vibration monitoring device and method based on dual-mode optical fiber. Background Art

[0002] With the advancement of my country's urbanization process and the rapid growth of the number of motor vehicles, the traffic volume on highways continues to increase, and the difficulty of monitoring high-speed vehicles is becoming increasingly greater. At present, relatively mature road vehicle monitoring methods mainly include discrete monitoring solutions such as ring coils, digital video, and ultrasound. Although the above technologies have high single-point monitoring accuracy, discrete monitoring solutions have problems such as blind spots and complex networking.

[0003] Fiber optic sensing technology is a new type of sensing technology that has emerged in recent decades along with fiber optic technology and optical communication technology. It uses light waves as sensing signals and optical fibers as signal transmission media to detect external signals. It is different from traditional electrical sensors in terms of sensing methods, signal detection, and signal processing. Among them, fiber optic sensing based on phase-sensitive optical time domain reflectometer is widely used in many fields such as border security and oil pipeline safety monitoring due to its long-distance, fully distributed, and high-sensitivity sensing characteristics. Using the sensing characteristics of optical fibers to monitor vehicle vibrations on the road in real time and throughout the entire process has important application value for building a road traffic monitoring network with multi-sensor fusion.

[0004] The traditional phase-sensitive optical time-domain reflectometer uses single-mode optical fiber as the sensing medium and realizes vibration position positioning by amplitude differential method. Specifically, the system first collects the amplitude data of the backward Rayleigh scattering signal; when the amount of data reaches the number set by the user, the collected data is divided into rows according to the length of the optical fiber, and each row of data is the Rayleigh scattering signal generated when the light pulse completes one transmission in the sensing optical fiber; each row of data is differentiated from the data of every k rows and the absolute value is calculated, and finally accumulated to obtain a frame of data, and the vibration peak position in the sensing link is determined accordingly.

[0005] Although the traditional single-mode fiber phase-sensitive optical time-domain reflectometer has shown good sensing performance in practical applications, it still has the following shortcomings: (1) The nonlinear threshold level and reflected light capture rate of single-mode fiber are low, which limits the injectable pulse light power, resulting in a weak Rayleigh scattered light signal received at the fiber injection end, which is not conducive to the collection and subsequent processing of scattered signals; (2) Only one mode of light signal can be transmitted in a single-mode fiber, which is greatly affected by the inhomogeneity of the fiber medium and the mismatch of the polarization state. It cannot compensate for the sudden change of vibration peaks caused by noise, resulting in a decrease in the signal-to-noise ratio; (3) The differential accumulation algorithm needs to select the number of differential intervals (i.e., k value) according to the vibration frequency. Whenever the vibration frequency changes, the differential k value needs to be changed, but there is no clear selection method for the k value; and the algorithm has high requirements on the waveform, and is less effective when detecting irregular vibration sources such as vehicle vibration. Summary of the invention

[0006] The present invention aims at the problems existing in the existing single-mode optical fiber phase-sensitive optical time domain reflectometer and provides a vehicle vibration monitoring device and method based on dual-mode optical fiber.

[0007] To solve the above problems, the present invention is achieved through the following technical solutions:

[0008] A vehicle vibration monitoring device based on a dual-mode optical fiber comprises a laser, a pulse light generator, a circulator, a mode division multiplexer, a dual-mode sensing optical fiber, an acquisition card, a digital signal processing unit, two polarization controllers, two photoelectric conversion and electrical signal processing modules, and four optical couplers; the output end of the laser is connected to the input end of the first optical coupler, one output end of the first optical coupler is connected to the input end of the pulse light generator, and the other output end of the first optical coupler is connected to the input end of the second optical coupler; the output end of the pulse light generator is connected to the first port of the optical circulator, and the second port of the optical circulator is connected to the LP of the mode division multiplexer. 01 The multiplexing port of the mode division multiplexer is connected to the two-mode sensing optical fiber; the two-mode sensing optical fiber is laid along the road surface; the third port of the optical circulator is connected to an input end of the third optical coupler, and the LP of the mode division multiplexer is connected to the input end of the third optical coupler. 11The mode port is connected to an input end of a fourth optical coupler; an output end of the second optical coupler is connected to another input end of the third optical coupler via a first polarization controller; another output end of the second optical coupler is connected to another input end of the fourth optical coupler via a second polarization controller; two output ends of the third optical coupler are simultaneously connected to two input ends of the first photoelectric conversion and electrical signal processing module, and two output ends of the first photoelectric conversion and electrical signal processing module are simultaneously connected to an acquisition card; two output ends of the fourth optical coupler are simultaneously connected to two input ends of the second photoelectric conversion and electrical signal processing module, and two output ends of the second photoelectric conversion and electrical signal processing module are simultaneously connected to an acquisition card; the output end of the acquisition card is connected to the input end of the digital signal processing unit.

[0009] In the above solution, the splitting ratio of the first optical coupler is 80:20, wherein the 80% output end is connected to the input end of the pulse light generator, and the 20% output end is connected to the input end of the second optical coupler.

[0010] In the above solution, the splitting ratio of the second optical coupler, the third optical coupler and the fourth optical coupler is 50:50.

[0011] In the above scheme, the pulse light generator is a fiber-coupled acousto-optic modulator, and the pulse period of the pulse light generated by the pulse light is greater than the time it takes for the light to propagate back and forth in the two-mode sensing optical fiber.

[0012] In the above scheme, the two output ends of the first photoelectric conversion and electrical signal processing module output two IQ RF signals, both of which carry LP 01 The two output ends of the second photoelectric conversion and electrical signal processing module output two IQ RF signals, both of which carry LP 11 Mode backscattered Rayleigh light information.

[0013] A vehicle vibration monitoring method based on dual-mode optical fiber comprises the following steps:

[0014] Step 1: LP 01 The continuous laser of the mode is divided into two paths: one path is modulated to generate LP 01 Mode light pulses are injected into the two-mode sensing optical fiber; the other path generates local oscillator light after polarization control;

[0015] Step 2: Let the sample vehicle pass through a sampling point of the two-mode sensing optical fiber laid on the road surface, that is, the sample sampling point. The ground generates vibrations that disturb the two-mode sensing optical fiber, thereby generating LP carrying vibration information. 01 Module and LP 11 The Rayleigh scattered light in the backward direction;

[0016] Step 3: LP carrying vibration information 01 and LP11 The Rayleigh scattered light of the backward mode is transmitted back to the injection end of the two-mode sensing fiber and combined with the local oscillator light to obtain LP 01 Module and LP 11 Mode vibration combined optical signal;

[0017] Step 4: LP obtained in step 3 01 Module and LP 11 After the optical signal processing, photoelectric conversion and electrical signal processing, the two-mode sensing fiber LP at the sample sampling point is obtained by sampling. 01 Module and LP 11 The vibration time domain signal of the module;

[0018] Step 5: LP at the sample sampling points obtained in step 4 01 Module and LP 11 The variance and range of the vibration time domain signal of the model are calculated to obtain the LP at the sample sampling point. 01 Module and LP 11 Variance and range of the analog time domain signal;

[0019] Step 6: Use the empirical mode decomposition method to calculate the LP at the sample sampling points obtained in step 4. 01 Module and LP 11 The vibration time domain signal of the model is decomposed to obtain LP 01 Module and LP 11 The eigenmode components of the mode vibration signal; and LP 01 Module and LP 11 After the intrinsic mode component of the mode vibration signal is integrated, the LP at the sample sampling point is obtained. 01 Module and LP 11 Energy value of the eigenmode component of the modal vibration signal;

[0020] Step 7: Place the sample sampling point obtained in step 5 at LP 01 Module and LP 11 The variance and range of the time domain signal and the LP at the sample sampling point obtained in step 6 01 Module and LP 11 The energy value of the intrinsic mode component of the mode vibration signal is used as sample data and is correspondingly input into the neural network for training to obtain a trained neural network;

[0021] Step 8: Real-time acquisition of LP returned from the dual-mode sensing fiber 01 Module and LP 11 The Rayleigh scattered light of the mode is combined with the local oscillator light to obtain LP 01 Module and LP 11 Mode combined beam optical signal;

[0022] Step 9: LP obtained in step 8 01 Module and LP 11 After the combined mode optical signal is processed by optical signal processing, photoelectric conversion and electrical signal processing, the LP of the two-mode sensing fiber at each sampling point is obtained by sampling. 01 Module and LP 11 The time domain signal of the module;

[0023] Step 10: For each sampling point obtained in step 9, 01 Module and LP 11 The time domain signal of the model is analyzed by using the coefficient of variation method. 01 Module and LP 11 The time domain signal of the model is processed by the coefficient of variation, and the LP 01 Analog signal and LP 11 The mode signals are multiplied to obtain the disturbance curves along the two-mode sensing optical fiber;

[0024] Step 11: If there is a vibration peak in the disturbance curve obtained in step 10, the LP at each sampling point obtained in step 9 is 01 Module and LP 11 The vibration peak LP in the time domain signal of the mode 01 Module and LP 11 The time domain signal of the model is sent to steps 12 and 13 at the same time; if there is no vibration peak in the disturbance curve obtained in step 10, the data of this frame is ignored and the process returns to step 8;

[0025] Step 12: LP at the vibration peak 01 Module and LP 11 The variance and range of the time domain signal of the model are calculated to obtain the vibration peak P 01 Module and LP 11 Variance and range of the analog time domain signal;

[0026] Step 13: Use the empirical mode decomposition method to analyze the LP at the vibration peak. 01 Module and LP 11 The time domain signal of the module is decomposed to obtain LP 01 Module and LP 11 The eigenmode components of the mode vibration signal; and LP 01 Module and LP 11 After integrating the intrinsic mode components of the time domain signal, the vibration peak LP is obtained. 01 Module and LP 11 Energy value of the intrinsic mode component of the mode time domain signal;

[0027] Step 14: The vibration peak LP obtained in step 12 01 Module and LP 11The variance and range of the time domain signal and the LP at the vibration peak obtained in step 13 01 Module and LP 11 The energy value of the intrinsic mode component of the time domain signal is used as the measured data and is sent to the neural network for processing accordingly. The neural network determines and identifies whether the vibration at that location is vehicle vibration.

[0028] Compared with the prior art, the present invention has the following characteristics:

[0029] 1. Compared with single-mode fiber, dual-mode fiber has higher threshold level and reflected light capture rate. It can inject higher power detection pulse into dual-mode fiber to generate stronger backward Rayleigh scattering signal, which is beneficial to the later signal collection and processing. At the same time, dual-mode fiber allows two modes (LP 01 Module and LP 11 The optical transmission of the two modes can be separated into two modes for the upper computer to perform noise compensation, thereby reducing the misjudgment of a certain mode of signal due to noise, improving the signal-to-noise ratio of the system and the accuracy of vibration judgment.

[0030] 2. Use a combination of EMD and neural network to identify vibration signals. Compared with other signal analysis methods, EMD does not need to set basis functions for the analysis and extraction of signal features. It only decomposes the signal according to its own time scale characteristics. It is adaptive and especially suitable for nonlinear signals in this system. After extracting features through EMD, the features of the signal are used as eigenvalues ​​to input into the neural network for training. Compared with directly using the time domain signal of the vibration point as the eigenvalue for training, it reduces the number of eigenvalues ​​and a large amount of invalid data, and improves the speed and accuracy of training the neural network.

[0031] 3. Compared with the differential accumulation algorithm's processing of time domain signals, the coefficient of variation method does not require setting the differential k value, and does not have high requirements for the measured waveform. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a schematic structural diagram of a vehicle vibration monitoring device based on dual-mode optical fiber according to the present invention;

[0033] Figure 2 It is a flow chart of the training phase of a vehicle vibration monitoring method based on dual-mode optical fiber of the present invention;

[0034] Figure 3 To define relevant data types in the data processing method of the present invention;

[0035] Figure 4 It is a schematic diagram of decomposing the time domain data of the vibration point by the empirical mode decomposition method used in the present invention;

[0036] Figure 5The present invention is a flow chart of the actual measurement phase of vehicle vibration monitoring based on dual-mode optical fiber. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific examples.

[0038] A vehicle vibration monitoring device based on dual-mode optical fiber, such as Figure 1 As shown, it includes a high-power narrow linewidth laser, i.e., a laser, a pulse light generator, a circulator, a mode division multiplexer, a two-mode sensing fiber, an acquisition card, a digital signal processing unit, two polarization controllers, two photoelectric conversion and electrical signal processing modules, and four optical couplers. The output end of the high-power narrow linewidth laser is connected to the input end of the first optical coupler, one output end of the first optical coupler is connected to the input end of the pulse light generator, and the other output end of the first optical coupler is connected to the input end of the second optical coupler. The output end of the pulse light generator is connected to the first port of the optical circulator, and the second port of the optical circulator is connected to the LP of the mode division multiplexer. 01 The mode port of the mode division multiplexer is connected to the two-mode sensing optical fiber. The third port of the optical circulator is connected to an input end of the third optical coupler. The LP port of the mode division multiplexer is connected to the two-mode sensing optical fiber. 11 The mode port is connected to an input end of the fourth optical coupler. An output end of the second optical coupler is connected to another input end of the third optical coupler via the first polarization controller. Another output end of the second optical coupler is connected to another input end of the fourth optical coupler via the second polarization controller. Two output ends of the third optical coupler are simultaneously connected to two input ends of the first photoelectric conversion and electrical signal processing module, and two output ends of the first photoelectric conversion and electrical signal processing module are simultaneously connected to the acquisition card. Two output ends of the fourth optical coupler are simultaneously connected to two input ends of the second photoelectric conversion and electrical signal processing module, and two output ends of the second photoelectric conversion and electrical signal processing module are simultaneously connected to the acquisition card. The output end of the acquisition card is connected to the input end of the digital signal processing unit.

[0039] High power narrow linewidth lasers generate LP 01 In this embodiment, the high-power narrow-linewidth laser emits a continuous laser with a wavelength of 1550nm, a laser linewidth of 3kHz, an output power of 25mW, and its working mode is LP. 01 mold.

[0040] The first optical coupler splits the output laser of the high-power narrow-linewidth laser into two paths with a splitting ratio of 80:20. The 80% output end of the first optical coupler is connected to the input end of the pulse light generator to generate a detection light pulse. The 20% output end of the first optical coupler is connected to the input end of the second optical coupler to serve as an LP 01Mode and LP 11 The local oscillator light required for coherent detection of the two-way backward Rayleigh scattering signal of the mode.

[0041] The pulsed light generator is a fiber-coupled acousto-optic modulator with a modulation bandwidth of 200 MHz. The pulsed light generator modulates the continuous laser input by the first optical coupler into pulsed light. The pulse width of the pulsed light is 100 ns, the rise and fall times are both 2.5 ns, and the repetition pulse period is determined by the fiber length. Specifically, it is required that the repetition pulse period is greater than the time for light to travel back and forth in the fiber. In this embodiment, the fiber length is 2 km, and the time for light to travel back and forth in the fiber is approximately 19.34 us, so the pulse period is set to 20 us. While modulating into pulsed light, its output frequency is shifted down by 200 MHz.

[0042] The circulator transmits the LP mode optical pulse generated by the pulsed light generator 01 through the LP mode port of the mode division multiplexer into the two-mode sensing fiber. The LP mode backward Rayleigh scattering signal generated by the two-mode sensing fiber 01 is then returned to the second port of the circulator through the LP mode port of the mode division multiplexer and output to one input end of the third optical coupler through the third port of the circulator. The LP mode backward Rayleigh scattering signal generated by the two-mode sensing fiber 01 is sent to one input end of the fourth optical coupler through the LP mode port of the mode division multiplexer. 01 11 11

[0043] The mode division multiplexer supports the multiplexing and demultiplexing of LP mode and LP mode optical waves within the C band. The mode division multiplexer injects LP mode laser into the two-mode sensing fiber. The mode division multiplexer demultiplexes the LP mode and LP mode backward Rayleigh scattering signals returned from the two-mode sensing fiber. One LP mode backward Rayleigh scattering signal 01 11 01 is sent to the third optical coupler through the circulator via the LP mode port, and one LP mode backward Rayleigh scattering signal 01 11 is sent to the fourth optical coupler through the LP mode port. 01 01 11 11

[0044] The two-mode sensing fiber is laid along the road for vehicle sensing. For the two-mode sensing fiber in this embodiment, its normalized frequency V within the C band satisfies: 2.4048 < V < 3.8327, supporting the transmission of LP mode and LP mode optical waves, with a length of 2 km. 01 11 01When the mode light wave is transmitted forward in the optical fiber, Rayleigh scattering will occur continuously, and the Rayleigh scattered light exists in the LP 01 and LP 11 In both modes, LP 01 Module and LP 11 The backward Rayleigh scattered light propagates along the sensing fiber back to the fiber injection end.

[0045] The second optical coupler splits the laser output by the first optical coupler into the first local oscillation light and the second local oscillation light, with a splitting ratio of 50:50, and two output ends of the second optical coupler are respectively connected to the input ends of the first polarization controller and the second polarization controller.

[0046] The first polarization controller PC1 performs polarization control on the first local oscillation light output by the second optical coupler.

[0047] The second polarization controller PC2 performs polarization control on the second local oscillation light output by the second optical coupler.

[0048] The third optical coupler first couples the first local oscillator light output by the first polarization controller and the LP light output by the third port of the circulator. 01 The backward Rayleigh scattered light is combined, and the electric field equation of the combined light wave is equation (1). The combined light is then split 1:1, and the electric field equation of the split light wave is equation (2) and equation (3). The two paths of split light have a phase difference of π / 2.

[0049] E=E s exp(iω s t+φ S )+E L exp(iω L t+φ L ) (1)

[0050]

[0051]

[0052] Among them, E s is the signal light amplitude, E L is the local oscillation amplitude.

[0053] The fourth optical coupler first couples the second local oscillator light output by the second polarization controller to the mode division multiplexer LP 11 Mode port output LP 11 The backward Rayleigh scattered light is combined, and then the combined light is split 1:1, and the two paths of split light have a phase difference of π / 2. The electric field equation of the light wave after the fourth optical coupler is combined and split as described in the third optical coupler.

[0054] The first photoelectric conversion and electrical signal processing module first performs photoelectric conversion and beat frequency detection on the two optical signals output by the third optical coupler with a phase difference of π / 2 to reduce the common mode noise of the system and improve the signal-to-noise ratio, detection sensitivity and dynamic range. Then the electrical signal is amplified, filtered and IQ demodulated, and the demodulated IQ two-way RF signal is sent to the acquisition card. The RF signal carries the LP signal in the two-mode optical fiber. 01 Mode backscattered Rayleigh light information.

[0055] The second photoelectric conversion and electrical signal processing module first performs photoelectric conversion and beat frequency detection on the two optical signals output by the fourth optical coupler with a phase difference of π / 2 to reduce the common mode noise of the system and improve the signal-to-noise ratio, detection sensitivity and dynamic range. Then the electrical signal is amplified, filtered and IQ demodulated, and the demodulated IQ two-way RF signal is sent to the acquisition card. The RF signal carries the LP signal in the two-mode optical fiber. 11 Mode backscattered Rayleigh light information.

[0056] The acquisition card performs A / D conversion on both the IQ two-channel RF signals sent by the first photoelectric conversion and electrical signal processing module and the IQ two-channel RF signals sent by the second photoelectric conversion and electrical signal processing module, and sends them to the digital signal processing module. In addition, the acquisition card sets different sampling rates for the two modes of backward Rayleigh scattering signals to compensate for the different propagation speeds of the two modes of light in the optical fiber. The acquisition card will continuously collect a portion of the data and then centrally transmit the data to the digital signal processing module. In this example, the acquisition card provides a maximum real-time sampling rate of 10GS / s on each of the four channels and a record length of 500M points.

[0057] The digital signal processing module processes the digital signal sent by the acquisition card, including the extraction of sampling point time domain signals, feature analysis, and neural network training in the training phase, and the judgment of vibration position, vibration position time domain signal feature analysis, and neural network processing to identify vibration signals in the actual measurement phase.

[0058] The working process of the above vehicle vibration monitoring device based on dual-mode optical fiber is as follows:

[0059] High power narrow linewidth lasers generate LP 01 The continuous laser of the mode is input into the first optical coupler, and the first optical coupler divides the laser output by the high-power narrow-linewidth laser into two: in the first branch, the pulse light generator modulates the laser output by the first optical coupler (80%) into a detection pulse, and sequentially passes through the first port of the circulator, the second port of the circulator and the mode division multiplexer LP 01 The LP in the two-mode sensing fiber is injected into the two-mode sensing fiber. 01 The backscattered Rayleigh light of the mode is sequentially transmitted through the mode division multiplexer LP 01The LP in the two-mode sensing fiber is input to an input end of the third optical coupler. 11 The backscattered Rayleigh light is sent to the mode division multiplexer LP 11 The mode port is input to an input end of the fourth optical coupler. In the second branch, the second optical coupler divides the other part of the laser (20%) separated by the first optical coupler into two local oscillator lights, which are recorded as the first local oscillator light and the second local oscillator light. The first local oscillator light and the second local oscillator light are input to the third optical coupler and the fourth optical coupler after polarization control by the first polarization controller and the second polarization controller respectively. The third optical coupler combines the first local oscillator light and the LP light returned by the third port of the circulator. 01 The fourth optical coupler combines the second local oscillator light and the mode multiplexer LP 11 LP returned by the module port 11 The backward Rayleigh scattered light is combined and input into the second photoelectric conversion and electrical signal processing module. The first photoelectric conversion and electrical signal processing module and the second photoelectric conversion and electrical signal processing module send the processed optical signal into the acquisition card for acquisition and then input into the digital signal processing unit for signal processing.

[0060] A vehicle vibration monitoring method based on dual-mode optical fiber implemented by the above device includes the following specific steps:

[0061] Step 1: LP 01 The continuous laser of the mode is divided into two paths: one path is modulated to generate LP 01 The first mode light pulse is injected into the two-mode sensing fiber, and the other mode light pulse is polarized to generate local oscillator light.

[0062] (1) Training phase (see Figure 2 ):

[0063] Step 2: The two-mode optical fiber is buried on one side of the road. A sample vehicle is allowed to pass through a sampling point of the two-mode sensing optical fiber laid on the road surface. The ground generates vibrations that disturb the two-mode sensing optical fiber, changing the refractive index of the two-mode optical fiber at that point, thereby changing the backward Rayleigh scattered light occurring in the optical fiber at that point, LP 01 Module and LP 11 The backward Rayleigh scattered light of the mode carries vibration information.

[0064] Step 3: LP carrying vibration information 01 and LP 11 The Rayleigh scattered light of the backward mode is transmitted back to the injection end of the two-mode sensing fiber and combined with the local oscillator light to obtain LP 01 Module and LP 11 Mode combined optical signal.

[0065] Step 4: LP 01 Module and LP 11 After the combined mode optical signal is processed by optical signal processing, photoelectric conversion and electrical signal processing, the LP at the sampling point of the two-mode sensing optical fiber where the sample vehicle is located is obtained by sampling. 01 Module and LP 11 The vibration time domain signal of the module.

[0066] Different sampling rates are designed for backscattered Rayleigh signals of different light wave modes to compensate for LP 01 Module and LP 11 The speed difference of LP mode in the sensing fiber. 01 Module and LP 11 The modes have different effective refractive indices, and their propagation speeds in the sensing optical fiber are inconsistent. If they are not compensated, the two signals will be directly integrated after data processing, which will reduce the signal-to-noise ratio and result in a phenomenon where one vibration source corresponds to two vibration peaks. In response to this problem, the present invention proposes a compensation method that uses different sampling rates for different light wave modes. Specifically, the sampling frequency of the acquisition card is set to ensure that its sampling period is the time it takes for the corresponding mode light wave to propagate 1 meter in the sensing optical fiber, that is, each data collected by the acquisition card corresponds to the Rayleigh scattering signal of a point on the sensing optical fiber, and each point is 1 meter apart. Sampling frequency The calculation formula is:

[0067]

[0068]

[0069] Where i is the propagation mode of light in the optical fiber (i = 01 means LP 01 Module. i=11 means LP 11 mode), c is the propagation speed of light in vacuum, is the effective refractive index of the corresponding mode light wave in the sensing fiber. From equations (4) and (5), it can be determined that the acquisition card for LP 01 Module and LP 11 The sampling frequency of the module is set.

[0070] In this embodiment, the sampling period is set to the time for the corresponding mode light wave to propagate 1 meter in the sensing optical fiber. Specifically, LP 01 The effective refractive index of the mode in the sensing fiber is 1.4498, LP 11 The effective refractive index of the mode is 1.4468, so the propagation time of the two modes in the optical fiber for 1m is 4.8360×10 -9 s and 4.8260×10 -9 s, and thus the sampling rates of the two modes should be set to 1.0339×10 8Hz and 1.0361×10 8 Hz.

[0071] In order to better describe the data, the data is defined as follows: Figure 3 As shown in the figure, suppose the acquisition card collects a Rayleigh scattering signal of a light wave mode returned by a single detection pulse as a backscattering data. Suppose the acquisition card collects two modes (LP 01 Module and LP 11 The Rayleigh scattering signal of the backscattering mode is a frame of backscattering Rayleigh scattering data. Assume that the time domain data corresponding to a specific sampling point in n backscattering Rayleigh scattering data is a sampling point time domain data. As can be seen from the above, the number of sampling points in a backscattering Rayleigh scattering data is determined by the length of the optical fiber. In this embodiment, the length of the sensing optical fiber is 2 km, there are 2000 sampling points in a backscattering Rayleigh scattering data, and there are 4000 sampling points in a frame of data. The number of time domain data of a sampling point is determined by the value of n.

[0072] The acquisition card continuously acquires n frames of backscattered Rayleigh scattering data to ensure that the two adjacent backscattered Rayleigh scattering data in each light wave mode are the amplitude information of the backscattered Rayleigh scattering signal generated by the adjacent detection pulses, thereby determining the sampling period of the time domain signal at the sampling point (the sampling period is the repetition period of the detection pulse light). The acquisition process is stored by the acquisition card and is centrally transmitted to the digital signal processing unit for signal processing after the acquisition is completed.

[0073] In this embodiment, the design rule of the n value is: the vibration frequency of the vehicle to the ground is in the range of 0-100 Hz, wherein the main energy is concentrated in the low-frequency range near 50 Hz. Half a cycle is sufficient to describe the characteristics of a signal, and the time to collect n data is half of the vibration signal cycle (not limited thereto). The sensing medium of this embodiment is 2 km of optical fiber, and the repetition period of the pulse is 20 μs. For a vibration signal of about 50 Hz, the n value is 10 ms / 20 μs, which is equal to 500. That is, each light wave mode collects 500 backward Rayleigh scattering data each time and transmits them to the digital signal processing unit, and the number of time domain data of a certain sampling point in each light wave mode is 500.

[0074] In this embodiment, the vehicle vibration occurs at 1 km of the optical fiber, so the time domain signal of the data at 1 km of the optical fiber is extracted, and the sampling frequency of each mode is the distance of 1 m that the corresponding light wave mode propagates in the optical fiber, so the data of the time domain signal at 1 km is the set of the 1000th data of each backward Rayleigh scattering signal, and the number of time domain data for each mode is 500.

[0075] Step 5: LP obtained in step 4 01 Module and LP 11 The variance and range of the vibration time domain signal of the model are calculated to obtain LP01 Module and LP 11 Variance and range of the analog time domain signal.

[0076] The range calculation formula is:

[0077] R = max(x i )-min(x i ) (6)

[0078] The variance calculation formula is:

[0079]

[0080] Among them, x i Represents a time domain data, is the average value of a time domain number, and n is the amount of data in a data line.

[0081] Step 6: Use the empirical mode decomposition (EMD) method to analyze the LP obtained in step 4. 01 Module and LP 11 The vibration point time domain signal of the model is decomposed to obtain LP 01 Module and LP 11 The eigenmode component of the mode vibration signal. 01 Module and LP 11 After integrating the intrinsic mode component of the mode vibration signal, we get LP 01 Module and LP 11 The energy value of the eigenmode component of the mode vibration signal.

[0082] Take the time domain data at the vehicle vibration point, use the empirical mode decomposition method to reduce noise and extract features, and add the variance and range of the time domain signal as eigenvalues ​​to input into the subsequent neural network for training. After empirical mode decomposition, each time domain data can be represented as multiple eigencomponents (IMF i ) and a linear sum of a trend term, that is, a non-stationary signal x(t) can be expressed as:

[0083]

[0084] Among them, IMF1 has the highest frequency, and the frequencies decrease thereafter. γ(t) is a trend term with an extremely low frequency.

[0085] Specifically, extract LP 01 Module and LP 11 The time domain signal data of the vehicle vibration point of the mode are subjected to empirical mode decomposition to obtain two groups of IMFs. After eliminating the extremely low-frequency components, the two groups of remaining IMFs are integrated to obtain the energy value of the eigenmode component of the vibration signal of the corresponding light wave mode.

[0086] Steps 5 and 6 are to extract features from the time domain signal. The feature extraction process is as follows: Figure 2 As shown. i (x) is the amplitude function of the vibration point time domain signal of the corresponding light wave mode (i = 01 represents LP 01 Module. i=11 means LP 11 mold). is EMD to F i (x)Eigenmode functions of various orders obtained by decomposition. is the integral value of the eigenmode function of each order of the corresponding mode. and R i are the vibration time domain signals F corresponding to the light wave mode i The variance and range of (x) are obtained using the LP 01 Module and LP 11 The energy value, variance and range corresponding to the mode are input into the neural network as feature values ​​for training.

[0087] In this embodiment, firstly, one time domain data of each of the two light wave mode vibration points is taken, and the variance and range are calculated respectively. Then, the two data are decomposed and feature extracted using EMD, and 9-10 IMFs (including trend items with very low frequency) are obtained for each data. The EMD decomposition effect diagram of one data is shown as follows: Figure 4 As shown in the figure, IMF1-IMF8 is taken as the signal features, and integrated to obtain 16 (two modes, 8 in one mode) energy values. The 16 energy values ​​plus the variance and range of the corresponding mode time domain data, a total of 20 feature values, are input as a sample into the neural network for training.

[0088] Step 7: Substitute the LP obtained in step 5 01 Module and LP 11 The variance and range of the modulus and the LP obtained in step 6 01 Module and LP 11 The energy values ​​of the intrinsic mode components of the mode vibration signal are correspondingly input into the neural network for training to obtain a trained neural network.

[0089] In the present embodiment, the above-mentioned eigenvalues ​​are trained using an improved neural network algorithm based on Dropout and ADAM optimizer (not limited to this neural network), the number of hidden layer nodes of the network is set to 30, the input layer activation function is ReLU, the output layer activation function is the sigmoid function, Dropout is set to 0.4, and the ADAM algorithm is used to optimize it. The number of iterations is first set to 5000, and the optimal number of iterations is selected by the curve graph of the number of iterations and the accuracy. The number of training sets, cross validation sets, and test sets meets 6:2:2.

[0090] The trained neural network is obtained through the above steps and stored in the data signal processing unit of the vehicle vibration monitoring device based on dual-mode optical fiber.

[0091] (2) Measurement phase (see Figure 5 ):

[0092] Step 8: Real-time acquisition of LP returned from the dual-mode sensing fiber 01 Module and LP 11 The Rayleigh scattered light of the mode is combined with the local oscillator light to obtain LP 01 Module and LP 11 Mode combined optical signal.

[0093] The raw data collected in the training phase is data carrying the vehicle vibration signal. The raw data collected in this step may carry a vibration signal or may not carry a vibration signal. If it carries a vibration signal, it may be a vibration signal of a vehicle or a vibration signal of a non-vehicle.

[0094] Step 9: LP 01 Module and LP 11 After the combined mode optical signal is processed by optical signal processing, photoelectric conversion and electrical signal processing, the LP at each sampling point of the two-mode sensing fiber is obtained by sampling. 01 Module and LP 11 The time domain signal of the model.

[0095] Different sampling rates are designed for backscattered Rayleigh signals of different light wave modes to compensate for LP 01 Module and LP 11 The sampling rate parameter setting must be the same as the sampling rate set during the training phase.

[0096] The number of frames collected by the acquisition card during the measurement phase should be the same as the number of frames collected by the acquisition card during the training phase. In this example, 500 frames of data are collected each time during the training phase, and the measurement phase should also collect 500 frames of data each time and then process them.

[0097] Step 10: For each LP at each sampling point 01 Module and LP 11 The time domain signal of the model is processed by the coefficient of variation method, and the two LP 01 Module and LP 11 The mode signals are multiplied to obtain a disturbance curve along the optical fiber with a higher signal-to-noise ratio, and the location of the vibration point is determined by this curve.

[0098] The vibration signal is located based on the coefficient of variation method. When the neural network obtained in the training phase is used to identify the vibration signal along the optical fiber, the digital signal processing unit cannot directly perform empirical mode decomposition, feature extraction and neural network recognition on the time domain data at all sampling points. Otherwise, the longer the optical fiber length and the more sampling points, the more calculations will be multiplied. The present invention first uses the coefficient of variation method to perform preliminary processing on the time domain data of each sampling point to determine the location where vibration may occur, and then extracts the time domain data of the location where vibration may occur for neural network calculation.

[0099] The discrete degree of the time domain data at the vibration point is much greater than that at the non-vibration point. The coefficient of variation method in statistics is used to calculate the discrete degree of the time domain data and determine the location where vibration may occur. This method does not need to consider the influence of measurement scale and dimension. The specific calculation method is: calculate LP 01 and LP 11 The coefficient of variation of the time domain data at each point along the optical fiber (the sampling rate designed in step 4, in this example, the interval between two adjacent points is 1m) is used to determine the location information of the vibration along the optical fiber. The calculation formula of the coefficient of variation CV is:

[0100]

[0101]

[0102] Among them, x i Represents a time domain data, is the average value of a time domain number, and n is the amount of data in a data. 01 and LP 11 The vibration information along the optical fiber corresponding to the mode is obtained. Then the data processing results of the two light wave modes are multiplied to remove the noise interference in a single light wave mode, obtain a vibration peak with a higher signal-to-noise ratio, and obtain the accurate position of the vibration peak.

[0103] Step 11: As described in step 8, the data obtained in the actual measurement stage does not necessarily carry a vibration signal. Therefore, if there is a vibration peak in the disturbance curve obtained in step 10, extract the LP at the vibration peak from the original data obtained in step 9. 01 Module and LP 11 The time domain signal of the model is sent to steps 12 and 13. If no vibration peak is obtained in the disturbance curve obtained in step 10, it means that the current frame data does not carry a vibration signal, and the current frame data is ignored, and the subsequent steps are not entered, and the process returns to step 8.

[0104] Step 12: Extract the LP of the vibration point from the raw data obtained in step 9 01 Module and LP 11 The time domain data of the model is respectively 01 Module and LP11 The variance and range of the time domain signal of the model are calculated to obtain LP 01 Module and LP 11 Variance and range of the model.

[0105] Step 13: Extract the LP of the vibration point from the raw data obtained in step 9 01 Module and LP 11 The empirical mode decomposition method is used to analyze the LP 01 Module and LP 11 The time domain signal of the module is decomposed to obtain LP 01 Module and LP 11 The eigenmode component of the mode vibration signal. 01 Module and LP 11 After integrating the intrinsic mode component of the mode vibration signal, we get LP 01 Module and LP 11 The energy value of the eigenmode component of the mode vibration signal.

[0106] The feature extraction method of the signal at the extracted vibration point in steps 12 and 13 is the same as the feature extraction method of the signal at the sample sampling point in steps 5 and 6.

[0107] Step 14: LP obtained in step 12 01 Module and LP 11 The variance and range of the modulus and the LP obtained in step 13 01 Module and LP 11 The energy value of the intrinsic mode component of the mode vibration signal is sent to the neural network for processing, and the neural network determines and identifies whether the vibration at that location is vehicle vibration.

[0108] It should be noted that although the embodiments of the present invention described above are illustrative, they are not intended to limit the present invention, and therefore the present invention is not limited to the above specific embodiments. Without departing from the principles of the present invention, any other embodiments obtained by those skilled in the art under the guidance of the present invention are deemed to be within the protection of the present invention.

Claims

1. A vehicle vibration monitoring method based on a dual-mode optical fiber, wherein the vehicle vibration monitoring device comprises a laser, a pulse light generator, a circulator, a mode division multiplexer, a dual-mode sensing optical fiber, an acquisition card, a digital signal processing unit, two polarization controllers, two photoelectric conversion and electrical signal processing modules, and four optical couplers; the output end of the laser is connected to the input end of the first optical coupler, one output end of the first optical coupler is connected to the input end of the pulse light generator, and the other output end of the first optical coupler is connected to the input end of the second optical coupler; the output end of the pulse light generator is connected to the first port of the optical circulator, and the second port of the optical circulator is connected to the LP of the mode division multiplexer. 01 The multiplexing port of the mode division multiplexer is connected to the two-mode sensing optical fiber; the two-mode sensing optical fiber is laid along the road surface; the third port of the optical circulator is connected to an input end of the third optical coupler, and the LP of the mode division multiplexer is connected to the input end of the third optical coupler. 11 The mode port is connected to an input end of the fourth optical coupler; an output end of the second optical coupler is connected to another input end of the third optical coupler via the first polarization controller; another output end of the second optical coupler is connected to another input end of the fourth optical coupler via the second polarization controller; two output ends of the third optical coupler are simultaneously connected to two input ends of the first photoelectric conversion and electrical signal processing module, and two output ends of the first photoelectric conversion and electrical signal processing module are simultaneously connected to the acquisition card; two output ends of the fourth optical coupler are simultaneously connected to two input ends of the second photoelectric conversion and electrical signal processing module, and two output ends of the second photoelectric conversion and electrical signal processing module are simultaneously connected to the acquisition card; the output end of the acquisition card is connected to the input end of the digital signal processing unit; Its characteristics are: Included steps as follows: Step 1: LP 01 The continuous laser of the mode is divided into two paths: one path is modulated to generate LP 01 Mode light pulses are injected into the two-mode sensing optical fiber; the other path generates local oscillator light after polarization control; Step 2: Let the sample vehicle pass through a sampling point of the two-mode sensing optical fiber laid on the road surface, that is, the sample sampling point. The ground generates vibrations that disturb the two-mode sensing optical fiber, thereby generating LP carrying vibration information. 01 Module and LP 11 The Rayleigh scattered light in the backward direction; Step 3: LP carrying vibration information 01 and LP 11 The Rayleigh scattered light of the backward mode is transmitted back to the injection end of the two-mode sensing fiber and combined with the local oscillator light to obtain LP 01 Module and LP 11 Mode vibration combined optical signal; Step 4: LP obtained in step 3 01 Module and LP 11 After the optical signal processing, photoelectric conversion and electrical signal processing, the two-mode sensing fiber LP at the sample sampling point is obtained by sampling. 01 Module and LP 11 The vibration time domain signal of the module; Step 5: LP at the sample sampling points obtained in step 4 01 Module and LP 11 The variance and range of the vibration time domain signal of the model are calculated to obtain the LP at the sample sampling point. 01 Module and LP 11 Variance and range of the analog time domain signal; Step 6: Use the empirical mode decomposition method to calculate the LP at the sample sampling points obtained in step 4. 01 Module and LP 11 The vibration time domain signal of the model is decomposed to obtain LP 01 Module and LP 11 The eigenmode components of the mode vibration signal; and LP 01 Module and LP 11 After the intrinsic mode component of the mode vibration signal is integrated, the LP at the sample sampling point is obtained. 01 Module and LP 11 Energy value of the eigenmode component of the modal vibration signal; Step 7: Place the sample sampling point obtained in step 5 at LP 01 Module and LP 11 The variance and range of the time domain signal and the LP at the sample sampling point obtained in step 6 01 Module and LP 11 The energy value of the intrinsic mode component of the mode vibration signal is used as sample data and is correspondingly input into the neural network for training to obtain a trained neural network; Step 8: Real-time acquisition of LP returned from the dual-mode sensing fiber 01 Module and LP 11 The Rayleigh scattered light of the mode is combined with the local oscillator light to obtain LP 01 Module and LP 11 Mode combined beam optical signal; Step 9: LP obtained in step 8 01 Module and LP 11 After the combined mode optical signal is processed by optical signal processing, photoelectric conversion and electrical signal processing, the LP of the two-mode sensing fiber at each sampling point is obtained by sampling. 01 Module and LP 11 The time domain signal of the module; Step 10: For each sampling point obtained in step 9, 01 Module and LP 11 The time domain signal of the model is analyzed by using the coefficient of variation method. 01 Module and LP 11 The time domain signal of the model is processed by the coefficient of variation, and the LP 01 Analog signal and LP 11 The mode signals are multiplied to obtain the disturbance curves along the two-mode sensing optical fiber; Step 11: If there is a vibration peak in the disturbance curve obtained in step 10, the LP at each sampling point obtained in step 9 is 01 Module and LP 11 The vibration peak LP in the time domain signal of the mode 01 Module and LP 11 The time domain signal of the model is sent to steps 12 and 13 at the same time; if there is no vibration peak in the disturbance curve obtained in step 10, the data of this frame is ignored and the process returns to step 8; Step 12: LP at the vibration peak 01 Module and LP 11 The variance and range of the time domain signal of the model are calculated to obtain the vibration peak P 01 Module and LP 11 Variance and range of the analog time domain signal; Step 13: Use the empirical mode decomposition method to analyze the LP at the vibration peak. 01 Module and LP 11 The time domain signal of the module is decomposed to obtain LP 01 Module and LP 11 The eigenmode components of the mode vibration signal; and LP 01 Module and LP 11 After integrating the intrinsic mode components of the time domain signal, the vibration peak LP is obtained. 01 Module and LP 11 Energy value of the intrinsic mode component of the mode time domain signal; Step 14: The vibration peak LP obtained in step 12 01 Module and LP 11 The variance and range of the time domain signal and the LP at the vibration peak obtained in step 13 01 Module and LP 11 The energy value of the intrinsic mode component of the time domain signal is used as the measured data and is correspondingly sent to the neural network for processing. The neural network determines and identifies whether the vibration at that location is vehicle vibration.

2. A vehicle vibration monitoring method based on dual-mode optical fiber according to claim 1, characterized in that: The splitting ratio of the first optical coupler is 80:20, wherein the 80% output end is connected to the input end of the pulse light generator, and the 20% output end is connected to the input end of the second optical coupler.

3. A vehicle vibration monitoring method based on dual-mode optical fiber according to claim 1, characterized in that: The splitting ratio of the second optical coupler, the third optical coupler and the fourth optical coupler is 50:

50.

4. The vehicle vibration monitoring method based on dual-mode optical fiber according to claim 1 is characterized in that: The pulse light generator is a fiber-coupled acousto-optic modulator, and the pulse period of the pulse light generated by the pulse light is greater than the time it takes for the light to propagate back and forth in the two-mode sensing optical fiber.

5. The vehicle vibration monitoring method based on dual-mode optical fiber according to claim 1 is characterized in that: The two output ends of the first photoelectric conversion and electrical signal processing module output two IQ RF signals, both of which carry LP 01 The two output ends of the second photoelectric conversion and electrical signal processing module output two IQ RF signals, both of which carry LP 11 Mode backscattered Rayleigh light information.

Images