A distributed optical fiber microseismic system adaptive dynamic noise reduction method

Abstract

The present application relates to the field of optical fiber microseism, and more particularly to a kind of distributed optical fiber microseismic system adaptive dynamic noise reduction method, the whole optical fiber link is divided into equal length section, obtain its time series signal;The amplitude spectrum of each frame signal in each monitoring channel is calculated;The whole optical fiber link is divided into service monitoring section and signal blind area;In service monitoring section, select front reference optical fiber and rear reference optical fiber, calculate the front reference optical fiber noise power spectrum and rear reference optical fiber noise power spectrum, and then obtain the noise amplitude spectrum;The amplitude of each frame signal in service monitoring section is executed spectrum subtraction processing, the noise reduction amplitude and the lower limit of residual noise are controlled by spectrum subtraction factor and spectrum bottom factor, the net signal amplitude spectrum is obtained, and then the noise reduction complex spectrum is reconstructed;After the noise reduction complex spectrum and signal blind area spectrum are spliced, inverse fourier transform is executed to restore to time domain, and signal noise reduction processing of the whole optical fiber link is realized.The present application reduces the noise influence caused by instability.

Description

Technical Field

[0001] This invention relates to the field of fiber optic microseismic systems, and more particularly to an adaptive dynamic noise reduction method for distributed fiber optic microseismic systems. Background Technology

[0002] Distributed fiber optic microseismic systems primarily utilize high-sensitivity fiber optic sensor networks to monitor in real-time weak seismic wave signals generated within rock masses, soils, or engineering structures due to stress changes and damage / fracture. This allows for precise location of micro-fracture events, stability assessment, and disaster warning. In the early stages of seismic wave formation, the signals are extremely weak, and there is significant interference from the surrounding environment. This means that the microseismic signals acquired by the distributed fiber optic microseismic system along the optical cable are affected not only by the system itself (e.g., light source phase noise and shot noise) but also by the superposition of various factors such as environmental vibrations, resulting in a low signal-to-noise ratio in the raw data. This affects the sensitivity and false alarm rate of microseismic event detection. Current noise reduction methods include bandpass filtering, notch filtering, wavelet thresholding, wavelet packet decomposition, singular spectrum analysis, and deep learning. While these methods each have advantages for specific types of noise, they generally rely on statistically stable noise or require extensive training samples. When the noise composition is complex and varies with time and distance, the denoising effect is prone to fluctuations, and algorithm parameters require frequent adjustments.

[0003] In long-distance optical cable lines, there are often several reference fibers located far from the vibration source and almost unaffected by microseismic events. Noise continuously acquired from these reference fibers can be used to characterize the dynamic changes in system noise in real time. If the signal from these reference fibers is used as a noise sample to process the signal in the section to be monitored, it is hoped that various non-stationary noises can be dynamically and accurately canceled, providing a higher signal-to-noise ratio basis for microseismic events. While current literature on fiber optic microseismic signal denoising has explored various methods, it has not fully utilized available reference fiber resources and lacks adaptive spectral subtraction implementations for real-time engineering scenarios. Therefore, spectral subtraction denoising methods based on reference fibers are necessary for further research and engineering applications, providing a new technical approach for high-reliability monitoring of distributed microseismic systems under harsh environments and low signal-to-noise ratio conditions. Summary of the Invention

[0004] The present invention aims to address the above-mentioned problems by proposing an adaptive dynamic noise reduction method for distributed optical fiber microseismic systems.

[0005] The technical solution of this invention is as follows: An adaptive dynamic noise reduction method for a distributed fiber optic microseismic system is as follows: The entire optical fiber link is divided into segments of equal length, each segment serving as a monitoring channel. Vibration signals are collected synchronously in each monitoring channel to obtain its time series signal. After processing each time series signal in frames, Fourier transform is performed to calculate the amplitude spectrum of each frame signal in each monitoring channel. The entire fiber optic link is divided into service monitoring segments and signal blind spots; Within the service monitoring section, a front reference fiber and a back reference fiber are selected, and the noise power spectrum of the front reference fiber and the noise power spectrum of the back reference fiber are calculated, thereby obtaining the noise amplitude spectrum. The amplitude of each frame signal within the service monitoring segment is subjected to spectral subtraction processing. The noise reduction amplitude and the lower limit of residual noise are controlled by the spectral subtraction factor α and the spectral base factor β to obtain the net signal amplitude spectrum; then the noise-reduced complex spectrum is reconstructed. After splicing the denoised complex spectrum with the signal dead zone spectrum, an inverse Fourier transform is performed to restore it to the time domain, thus achieving signal denoising processing for the entire fiber optic link.

[0006] The specific process for obtaining the amplitude spectrum of each frame signal is as follows: each time series signal is divided into frame signals with a fixed frame length and a 50% overlap rate, and a window function is applied to each frame signal. Then, a Fourier transform is performed to transform the time domain to the frequency domain, and the amplitude spectrum of each frame signal in each monitoring channel is calculated.

[0007] The principle for dividing the service monitoring segment is as follows: the average energy of each monitoring channel during the observation time is obtained based on the amplitude spectrum of each frame signal in each monitoring channel, and the signal blind zone in the entire optical fiber link is identified based on the preset energy threshold, with the non-signal blind zone being the service monitoring segment.

[0008] The principle for dividing the service monitoring segment is as follows: there is a blind zone at the beginning of the entire optical fiber link and a blind zone at the end; a blind zone decision threshold T is set at the beginning. lead and end-point blind zone decision threshold T trail ; Starting from the beginning of the entire fiber optic link, when the average energy of the continuously monitored channel drops to the beginning-end blind zone decision threshold T... lead The following is recorded as the starting point. ; Starting from the end of the entire fiber optic link, the search proceeds in reverse order until the average energy of the continuously monitored channel drops to the end-blind zone decision threshold T. trail The following is recorded as the end point. Starting point To the finish line The section between these two points is the service monitoring segment containing valid signals.

[0009] The front and rear reference fibers contain only background noise, and the front reference fiber starts from the point of origin. Take 100 m from the starting point, then refer to the fiber optic cable from the endpoint. Take 100 m forward from the starting point.

[0010] The specific process for obtaining the noise amplitude spectrum is as follows: An autoregressive model is established for the background noise time-series signals of the front and rear reference fibers using a 20th-order Burg algorithm, respectively, to obtain the noise power spectra of the front and rear reference fibers; the noise power spectra of the front and rear reference fibers are then linearly combined according to their respective weights to obtain the noise amplitude spectrum. The specific process is as follows: In the formula: Noise amplitude spectrum; The weighting coefficients for the noise power spectrum of the front reference fiber; The noise power spectrum of the preceding reference fiber; For the weighting coefficients of the noise power spectrum of the subsequent reference fiber; For subsequent reference fiber noise power spectrum; For the preceding reference fiber energy; For subsequent reference fiber energy and...

[0011] Wherein, the spectral reduction factor α ≥ 1, 0 < spectral base factor β < 1; the spectral reduction factor α is dynamically adjusted with the signal-to-noise ratio; when the signal-to-noise ratio of a certain frame signal is < 40, the value of the spectral reduction factor α is 1~45.5; when the signal-to-noise ratio of a certain frame signal is ≥ 40, the value of the spectral reduction factor α is 1.

[0012] The specific process for obtaining the reconstructed, denoised complex spectrum is as follows: The original amplitude spectrum and original phase spectrum of each frame signal within the service monitoring segment are acquired; a spectral subtraction operation is performed on the original amplitude spectrum; if the original amplitude of a certain frame signal is higher than its noise amplitude, then the net signal amplitude spectrum is the sum of the original amplitude and the noise amplitude. The difference between the amplitudes of two times the noise amplitude; if the original amplitude of a certain frame of signal is lower than the noise amplitude, then the net signal amplitude spectrum is... The noise amplitude is multiplied by 1, and the net signal amplitude spectrum is ≥0; the original phase spectrum and net signal amplitude spectrum of each frame signal are reconstructed into a denoised complex spectrum.

[0013] The specific process of performing spectral subtraction on the original amplitude spectrum is as follows: In the formula: The net signal amplitude spectrum; The original amplitude spectrum; This is the noise amplitude spectrum.

[0014] The specific process of splicing the noise-reduced complex spectrum with the signal blind zone spectrum is as follows: the original spectrum of the signal blind zone is preserved and spliced ​​with the noise-reduced complex spectrum.

[0015] The technical effects of this invention are as follows: 1) By introducing adaptive spectral subtraction of the front and rear reference fibers, the signal-to-noise ratio of the distributed fiber microseismic signal is improved by about 8.92 dB, which significantly suppresses light source phase noise, shot noise and environmental vibration interference, enhances the characteristics of weak microseismic events, and provides a higher quality data foundation for subsequent microseismic type identification and accurate positioning. 2) Dynamically reduce the noise impact caused by system instability. Attached Figure Description

[0016] Figure 1 This is a flowchart of an adaptive dynamic noise reduction method for a distributed optical fiber microseismic system according to the present invention.

[0017] Figure 2 This is a spectrum diagram of an adaptive dynamic noise reduction method for a distributed optical fiber microseismic system according to the present invention. Detailed Implementation

[0018] Specific application case: An adaptive dynamic noise reduction method for a distributed optical fiber microseismic system, which reduces noise across the entire optical fiber link according to the method provided in this paper.

[0019] Step 1: Assume the entire fiber optic link is of length L, divided into M equal segments, each segment serving as a monitoring channel. Each monitoring channel synchronously acquires vibration signals (assuming a sampling frequency f). s = 1000 Hz), the time series signal of each monitoring channel is denoted as ;in =1,2,...,M represents the monitoring channel number; To obtain a stable spectrum in the frequency domain, the time series signal of each monitoring channel is processed by fixed-length framing: a fixed frame length N is selected (e.g., N = 512), adjacent frames overlap by N / 2 (i.e., 50% overlap, overlap length 256), and portions less than one frame are padded with zeros. For the... The first monitoring channel Frame signal multiplied by Hamming (Window length = fixed frame length N) is denoted as Each frame of signal is represented in the following form: (1) In the formula: For time sampling points; For window functions; right Perform a Fourier transform to convert the time-domain vibration signal to the frequency domain. The first monitoring channel The amplitude spectrum of the frame signal is denoted as The definition is as follows: (2) In the formula: For discrete frequencies in the frequency domain; Let be the imaginary part of the complex number, satisfying ; Then the first The first monitoring channel Power spectrum of frame signal , used to reflect the first The first monitoring channel The energy of each frequency component of the frame signal.

[0020] Step 2: Divide the entire fiber optic link into service monitoring segments and signal blind spots; To assess the overall signal energy level of each monitoring channel, Averaging is performed in the frequency domain and over time frames. Specifically, the first... Average energy of each monitoring channel during the observation period Represented as: (3) In the formula: It is the total number of time frames; For the first The first monitoring channel The total energy of the frame signal; Because fiber optic microseismic systems typically exhibit attenuation or saturation effects at both ends, dead zones with no effective signal often appear at the beginning and end of the entire fiber optic link. The average energy sequence of each monitoring channel during the observation time is calculated using formula (3). , ,..., Based on a preset energy threshold, the signal dead zone range in the entire optical fiber link is identified, and the starting-end dead zone decision threshold T is preset according to the noise floor level. lead and end-point blind zone decision threshold T trail (e.g., the threshold T for the initial blind zone decision) lead =1, End-point blind zone decision threshold T trail =5); Starting from the beginning of the entire fiber optic link, when the average energy of the continuously monitored channel drops to the beginning-end blind zone decision threshold T... lead The following is recorded as the starting point. ; Starting from the end of the entire fiber optic link, the search proceeds in reverse order until the average energy of the continuously monitored channel drops to the end-blind zone decision threshold T. trail The following is recorded as the end point. Starting point To the finish line The part between ( ) is the service monitoring segment containing valid signals, while and This is a signal blind zone; it should be noted that the initial blind zone decision threshold T lead and end-point blind zone decision threshold T trail The specific value is determined by statistical analysis of the pre-collected static noise data.

[0021] Step 3: Within the service monitoring segment, select two segments of front and back reference fibers containing only background noise. The front reference fiber starts from the starting point. Take 100 m from the starting point, then refer to the fiber optic cable from the endpoint. Take 100 m forward from the starting point; The 20th-order Burg algorithm was used to establish autoregressive models for the background noise time-series signals of the front and back reference fibers, respectively, to obtain the noise power spectrum of the front reference fiber. and the noise power spectrum of the back reference fiber ; The Burg algorithm is a parameterized spectrum estimation algorithm based on the maximum entropy principle. It solves for the autoregressive model parameters by minimizing the forward and backward prediction errors, and can obtain a smooth and high-resolution power spectrum within a short data window. (Autoregressive model) Represented as: (4) In the formula: These are the parameters of the autoregressive model. The order of the parameters in the autoregressive model; This is the error term for modeling white noise; According to the autoregressive model The obtained noise power spectral density The expression is: (5) In the formula: The variance of the error term in the white noise modeling; Angular frequency; The noise power spectrum of the front reference fiber is obtained using the above autoregressive model. and the noise power spectrum of the back reference fiber Then, the weighting coefficient of the noise power spectrum of the front reference fiber is obtained according to the following formula (6). Weighting coefficients of the noise power spectrum of the back reference fiber : (6) In the formula: For the preceding reference fiber energy; For subsequent reference fiber energy; This leads to the noise amplitude spectrum. : (7).

[0022] Step 4: Perform spectral subtraction on the amplitude of each frame signal within the service monitoring segment, that is, subtract the noise amplitude spectrum from the original signal amplitude spectrum, thereby reducing the noise component. To avoid excessive noise reduction, a spectral subtraction factor α and a spectral base factor β are introduced for adjustment: the spectral subtraction factor α (≥1) is used to control the noise reduction intensity; the larger the spectral subtraction factor α, the more noise is subtracted; the spectral base factor β (0 < β < 1) is used to set the residual noise limit. The spectral base factor β is taken as a very small constant, for example, 0.01. The selection of the spectral subtraction factor α can be dynamically adjusted according to the signal-to-noise ratio (hence the term "adaptive" spectral subtraction): when the signal-to-noise ratio of a certain frame signal is <40, a larger α is appropriately chosen to more effectively eliminate noise; conversely, when the signal-to-noise ratio of a certain frame signal is ≥40, a smaller α is chosen to avoid excessive attenuation of the actual signal. For example, the values ​​of the spectral subtraction factor α and the spectral base factor β are: (8) in, For the first The first monitoring channel The signal-to-noise ratio of the frame signal; Obtain the raw amplitude spectrum of each frame signal within the service monitoring segment. and the original phase spectrum Then, the original amplitude spectrum The net signal amplitude spectrum is obtained by performing the following spectral subtraction operation. |: (9) (10) That is, if the original amplitude of a certain frame of signal is higher than its noise amplitude, then the net signal amplitude spectrum... | for the original amplitude and The difference between the amplitudes of two noise levels; if the original amplitude of a signal frame is lower than the noise amplitude, then the net signal amplitude spectrum... |for Noise amplitude; net signal amplitude spectrum Perform testing to avoid negative spectral amplitudes; The original phase spectrum and the net signal amplitude spectrum of each frame are reconstructed into a denoised complex spectrum. : (11) In the formula: The phase angle; Adaptive noise reduction processing is performed on all frame signals within the service monitoring segment.

[0023] Step 5: After spectral subtraction, convert the frequency domain back to the time domain and perform noise reduction on the complex spectrum of each frame signal in each monitoring channel. Perform the inverse Fourier transform to reconstruct the corresponding time-domain frame. : (12) Temporal frames The original signal's phase and event components are preserved, while noise amplitude is significantly reduced. The signal dead zone is left unprocessed, treated as a completely zero signal, and then processed and spliced ​​with the denoised complex spectrum.

[0024] In this method, we preserve the signal dead zone in its original state, perform noise reduction on the service monitoring segment using the aforementioned spectral subtraction method, and then combine the two. Performing an inverse Fourier transform on the entire fiber optic link yields the complete denoised time-domain signal matrix.

[0025] In summary, this method utilizes the front and rear reference fibers to estimate the noise amplitude spectrum in real time and performs adaptive spectral subtraction on the effective signal, significantly reducing background noise interference in the fiber optic microseismic system signal while achieving dynamic denoising. Experimental results show that, while maintaining the time-frequency structure of the vibration event without distortion, this method improves the average signal-to-noise ratio of the entire fiber optic link by approximately 8.92 dB. Visually, this results in a significant reduction in phase noise speckle that originally filled the time-distance map background, and makes the features of weak events more prominent relative to the background noise. Figure 2 The spectral subtraction algorithm based on front and back reference fibers provides a practical engineering approach for signal noise reduction in fiber optic microseismic systems under complex background noise conditions, and can provide higher signal-to-noise ratio and more stable and reliable input data for subsequent event recognition and precise positioning.

Claims

1. An adaptive dynamic noise reduction method for a distributed fiber optic microseismic system, characterized in that, The method is as follows: The entire optical fiber link is divided into segments of equal length, each segment serving as a monitoring channel. Vibration signals are collected synchronously in each monitoring channel to obtain its time series signal. After processing each time series signal in frames, Fourier transform is performed to calculate the amplitude spectrum of each frame signal in each monitoring channel. The entire fiber optic link is divided into service monitoring segments and signal blind spots; Within the service monitoring section, a front reference fiber and a back reference fiber are selected, and the noise power spectrum of the front reference fiber and the noise power spectrum of the back reference fiber are calculated, thereby obtaining the noise amplitude spectrum. The amplitude of each frame signal within the service monitoring segment is subjected to spectral subtraction processing. The noise reduction amplitude and the lower limit of residual noise are controlled by the spectral subtraction factor α and the spectral base factor β to obtain the net signal amplitude spectrum; then the noise-reduced complex spectrum is reconstructed. After splicing the denoised complex spectrum with the signal dead zone spectrum, an inverse Fourier transform is performed to restore it to the time domain, thus achieving signal denoising processing for the entire fiber optic link.

2. The adaptive dynamic noise reduction method for distributed fiber optic microseismic systems according to claim 1, characterized in that, The specific process for obtaining the amplitude spectrum of each frame signal is as follows: each time series signal is divided into frame signals with a fixed frame length and a 50% overlap rate, and a window function is applied to each frame signal. Then, a Fourier transform is performed to transform the time domain to the frequency domain, and the amplitude spectrum of each frame signal in each monitoring channel is calculated.

3. The adaptive dynamic noise reduction method for distributed fiber optic microseismic systems according to claim 2, characterized in that, The principle for dividing the service monitoring segment is as follows: the average energy of each monitoring channel during the observation time is obtained based on the amplitude spectrum of each frame signal in each monitoring channel, and the signal blind zone in the entire optical fiber link is identified based on the preset energy threshold, with the non-signal blind zone being the service monitoring segment.

4. The adaptive dynamic noise reduction method for distributed fiber optic microseismic systems according to claim 3, characterized in that, The principle for dividing the service monitoring segment is as follows: there is a blind zone at the beginning of the entire optical fiber link and a blind zone at the end; a blind zone decision threshold T is set at the beginning. lead and end-point blind zone decision threshold T trail ; Starting from the beginning of the entire fiber optic link, when the average energy of the continuously monitored channel drops to the beginning-end blind zone decision threshold T... lead The following time, the last monitoring channel is taken as the starting point. ; Starting from the end of the entire fiber optic link, the search proceeds in reverse order until the average energy of the continuously monitored channel drops to the end-blind zone decision threshold T. trail The last monitoring channel is considered the endpoint. Starting point To the finish line The section between these two points is the service monitoring segment containing valid signals.

5. The adaptive dynamic noise reduction method for a distributed fiber optic microseismic system according to claim 1, characterized in that, The front and rear reference fibers contain only background noise, and the front reference fiber starts from the point... Take 100 m from the starting point, then refer to the fiber optic cable from the endpoint. Take 100 m forward from the starting point.

6. The adaptive dynamic noise reduction method for a distributed fiber optic microseismic system according to claim 1, characterized in that, The specific process for obtaining the noise amplitude spectrum is as follows: An autoregressive model is established for the background noise time-series signals of the front and rear reference fibers using a 20th-order Burg algorithm, respectively, to obtain the noise power spectra of the front and rear reference fibers; the noise power spectra of the front and rear reference fibers are then linearly combined according to their respective weights to obtain the noise amplitude spectrum. The specific process is as follows: In the formula: Noise amplitude spectrum; The weighting coefficients for the noise power spectrum of the front reference fiber; The noise power spectrum of the preceding reference fiber; For the weighting coefficients of the noise power spectrum of the subsequent reference fiber; For subsequent reference fiber noise power spectrum; For the preceding reference fiber energy; For subsequent reference fiber energy and...

7. The adaptive dynamic noise reduction method for distributed fiber optic microseismic systems according to claim 1, characterized in that, The spectral subtraction factor α ≥ 1, 0 < spectral base factor β < 1; The spectral reduction factor α is dynamically adjusted with the signal-to-noise ratio; When the signal-to-noise ratio (SNR) of a certain frame signal is <40, the value of the spectral reduction factor α ranges from 1 to 45.5; when the SNR of a certain frame signal is ≥40, the value of the spectral reduction factor α is 1.

8. The adaptive dynamic noise reduction method for a distributed fiber optic microseismic system according to claim 1, characterized in that, The specific process for obtaining the reconstructed and denoised complex spectrum is as follows: Obtain the original amplitude spectrum and original phase spectrum of each frame signal within the service monitoring segment; perform spectral subtraction on the original amplitude spectrum; if the original amplitude of a certain frame signal is higher than its noise amplitude, then the net signal amplitude spectrum is the sum of the original amplitude and the noise amplitude. The difference between the noise amplitude and the original amplitude of a certain frame signal; if the original amplitude of a certain frame signal is lower than the noise amplitude, then the net signal amplitude spectrum is β times the noise amplitude, and the net signal amplitude spectrum is ≥0; the original phase spectrum and the net signal amplitude spectrum of each frame signal are reconstructed into a noise-reduced complex spectrum.

9. The adaptive dynamic noise reduction method for a distributed fiber optic microseismic system according to claim 8, characterized in that, The specific process of performing spectral subtraction on the original amplitude spectrum is as follows: In the formula: The net signal amplitude spectrum; The original amplitude spectrum; This is the noise amplitude spectrum.

10. The adaptive dynamic noise reduction method for a distributed fiber optic microseismic system according to claim 1, characterized in that, The specific process of splicing the noise-reduced complex spectrum with the signal blind zone spectrum is as follows: the original spectrum of the signal blind zone is preserved and spliced ​​with the noise-reduced complex spectrum.

Images