Method for extracting low-frequency components of signals from UWFBG distributed fiber optic acoustic wave sensing system

Abstract

This invention relates to a method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system, comprising: acquiring phase signals from each sensing grating segment; performing a first low-pass filter on the obtained signal to remove high-frequency components; downsampling the obtained signal to reduce the data volume; performing least-squares smoothing filter on the obtained signal; then performing a second low-pass filter; performing differential processing on the obtained signal to remove trend terms and filter out noise; and outputting the obtained low-frequency signal. This invention employs a double low-pass filtering method, which improves filtering performance, reduces the design difficulty of the filter, and ensures the real-time performance of low-frequency component extraction of the UWDAS signal. The downsampling method used between the two filters effectively reduces the impact of spectral aliasing on downsampling and also reduces the difficulty of designing a high-precision second low-pass filter. The differential processing method effectively removes trend terms from the low-frequency components of the UWDAS signal and filters out noise.

Description

Technical Field

[0001] This invention belongs to the field of signal processing, specifically relating to a method for extracting low-frequency components of signals from a UWFBG distributed optical fiber acoustic wave sensing system. Background Technology

[0002] The Distributed Optical Fiber Acoustic Sensing (DAS) system (UWDAS) based on the ultra-weak fiber Bragg grating (UWFBG) is a novel sensing technology that enables continuous distributed detection of vibration and sound fields. It has advantages such as wide range, high resolution, high sensitivity, wide bandwidth, and resistance to electromagnetic interference, and has broad application prospects in fields such as seismic exploration, pipeline monitoring, power cables, and perimeter security.

[0003] With a reflectivity of less than 0.1%, UWFBG (Ultra-Wide Fiber Bragg Grating) acts like a broadband filter or mirror within the fiber core. When incident pulsed light enters the fiber grating array, the reflected light from gratings at different positions meets, creating Fizeau interference. UWDAS (Ultra-Wide Fiber Bragg Assay) utilizes the Fizeau interference principle formed by the UWFBG grating structure to measure external physical quantities. Its working principle is relatively complex, requiring high-sensitivity, low-noise optical components to improve sensor performance. Despite this, UWDAS systems still exhibit significant noise, including electrical noise, laser noise, environmental noise, and polarization fading noise. Most system noise is concentrated in the operating frequency band below 10 Hz. In weak signal conditions, the low signal-to-noise ratio directly impacts the quality of the transmitted signal. Since the signal acquisition frequency of a single grating point reaches tens of kHz, and UWDAS has thousands of grating points, the amount of data processed in real-time is large, placing complex demands on signal processing. Considering system hardware and software resources, improving the signal-to-noise ratio and reducing the data volume of UWDAS signals has been a hot research topic in this field.

[0004] While UWDAS acquires signals with a wide frequency range, in practical engineering applications, useful signals are sometimes concentrated in the low-frequency range, such as in the detection of hydraulic fractures, microseismic monitoring, and underwater acoustic signal listening. In microseismic monitoring, different types of seismic events generate seismic waves with varying frequency ranges. For example, distant and intermediate earthquakes generate seismic waves with frequencies ranging from approximately 1 Hz to 20 Hz. Therefore, only the low-frequency components of the UWDAS signals need to be extracted for analysis, rather than the entire frequency band, which greatly improves the efficiency of microseismic monitoring. Thus, research on extracting the low-frequency components of UWDAS signals is particularly important.

[0005] UWDAS signals have significant background noise, often submerging the useful signal within it, resulting in a low signal-to-noise ratio and extremely large data volume, making them difficult to process in engineering applications. Existing common signal processing methods struggle to effectively extract the low-frequency components of UWDAS signals. While wavelet transform performs well in handling high-frequency noise, the threshold and threshold function are difficult to select, and there is a lack of suitable wavelet basis functions for decomposing UWDAS signals, leading to less than ideal processing of low-frequency components. Empirical Mode Decomposition (EMD) and its related algorithms suffer from mode aliasing, affecting the separation of effective and noisy signals and resulting in poor separation performance. For conventional filtering methods, such as downsampling after using an anti-aliasing filter, the filtering effect is less than ideal when the filter order is low, while the computational burden becomes excessive for high-order UWDAS signals due to the large data volume, reducing efficiency in practical engineering applications and hindering their application. Furthermore, these methods are unable to handle the noise present in the signal. Summary of the Invention

[0006] The purpose of this invention is to address the aforementioned problems by providing a method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system. The method involves performing a first low-pass filter on the UWDAS signal, followed by downsampling of the downsampled UWDAS signal. A second low-pass filter is then applied to extract the low-frequency components. Finally, a differential processing method is used to remove the trend term from the low-frequency components. This double low-pass filtering reduces the design complexity of the filter. Downsampling between the two low-pass filtering processes reduces the amount of data in the signal, improving the efficiency of subsequent data processing. Furthermore, least-squares smoothing filtering is incorporated to enhance signal smoothness, thereby improving the real-time performance and accuracy of low-frequency signal extraction.

[0007] The technical solution of this invention is a method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system, comprising the following steps: Step 1: Acquire the phase signals of each sensing grating segment using a distributed fiber optic acoustic wave sensing system with ultra-weak fiber Bragg gratings. Step 2: Design the first low-pass filter to filter out the high-frequency components of the signal obtained in Step 1. Step 3: Based on the Nyquist sampling theorem, select and determine the downsampling factor, and downsample the signal obtained in Step 2 to reduce the amount of data in the signal and improve the efficiency of subsequent data processing. Step 4: Use the least squares method to smooth the signal obtained in Step 3 to improve the smoothness of the signal; Step 4.1: Determine the size of the filter window. The size of the filter window determines the neighborhood range around each data point. Step 4.2: For each data point in the window, fit a polynomial function using the least squares method. Determine whether the polynomial is first-order, second-order, or multi-order based on the accuracy and real-time requirements of the output signal. Step 4.3: Use the polynomial obtained from the fitting in Step 4.2 to determine the value of the center point of the filter window; Step 4.4: Slide the filter window forward or backward, and repeat steps 4.2 and 4.3 until the smoothing filtering of the signal sequence obtained in step 3 is completed; Step 5: Design a second low-pass filter to perform low-pass filtering on the signal obtained in Step 4; Step 6: Perform differential processing on the signal obtained in Step 5 to remove the signal trend and filter out background noise; Step 7: Output the low-frequency signal obtained in step 6.

[0008] Furthermore, in step 2, the design parameters of the first low-pass filter include the filter order, sampling rate Fs1, passband frequency Fpass1, passband gain Apass1, and stopband attenuation Astop1.

[0009] Preferably, the filter order of the first low-pass filter is no greater than 5.

[0010] Furthermore, in step 3, based on the accuracy requirements of the output signal and the host computer configuration, the downsampling factor is selected while satisfying the Nyquist sampling theorem. When the signal frequency is less than 10Hz and the data volume is large, a larger downsampling factor is selected, with a downsampling factor greater than or equal to 64; when the signal frequency is greater than 50Hz and the data volume is small, a lower downsampling factor is selected, with a downsampling factor less than or equal to 64.

[0011] Preferably, in step 5, the design parameters of the second low-pass filter include sampling rate Fs2, passband frequency Fpass2, stopband frequency Fstop, passband gain Apass2, and stopband attenuation Astop2.

[0012] Preferably, the quotient obtained by dividing the original signal sampling frequency by the downsampling factor is taken as the value of Fs2, and the value of Fpass2 is equal to the frequency of the required low-frequency signal.

[0013] Preferably, the constraint relationship between the passband frequency Fpass2 and the stopband frequency Fstop is as follows: Fpass2 <Fstop≤ Fpass2+ 0.1。

[0014] Preferably, in step 6, the signal obtained in step 5 is differentially processed using a first-order backward difference method to remove the trend term of the signal. The calculation formula for the difference processing is as follows: x ( t ) = x ( t ) - x ( t – 1) in x ( t ), x ( t – 1) respectively represent t , t The signal at time -1 t Indicates the time.

[0015] Compared with the prior art, the beneficial effects of the present invention include: 1) This invention first performs a low-pass filter on the UWDAS signal to remove high-frequency information, then downsamples the low-pass filtered UWDAS signal, and performs a second low-pass filter on the downsampled UWDAS signal to effectively extract the low-frequency components of the UWDAS signal. The use of a double low-pass filter improves the filtering performance, reduces the design difficulty of the filter, and ensures the real-time extraction of the low-frequency components of the UWDAS signal. This invention also uses a differential processing method to effectively remove the trend term in the low-frequency components of the UWDAS signal and filter out the background noise.

[0016] 2) This invention employs downsampling after the first low-pass filtering and before the second low-pass filtering, which effectively reduces the impact of spectral aliasing on downsampling, greatly reduces the data volume of the UWDAS signal, reduces the design difficulty of the second low-pass filter, improves signal processing efficiency, and enhances the real-time performance of low-frequency component extraction and output of the UWDAS signal, which is beneficial for practical engineering applications.

[0017] 3) This invention greatly reduces the computation time and resource consumption of the signal processing process by flexibly selecting the downsampling factor.

[0018] 4) The method of the present invention effectively reduces the noise of the UWDAS signal and improves the signal-to-noise ratio and smoothness of the UWDAS signal. Attached Figure Description

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] Figure 1 This is a flowchart illustrating the method for extracting low-frequency components of UWDAS signals according to an embodiment of the present invention.

[0021] Figure 2 aThis is a time-domain diagram of the original UWDAS signal in an embodiment of the present invention.

[0022] Figure 2b This is a spectrum diagram of the original UWDAS signal in an embodiment of the present invention.

[0023] Figure 3 a This is a time-domain diagram of the UWDAS signal after the first low-pass filtering according to an embodiment of the present invention.

[0024] Figure 3b This is a spectrum diagram of the UWDAS signal after the first low-pass filtering according to an embodiment of the present invention.

[0025] Figure 4 a This is a time-domain diagram of the downsampled UWDAS signal according to an embodiment of the present invention.

[0026] Figure 4b This is a spectrum diagram of the downsampled UWDAS signal according to an embodiment of the present invention.

[0027] Figure 5 a This is a time-domain diagram of the UWDAS signal after least-squares smoothing filtering according to an embodiment of the present invention.

[0028] Figure 5b This is a spectrum diagram of the UWDAS signal after least-squares smoothing filtering according to an embodiment of the present invention.

[0029] Figure 6 a This is a time-domain diagram of the UWDAS signal after the second low-pass filtering according to an embodiment of the present invention.

[0030] Figure 6b This is a spectrum diagram of the UWDAS signal after the second low-pass filtering according to an embodiment of the present invention.

[0031] Figure 7 a This is a time-domain diagram of the UWDAS signal after differential processing according to an embodiment of the present invention.

[0032] Figure 7b This is a spectrum diagram of the UWDAS signal after differential processing according to an embodiment of the present invention. Detailed Implementation

[0033] In this embodiment, the frequency of a signal below 5 Hz is extracted from the UWDAS measured signal at a vibration table frequency of 4 Hz.

[0034] like Figure 1 As shown, the method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system includes the following steps: Step 1: Acquire the UWDAS signal from the vibration table at a frequency of 4 Hz. The UWDAS output is the phase, and the sampling rate is 10000. The time domain and spectrum of the original signal are as follows: Figure 2a , 2b As shown, the UWDAS signal has a sampling time of approximately 2 seconds, with 20,480 sampling points. The data volume is relatively small, the useful signal is a 4Hz sine wave, and the signal contains multiple frequency components, including signal components above 5Hz, resulting in a low signal-to-noise ratio.

[0035] Step 2: Perform a first low-pass filter on the acquired UWDAS signal. Set the sampling rate Fs1 to 10000, passband frequency Fpass1 to 5 Hz, passband gain Apass1 to 1 dB, stopband attenuation Astop1 to 100 dB, and filter order to 5. Design a 5th-order IIR elliptic low-pass filter to remove the high-frequency components of the signal. The time domain and spectrum of the UWDAS signal after the first low-pass filter are shown below. Figure 3a , 3b As shown, most of the high-frequency information above 5Hz in the UWDAS signal was removed, reducing the impact of spectral aliasing that might occur during subsequent downsampling, while perfectly preserving the low-frequency information of the signal.

[0036] Step 3: Downsample the low-pass filtered signal. Select a downsampling factor 'a' of 32 to reduce the data volume. The time domain and spectrum of the downsampled UWDAS signal are shown below. Figure 4a , 4b As shown, the number of UWDAS signal sampling points has been reduced from the original 20480 to 640, which greatly improves the efficiency of data use in engineering applications and reduces the design difficulty of the low-pass filter during the second low-pass filtering, enabling it to meet the work efficiency requirements of practical engineering applications while ensuring good filtering effect.

[0037] Step 4: Perform least-squares smoothing filtering on the downsampled signal. Set the least-squares smoothing filter window length L to 5 and the polynomial fitting order N to 3. This removes some noise from the signal and improves its smoothness. The time domain and spectrum of the UWDAS signal after least-squares smoothing filtering are shown below. Figure 5a , 5b As shown, random noise in the UWDAS signal is removed, improving the smoothness of the signal.

[0038] Step 5: Perform a second low-pass filter on the least-squares smoothed signal. The filter order is not considered; design the low-pass filter as an IIR elliptic filter.

[0039] In this embodiment, the sampling rate Fs2 of the second low-pass filter is 312.5, the passband frequency Fpass2 of the second low-pass filter is 5 Hz, the stopband frequency Fstop of the second low-pass filter is 5.1 Hz, the passband gain Apass2 of the second low-pass filter is 1 dB, and the stopband attenuation Astop2 of the second low-pass filter is 60 dB, thus filtering out the high-frequency part of the signal.

[0040] The time domain and spectrum of the UWDAS signal after the second low-pass filtering are as follows: Figure 6a , 6b As shown, the second low-pass filter removes the portion of the UWDAS signal with frequencies higher than 5 Hz, thus enabling the extraction of the low-frequency components of the UWDAS signal.

[0041] Step 6: Perform differential processing on the signal after the second filtering. This removes the trend term and filters out some background noise. The time domain and spectrum of the differentially processed UWDAS signal are shown below. Figure 7a , 7b As shown, differential processing effectively removes the trend term from the signal.

[0042] Step 7: Data output.

[0043] In this embodiment, a first low-pass filter removes some high-frequency information from the UWDAS signal and reduces the impact of spectral aliasing on downsampling, enabling downsampling of the UWDAS signal. A second downsampling reduces the data volume of the UWDAS signal. Then, a least-squares smoothing filter removes random noise and improves signal smoothness. A second low-pass filter is then applied. Since the UWDAS signal has undergone the previous processing steps, the filter design is less complex, improving filter performance while reducing its impact on the efficiency of practical engineering applications. This allows for the extraction of low-frequency components from the UWDAS signal. Finally, differential processing removes signal trends.

[0044] During the first low-pass filtering, the signal data volume is relatively large. To reduce computational complexity for practical engineering purposes, the filter order is controlled, leading to reduced filter accuracy and poor filtering performance. Figure 3a , 3b As shown, the designed passband frequency is 5 Hz, but it can be seen that there are still signal frequencies above 5 Hz. Therefore, a second low-pass filter is needed. However, due to the data volume, a first downsampling is chosen. Since a low-pass filter has already been applied, the impact of spectral aliasing on downsampling is reduced, and the range of downsampling factor selection is widened. However, this will lead to a decrease in signal smoothness, such as... Figure 4a , 4b As shown. Considering this, least squares smoothing filtering is used to improve signal smoothness. Due to downsampling, the data volume is small, so a higher polynomial fitting order can be used to improve fitting accuracy, enhance signal smoothness, and remove random noise, spike noise, etc., thus improving the signal-to-noise ratio. Figure 5a , 5b As shown.

[0045] Then a second low-pass filter is performed, such as... Figure 6a , 6b As shown, the frequency range above 5Hz is basically filtered out, completing the extraction of low-frequency signals, and the signal trend is removed through the final differential processing. Figure 7a , 7b As shown.

[0046] Compared to traditional low-pass filtering followed by downsampling, this invention addresses the issue that high-precision low-pass filters are computationally expensive and inefficient when dealing with large datasets. Therefore, it employs two low-pass filters to extract the low-frequency components of the signal. Furthermore, it addresses the reduced signal smoothness resulting from downsampling. Compared to median or mean filtering, the least-squares smoothing filter offers higher accuracy and simultaneously reduces signal noise. Finally, differential detrending is used, which is simpler to operate than EMD and unaffected by mode aliasing. Traditional fitting detrending methods, on the other hand, struggle to determine the fitting order, hindering engineering applications.

[0047] This invention achieves low-frequency signal extraction through a combination of several filtering methods. In the entire workflow, the various filtering methods complement each other, reducing the impact of a single filter on the signal and improving their respective working efficiency. The principle is simple, easy to implement, highly practical, and effective.

Claims

1. A method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system, characterized in that, Includes the following steps: Step 1: Acquire the phase signals of each sensing grating segment using a distributed fiber optic acoustic wave sensing system with ultra-weak fiber Bragg gratings. Step 2: Design the first low-pass filter to perform low-pass filtering on the signal obtained in Step 1, and filter out the high-frequency components of the signal; Step 3: Based on the Nyquist sampling theorem, determine the downsampling factor, and downsample the signal obtained in Step 2 to reduce the amount of data; Step 4: Use the least squares method to smooth the signal obtained in Step 3 to improve the smoothness of the signal; Step 4.1: Determine the size of the filter window. The size of the filter window determines the neighborhood range around each data point. Step 4.2: For each data point in the window, fit a polynomial function using the least squares method. Determine whether the polynomial is first-order, second-order, or multi-order based on the accuracy and real-time requirements of the output signal. Step 4.3: Use the polynomial obtained from the fitting in Step 4.2 to determine the value of the center point of the filter window; Step 4.4: Slide the filter window forward or backward, and repeat steps 4.2 and 4.3 until the smoothing filtering of the signal sequence obtained in step 3 is completed; Step 5: Design a second low-pass filter to perform low-pass filtering on the signal obtained in Step 4; Step 6: Perform differential processing on the signal obtained in Step 5 to remove the trend term and filter out background noise; Step 7: Output the low-frequency signal obtained in step 6.

2. The method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system according to claim 1, characterized in that, In step 2, the design parameters of the first low-pass filter include the filter order of the first low-pass filter, the sampling rate Fs1 of the first low-pass filter, the passband frequency Fpass1 of the first low-pass filter, the passband gain Apass1 of the first low-pass filter, and the stopband attenuation Astop1 of the first low-pass filter.

3. The method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system according to claim 2, characterized in that, The first low-pass filter has a filter order no greater than 5.

4. The method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system according to claim 2 or 3, characterized in that, In step 3, the downsampling factor is selected based on the accuracy requirements of the output signal and the configuration of the host computer.

5. The method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system according to claim 4, characterized in that, In step 5, the design parameters of the second low-pass filter include the sampling rate Fs2, the passband frequency Fpass2, the stopband frequency Fstop, the passband gain Apass2, and the stopband attenuation Astop2. The value of Fpass2 is equal to the frequency of the desired low-frequency signal.

6. The method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system according to claim 5, characterized in that, The value of Astop2 ranges from 40 to 60 dB.

7. The method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system according to claim 5 or 6, characterized in that, The constraint relationship between the passband frequency Fpass2 and the stopband frequency Fstop of the second low-pass filter is as follows: Fpass2 <Fstop ≤ Fpass2 + 0.1。 8. The method for extracting low-frequency components of signals from a UWFBG distributed fiber optic acoustic wave sensing system according to claim 1, 2, 3, 5, or 6, characterized in that, In step 6, the first-order backward difference method is used to perform differential processing on the signal obtained in step 5 to remove the trend term of the signal.

Images