DAS signal noise floor cancellation method, system, and device based on statistical analysis

Abstract

The application relates to the technical field of optical fiber sensing signal processing, in particular to a DAS signal bottom noise elimination method, system and device based on statistical analysis, which comprises the following steps: initializing system parameters, and loading a DAS application field data matrix; slicing and performing frequency domain transformation on DAS original data; dynamically determining a peak threshold value, and adaptively searching for a peak according to the peak threshold value; performing descending order sorting on local maximum values screened out through identification, and selecting a frequency corresponding to the local maximum values as a main frequency candidate; performing clustering analysis on the selected main frequency candidate; calculating the cumulative proportion of each main frequency in total statistics, selecting a center frequency, forming a band-stop frequency band by different center frequencies, and composing a multi-frequency band-stop of the monitoring unit; adding the multi-frequency band-stop to a statistical filtering frequency band set of all point positions, saving the statistical filtering frequency band set of all point positions as a statistical model, and eliminating bottom noise of a DAS original signal based on the statistical model.

Description

Technical Field

[0001] This invention relates to the field of fiber optic sensing signal processing technology, specifically to a method, system, and device for eliminating the noise floor of DAS signals based on statistical analysis. Background Technology

[0002] Mechanical equipment operation, fluid transportation, and geological activities can all cause vibrations or sound signals of varying degrees. While some vibrations are characteristic of normal operation, abnormal vibrations often imply potential risks. For example, in oil and gas transportation systems, buried pipelines may generate abnormal vibration signals when subjected to third-party mechanical construction or external force damage. If these are not monitored and addressed in a timely manner, they can easily lead to pipeline leaks or even major safety accidents. In coal mine transportation, when belt conveyor rollers experience wear, jamming, or other abnormalities, their vibration patterns will also change significantly. If this is not detected early, it may cause belt breakage, production stoppage, or secondary disasters.

[0003] Traditional vibration or acoustic wave monitoring methods mostly rely on point-based detection methods such as accelerometers and acoustic emission sensors. While these methods offer a certain level of accuracy in localized monitoring, they generally suffer from problems such as discrete monitoring points, limited detection range, and insufficient data continuity, making it difficult to achieve real-time, continuous, and blind-spot-free monitoring of large-scale targets. Distributed Acoustic Sensing (DAS) technology is gradually emerging as a new monitoring method. This technology utilizes existing fiber optic communication networks, transforming ordinary optical fibers into distributed sensors to achieve real-time sensing of vibration and acoustic wave signals along the pipeline. Compared to traditional point sensors, DAS offers advantages such as being passive, long-distance, providing full coverage, and being continuously distributed, making it particularly suitable for applications such as pipeline safety monitoring and early warning of transmission equipment failures.

[0004] However, in practical applications, DAS is inevitably subject to interference from multiple sources, such as environmental noise, system noise, and scattering noise. These background noise signals can mask or weaken abnormal vibration characteristics, reducing the accuracy of fault or risk event detection. Currently common filtering methods, such as wavelet transform and deep learning, typically require a large amount of labeled data, model training, and high computational resources, making them difficult to adapt to the real-time processing needs of different engineering scenarios. Summary of the Invention

[0005] To address the aforementioned problems, the first aspect of this invention provides a method for eliminating the noise floor of DAS signals based on statistical analysis, comprising:

[0006] S1. Initialize system parameters and load the DAS application field data matrix, which includes the total number of monitoring units and the total number of frames;

[0007] S2. Slice and transform the DAS raw data of a single monitoring unit's calculation frame to extract positive frequency spectrum data, and obtain the envelope curve by fitting local maxima.

[0008] S3. Calculate the local mean and noise standard deviation of the envelope curve within the preset window, dynamically determine the peak threshold, perform adaptive peak finding based on the peak threshold, and identify the local maximum value of the envelope curve.

[0009] S4. Sort the identified local maximum values ​​in descending order to obtain an indexed sequence. Select the frequencies corresponding to the first N local maximum values ​​in the indexed sequence as candidates for the dominant frequency.

[0010] S5. Perform cluster analysis on the selected candidate frequencies, count the occurrence frequency and maximum amplitude of each frequency group, and obtain the cluster result set.

[0011] S6. Calculate the clustering results set of DAS raw data of all calculation frames of the monitoring unit, calculate the cumulative proportion of each main frequency in the total statistics, select the main frequency whose cumulative proportion reaches the preset threshold as the center frequency, and form a band-stop frequency band based on the center frequency to expand the bandwidth; form a multi-frequency band-stop of the monitoring unit by band-stop frequency bands formed by different center frequencies.

[0012] S7. Add the multi-frequency bandstop of a single monitoring unit to the statistical filtering frequency band set of all points; determine whether the multi-frequency bandstop of all monitoring units has been obtained. If not, return to S2 and execute the loop; if completed, save the statistical filtering frequency band set of all points as a statistical model, and perform noise reduction on the DAS original signal based on the statistical model.

[0013] The adaptive peak finding based on the peak threshold described in S3, which identifies the local maximum value of the envelope curve, is specifically performed as follows:

[0014] Based on the preset window, the local mean and noise standard deviation of the data within the envelope curve window are calculated. The peak threshold of this window is then determined using the local mean and noise standard deviation.

[0015]

[0016] In the formula, T(i) represents the peak threshold; μ L (i) It is a local mean; σ L (i) The noise standard deviation is represented by α, which represents the threshold coefficient.

[0017] Adaptive peak finding is performed on the data points within the envelope curve window based on the peak threshold. When a data point satisfies x(i)>x(i−1), x(i)>x(i+1), and x(i)>T(i), it is determined to be a local maximum and its index i is extracted.

[0018] The clustering analysis of the selected candidate dominant frequencies described in S5, which involves counting the frequency of occurrence and the maximum amplitude of each frequency group to obtain a clustering result set, is specifically performed as follows:

[0019] The frequency range is determined based on the sampling rate, and the frequency range is divided into multiple intervals according to the preset step size;

[0020] The selected candidate frequencies are mapped to the corresponding intervals using a grouping function;

[0021] The number of samples in each group interval and the maximum amplitude in that interval are counted to form a cluster result set, which includes the cluster name, the count value of the number of samples, and the maximum amplitude value.

[0022] The formula for the grouping function is:

[0023] ,

[0024] In the formula, g(·) represents the grouping function; data i Indicates the candidate value for the main frequency; s Δ is the starting frequency of the interval; Δ is the preset step size.

[0025] The system parameters include sampling rate (sample_rate), moving step, calculation frame length (window_length), number of main frequency bands (N_PCA_fre), frequency clustering step length (N_PCA_fre_group_length), and moving times.

[0026] The system parameters also include the cluster maximum value statistics max_values, the cluster count statistics counter, and the total_fre_expand_range_fft_max of the total number of points for statistical filtering.

[0027] The method described in S3 for obtaining the envelope curve by fitting local maxima involves traversing the positive frequency spectrum to filter local maxima, saving the peak value and abscissa of the local maxima to the x-sequence and y-sequence respectively, and performing linear fitting on adjacent data points based on the x-sequence and y-sequence to generate the envelope curve.

[0028] The second aspect provides a statistical analysis-based DAS signal noise reduction system for implementing the statistical analysis-based DAS signal noise reduction method described above, comprising:

[0029] The initialization module is used to initialize system parameters and load the DAS application field data matrix, which includes the total number of monitoring units and the total number of frames.

[0030] The frequency domain transformation module is used to slice and transform the raw DAS data of a single monitoring unit in one calculation frame, extract positive frequency spectrum data, and obtain the envelope curve by fitting local maxima.

[0031] The adaptive peak finding module is used to calculate the local mean and noise standard deviation of the envelope curve within a preset window, dynamically determine the peak threshold, and perform adaptive peak finding based on the peak threshold to identify the local maximum value of the envelope curve.

[0032] The main frequency acquisition module is used to sort the identified local maximum values ​​in descending order to obtain an indexed sequence, and select the frequencies corresponding to the first N local maximum values ​​in the indexed sequence as candidates for the main frequency.

[0033] The dominant frequency clustering module is used to perform cluster analysis on the selected dominant frequency candidates, count the occurrence frequency and maximum amplitude of each frequency, and obtain a set of clustering results.

[0034] The frequency band generation module is used to statistically analyze the clustering results of the DAS raw data of all calculation frames of the monitoring unit, calculate the cumulative proportion of each main frequency in the total statistics, select the main frequency whose cumulative proportion reaches a preset threshold as the center frequency, and expand the bandwidth based on the center frequency to form a band-stop frequency band; and combine the band-stop frequency bands formed by different center frequencies to form the multi-frequency band-stop of the monitoring unit.

[0035] The statistical model module is used to add the multi-frequency bandstop of a single monitoring unit to the statistical filter frequency band set of all points, save the statistical filter frequency band set of all points as a statistical model, and perform noise reduction on the DAS raw signal based on the statistical model.

[0036] The statistical model module also includes determining whether the multi-frequency bandstops of all monitoring units have been added to the statistical filter frequency band set of all points. If not, it returns to the frequency domain transformation module to update the monitoring units.

[0037] The third aspect provides a DAS signal noise reduction device based on statistical analysis, including a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the DAS signal noise reduction method based on statistical analysis as described above.

[0038] Beneficial Effects: This invention provides a method, system, and device for DAS signal noise reduction based on statistical analysis. It achieves DAS signal noise reduction through statistical analysis and frequency domain filtering, directly eliminating DAS noise from different projects without requiring pre-training or large-scale labeled datasets, thus adapting to the noise reduction needs of acoustic signal monitoring in various application scenarios. Through steps such as dominant frequency candidate clustering analysis and interval partitioning mapping, it automatically captures the noise frequency band distribution characteristics of different monitoring units and dynamically generates band-stop filters suitable for each location. Simultaneously, it only eliminates high-frequency noise frequency bands with a cumulative proportion reaching a preset threshold (e.g., 90%), retaining low-proportion abnormal vibration characteristic signals. This effectively avoids the accidental deletion of key event information by traditional filtering methods, ensuring the accuracy of risk event detection. Attached Figure Description

[0039] The solutions and advantages of this application will become clear to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of the invention.

[0040] In the attached diagram:

[0041] Figure 1 This is a flowchart of the DAS signal noise reduction method of the present invention;

[0042] Figure 2 Flowchart for establishing the statistical model for this invention

[0043] Figure 3 The diagram shows the vibration phase curve of the original DAS signal.

[0044] Figure 4 This is a Fourier spectrum curve;

[0045] Figure 5 The vibration phase curve of the filtered DAS signal;

[0046] Figure 6 This is the FFT spectrum curve of the filtered DAS signal. Detailed Implementation

[0047] Exemplary embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings.

[0048] Example 1

[0049] See Figures 1-2 This embodiment provides a method for eliminating the noise floor of DAS signals based on statistical analysis. The specific implementation steps of the method are as follows:

[0050] S1. Initialize system parameters and load the DAS application field data matrix;

[0051] Load the DAS application field data matrix (total_frame, Total_unit_num). Total_frame represents the total number of frames, and total_frame = sample_rate (sampling rate) × t (monitoring time). Total_unit_num represents the total number of units, and total_unit_num = total length of optical cable / spatial resolution.

[0052] The system parameters are initialized. The system is the processor system that acquires the raw DAS signal and processes the signal data. The system parameters include the sampling rate (sample_rate), the moving step size (move_step), the calculation frame length (window_length), the number of main frequency bands (N_PCA_fre), the frequency clustering step size (N_PCA_fre_group_length), and the number of moving times (move_times).

[0053] At the same time, initialize the cluster maximum value statistics max_values, the cluster count statistics counter, and the total point statistical filter frequency band set to total_fre_expand_range_fft_max.

[0054] This embodiment uses the monitoring scenario of oil and gas pipeline as an example to explain the parameter initialization. The total length of the optical cable in the system is 10km, and the spatial resolution is 10m. Therefore, Total_unit_num=1000; the monitoring time is 10s, the sampling rate is 5000Hz, and total_frame=50000.

[0055] The other initialization parameters are as follows: move_step=2048, window_length=4096, number of main frequency bands N_PCA_fre=10, frequency cluster length N_PCA_fre_group_length = 20; move_times=int ((50000-4096) / 2048)≈22.

[0056] S2. Slice and transform the DAS raw data of a single monitoring unit's calculation frame to extract positive frequency spectrum data, and obtain the envelope curve by fitting local maxima.

[0057] For the i-th computation frame of any unit, the original data DAS_ori is sliced ​​based on the sliding step size and the computation frame length to obtain a single-frame one-dimensional time-domain array singe_frame, singe_frame=DAS_ori[d1,d2,d3....dwindow_length].

[0058] A discrete Fourier transform is performed on a one-dimensional time-domain array of a single frame to obtain a Fourier spectrum. The Fourier spectrum of the positive frequency portion is extracted as data for subsequent analysis. For example... Figure 3 , Figure 4 As shown, Figure 3 This is the original DAS signal. Figure 4 The Fourier spectrum corresponding to the original data has been locally amplified in terms of the amplitude of the low-frequency and high-frequency components.

[0059] The envelope curve is obtained by fitting local maxima based on positive frequency spectrum data. Specifically, the positive frequency spectrum is traversed to filter local maxima, and the peak value and abscissa of the local maxima are saved to the x-sequence and y-sequence respectively. Based on the x-sequence and y-sequence, linear fitting is performed on adjacent data points to generate the envelope curve.

[0060] S3. Calculate the local mean and noise standard deviation of the envelope curve within the preset window, dynamically determine the peak threshold, perform adaptive peak finding based on the peak threshold, and identify the local maximum value of the envelope curve.

[0061] S301. Calculate the local mean and noise standard deviation of the data within the envelope curve window based on the preset window, and determine the peak threshold of the window using the local mean and noise standard deviation:

[0062]

[0063] In the formula, T(i) represents the peak threshold; μ L (i) It is a local mean; σ L (i) α represents the noise standard deviation; α represents the threshold coefficient, which is set to 3 through testing in this embodiment.

[0064] S302. Based on the peak threshold, adaptively find the peak of the data points within the envelope curve window. When the data points satisfy x(i)>x(i−1), x(i)>x(i+1) and x(i)>T(i), they are determined to be local maxima and their index i is extracted.

[0065] S4. Sort the identified local maximum values ​​in descending order to obtain an indexed sequence. Select the frequencies corresponding to the first N local maximum values ​​in the indexed sequence as candidates for the dominant frequency.

[0066] S5. Perform cluster analysis on the selected candidate dominant frequencies, count the occurrence frequency and maximum amplitude of each frequency group, and obtain the clustering result set. The specific operation is as follows:

[0067] S501. Determine the frequency interval [s, e) based on the sampling rate, where s is the starting frequency of the interval and e is the ending frequency of the interval; and divide the frequency interval [s, e) into multiple intervals G according to the preset step size Δ; in this embodiment, the preset step size Δ corresponds to the frequency clustering step size N_PCA_fre_group_length.

[0068] S502. Map the selected candidate frequencies to the corresponding intervals using a grouping function;

[0069] The formula for the grouping function is:

[0070]

[0071] In the formula, g(·) represents the grouping function; data i Indicates the candidate value for the main frequency; s Δ is the starting frequency of the interval; Δ is the preset step size.

[0072] S503. Count the number of samples in each group interval and the maximum amplitude in that interval to form a clustering result set. The clustering result set includes the cluster name, the count value of the number of samples, and the maximum amplitude value.

[0073] S6. Calculate the clustering results set of DAS raw data of all calculation frames of the monitoring unit, calculate the cumulative proportion of each main frequency in the total statistics, select the main frequency whose cumulative proportion reaches the preset threshold as the center frequency, and form a band-stop frequency band based on the center frequency to expand the bandwidth; form a multi-frequency band-stop of the monitoring unit by band-stop frequency bands formed by different center frequencies.

[0074] By selecting the main frequency whose cumulative proportion reaches a preset threshold (i.e., 90%) as the center frequency, and extending the bandwidth of the center frequency according to the frequency cluster length, a multi-frequency bandstop is formed. The specific process is as follows:

[0075] For each selected center frequency f c ∈F, if it satisfies the constraints in a certain frequency interval set: s k =auto_start,e k =auto_end,Δ 3dB,k = N_PCA_fre_group_length; based on the 3dB bandwidth Δ 3dB,k Generate the symmetrical extended range of the center frequency and the corresponding upper and lower limits of the frequency. f + , f - ,

[0076]

[0077] The bandstops formed by different center frequencies are grouped into a set ε, which is the multi-frequency bandstop of the monitoring unit. .

[0078] S7. Add the multi-frequency bandstop of a single monitoring unit to the statistical filtering frequency band set of all points; determine whether the multi-frequency bandstop of all monitoring units has been obtained. If not, return to S2 and execute the loop; if completed, save the statistical filtering frequency band set of all points as a statistical model, and perform noise reduction on the DAS original signal based on the statistical model.

[0079] When all the multi-frequency bandstops of all monitoring units are saved and added to the filter frequency band set of all points, the set is saved as a statistical model. The statistical model is the noise floor filter parameter library integrated after statistical analysis. For the subsequent input DAS raw signal, the raw signal is filtered according to the noise floor filter of the corresponding monitoring unit to achieve noise floor elimination.

[0080] like Figure 5 , Figure 6 As shown, Figure 5 The image shows the vibration phase curve of the filtered DAS signal; an FFT transformation is then performed on the filtered data. Figure 6 The FFT spectrum curve of the filtered DAS signal shows that there are multiple frequency bands with zero values, indicating that frequency noise reduction based on statistical analysis has been performed.

[0081] In addition, a DAS signal noise reduction system based on statistical analysis is also provided, including:

[0082] The initialization module is used to initialize system parameters and load the DAS application field data matrix, which includes the total number of monitoring units and the total number of frames.

[0083] The frequency domain transformation module is used to slice and transform the raw DAS data of a single monitoring unit in one calculation frame, extract positive frequency spectrum data, and obtain the envelope curve by fitting local maxima.

[0084] The adaptive peak finding module is used to calculate the local mean and noise standard deviation of the envelope curve within a preset window, dynamically determine the peak threshold, and perform adaptive peak finding based on the peak threshold to identify the local maximum value of the envelope curve.

[0085] The main frequency acquisition module is used to sort the identified local maximum values ​​in descending order to obtain an indexed sequence, and select the frequencies corresponding to the first N local maximum values ​​in the indexed sequence as candidates for the main frequency.

[0086] The dominant frequency clustering module is used to perform cluster analysis on the selected dominant frequency candidates, count the occurrence frequency and maximum amplitude of each frequency, and obtain a set of clustering results.

[0087] The frequency band generation module is used to statistically analyze the clustering results of the DAS raw data of all calculation frames of the monitoring unit, calculate the cumulative proportion of each main frequency in the total statistics, select the main frequency whose cumulative proportion reaches a preset threshold as the center frequency, and expand the bandwidth based on the center frequency to form a band-stop frequency band; and combine the band-stop frequency bands formed by different center frequencies to form the multi-frequency band-stop of the monitoring unit.

[0088] The statistical model module is used to add the multi-frequency bandstop of a single monitoring unit to the statistical filter frequency band set of all points, save the statistical filter frequency band set of all points as a statistical model, and perform noise reduction on the DAS raw signal based on the statistical model.

[0089] The statistical model module also includes determining whether the multi-frequency bandstops of all monitoring units have been added to the statistical filter frequency band set of all points. If not, it returns to the frequency domain transformation module to update the monitoring units.

[0090] Finally, this embodiment also provides a DAS signal noise reduction device based on statistical analysis, including a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the DAS signal noise reduction method based on statistical analysis as described above.

Claims

1. A statistical analysis based DAS signal floor cancellation method, characterized in that, The method comprises the following steps: S1, initializing system parameters, loading a DAS application field data matrix, the data matrix comprising a total number of monitoring units and a total number of frames; S2, slicing and frequency domain transforming DAS original data of a calculation frame of a single monitoring unit, extracting positive frequency spectrum data, and obtaining an envelope curve through local maximum value fitting; S3, calculating a local mean value and a noise standard deviation of the envelope curve in a preset window, dynamically determining a peak threshold, adaptively searching for a peak according to the peak threshold, and identifying a local maximum value of the envelope curve; S4, sorting the identified and screened local maximum values in descending order to obtain an index arrangement sequence, and selecting frequencies corresponding to the first N local maximum values in the index arrangement sequence as main frequency candidates; S5, performing clustering analysis on the selected main frequency candidates, counting the occurrence times and maximum amplitudes of various frequencies, and obtaining a clustering result set; S6, counting the clustering result sets of the DAS original data of all calculation frames of the monitoring unit, calculating the cumulative proportion of each main frequency in the total statistics, selecting a main frequency with a cumulative proportion reaching a preset threshold as a center frequency, and forming a band-stop frequency band based on the center frequency to expand the bandwidth; and grouping the band-stop frequency bands formed by different center frequencies to form a multi-frequency band-stop of the monitoring unit; S7, adding the multi-frequency band-stop of the single monitoring unit to a set of all point position statistical filtering frequency bands; determining whether the multi-frequency band-stops of all monitoring units are obtained, and if not, returning to S2 for cyclic execution; If the process is completed, the set of all point position statistical filtering frequency bands is saved as a statistical model, and the DAS original signal is subjected to bottom noise elimination based on the statistical model.

2. The statistical analysis based DAS signal floor cancellation method according to claim 1, characterized in that, In S3, the adaptive peak searching according to the peak threshold to identify the local maximum value of the envelope curve is specifically performed as follows: The local mean value and the noise standard deviation of the data in the envelope curve window are calculated according to the preset window, and the peak threshold of the window is determined by the local mean value and the noise standard deviation as follows: ， where T(i) represents a peak threshold value; According to the peak threshold, the data points in the envelope curve window are adaptively searched for peaks, and when the data points satisfy x(i)>x(i−1), x(i)>x(i+1) and x(i)>T(i), it is determined that the data points are local maximum values and their indexes i are extracted. L (i) is a local mean; In S5, the clustering analysis on the selected main frequency candidates, the counting of the occurrence times and the maximum amplitudes of various frequencies, and the obtaining of the clustering result set are specifically performed as follows: L (i) is a noise standard deviation; and a represents a threshold coefficient. The frequency interval is determined based on the sampling rate, and the frequency interval is divided into multiple intervals according to a preset step length; 3. The statistical analysis based DAS signal floor cancellation method of claim 1, wherein, The selected main frequency candidates are mapped to the corresponding intervals through a grouping function; The sample number in each grouping interval and the maximum amplitude in the interval are counted to form a clustering result set, and the clustering result set comprises a clustering name, a count value of a sample number and a maximum amplitude. The grouping function formula is as follows: data 4. The statistical analysis based DAS signal floor cancellation method of claim 3, wherein, Δ is a preset step length. ， In the formula, g(·) represents a grouping function. The system parameters comprise a sampling rate sample_rate, a sliding step length move_step, a calculation frame length window_length, a main frequency band number N_PCA_fre, a frequency clustering step length N_PCA_fre_group_length, and a sliding number move_times. i represents a main frequency candidate value; s is an interval start frequency; ​ 5. The statistical analysis based DAS signal floor cancellation method of claim 1, wherein, ​ 6. The statistical analysis based DAS signal floor cancellation method of claim 5, wherein, The system parameters further include a cluster maximum value statistic max_values, a cluster count statistic counter, and a total point site statistic filter frequency band set total_fre_expand_range_fft_max.

7. The statistical analysis based DAS signal floor cancellation method of claim 1, wherein, In S2, the envelope curve is obtained by fitting the local maximum value, specifically, local maximum values are screened by traversing the positive frequency spectrum, and the peak values and abscissa of the local maximum values are saved to x sequence and y sequence respectively, linear fitting is performed on adjacent data points based on x sequence and y sequence, and the envelope curve is generated.

8. A statistical analysis based DAS signal floor cancellation system for implementing the statistical analysis based DAS signal floor cancellation method of claim 1, characterized by, It comprises: An initialization module for initializing system parameters, loading a DAS application field data matrix, the data matrix including a total number of monitoring units and a total number of frames; A frequency domain transformation module for slicing and frequency domain transforming DAS original data of a single monitoring unit and a calculation frame, extracting positive frequency spectrum data, and obtaining an envelope curve by fitting a local maximum value; An adaptive peak searching module for calculating a local mean value and a noise standard deviation of the envelope curve within a preset window, dynamically determining a peak threshold, adaptively searching for a peak according to the peak threshold, and identifying a local maximum value of the envelope curve; A main frequency acquisition module for sorting the identified and screened local maximum values in descending order to obtain an index arrangement sequence, and selecting frequencies corresponding to the first N local maximum values in the index arrangement sequence as main frequency candidates; A main frequency clustering module for performing clustering analysis on the selected main frequency candidates, counting the occurrence times and maximum amplitudes of each frequency, and obtaining a clustering result set; A frequency band generation module for counting the clustering result set of the DAS original data of all calculation frames of the monitoring unit, calculating the cumulative proportion of each main frequency in the total statistics, selecting a main frequency with a cumulative proportion reaching a preset threshold as a center frequency, and expanding the bandwidth based on the center frequency to form a band-stop frequency band; and grouping the band-stop frequency bands formed by different center frequencies to form a multi-frequency band-stop of the monitoring unit. A statistical model module for adding the multi-frequency band-stop of a single monitoring unit to the total point site statistic filter frequency band set, saving the total point site statistic filter frequency band set as a statistical model, and performing bottom noise elimination on the DAS original signal based on the statistical model.

9. The statistical analysis based DAS signal floor cancellation system of claim 8, wherein, The statistical model module further includes a judgment whether the multi-frequency band-stops of all monitoring units are added to the total point site statistic filter frequency band set, and if not, returns to the frequency domain transformation module to update the monitoring unit.

10. A statistical analysis based DAS signal floor cancellation device, characterized by, It comprises a processor and a memory, wherein the processor implements the statistical analysis-based DAS signal bottom noise elimination method of any one of claims 1-7 when executing the computer program saved in the memory. It comprises a processor and a memory, wherein the processor implements the statistical analysis-based DAS signal bottom noise elimination method of any one of claims 1-7 when executing the computer program saved in the memory.

Images