DAS road cross-correlation risk evaluation method based on SOM segmentation

Abstract

The invention belongs to the technical field of urban road structure safety monitoring, and discloses a DAS road cross-correlation risk evaluation method based on SOM segmentation, and the method specifically comprises the following steps: receiving a DAS signal, collected by a distributed optical fiber sound wave sensing system, of a road along an optical fiber, and carrying out the noise reduction and enhancement processing of the DAS signal; extracting feature parameter vectors of all signal channels in the distributed optical fiber acoustic wave sensing system based on silent time period data in the processed DAS signals; clustering signal channels with adjacent spatial positions and similar feature parameter vectors through a self-organizing mapping network, and dividing an optical fiber into a plurality of spatially continuous independent segments according to a clustering result; in each independent segment, multi-mode cross-correlation functions among the signal channels are calculated respectively, and peak sharpness indexes in the cross-correlation functions are extracted and synthesized; the problem of poor road risk assessment accuracy caused by non-uniform response of the DAS signal channel in the prior art is effectively solved.

Description

Technical Field

[0001] This invention belongs to the field of urban road structure safety monitoring technology, specifically involving a DAS road cross-correlation risk assessment method based on SOM segmentation. Background Technology

[0002] The long-term performance of urban roads is affected by various external factors, leading to degradation and potential safety hazards. Certain underground cavities or surface risks are particularly serious, potentially causing irreversible damage to transportation infrastructure and people's lives and property. Existing urban road structure monitoring methods largely rely on point sensors or manual inspection, making it difficult to achieve large-scale, continuous, and highly sensitive risk identification. Distributed fiber optic acoustic sensing (DAS) technology can achieve real-time vibration sensing along the entire fiber optic cable; however, due to strong signal noise, uneven channel response, and large data dimensions, traditional single-channel or fixed-window cross-correlation analysis cannot accurately reflect the consistency changes between fiber optic channels, limiting the early identification capability of underground cavities and road structural anomalies. Therefore, existing technologies suffer from poor accuracy in road risk assessment due to uneven DAS signal channel response. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a DAS road cross-correlation risk assessment method based on SOM segmentation, which solves the problem of poor accuracy in road risk assessment caused by uneven response of DAS signal channels in existing technologies.

[0004] The objective of this invention can be achieved through the following technical solutions: A method for assessing the cross-correlation risk of DAS roads based on SOM segmentation includes the following steps: It receives DAS signals along the road fiber collected by a distributed optical fiber acoustic wave sensing system and performs filtering and wave selection processing on the DAS signals. Based on the silent period data in the processed DAS signal, the feature parameter vectors of each signal channel in the distributed fiber optic acoustic wave sensing system are extracted. Signal channels that are spatially adjacent and have similar feature parameter vectors are clustered using a self-organizing map network. Based on the clustering results, the optical fiber is divided into multiple spatially continuous independent segments. Based on the effective excitation data in the DAS signal after filtering and wave selection processing, the multi-mode cross-correlation function between each signal channel is calculated in each independent segment. The peak sharpness index in each cross-correlation function is extracted and integrated to generate a channel-level reliability index for each signal channel. Based on the spatial distribution of reliability indicators along optical fibers, an identification method combining global statistical features and local mutations is used to determine the risky optical fiber regions corresponding to signal channels with abnormal reliability indicators, and generate a risk visualization map.

[0005] Furthermore, based on the silent period data in the processed DAS signal, the feature parameter vectors of each signal channel in the distributed fiber optic acoustic wave sensing system are extracted, specifically including the following steps: Steady-state characteristic parameters of each signal channel in the distributed fiber optic acoustic wave sensing system during the silent period were calculated and extracted. These steady-state characteristic parameters include power spectral density. ,variance kurtosis and time mean The specific calculation formula is as follows: in, Indicates the first Each signal channel in time DAS signal; The sampling length for the silent period; The steady-state characteristic parameters together constitute the characteristic parameter vector of the corresponding signal channel, and the specific expression is as follows: in, Indicates the first The feature parameter vector corresponding to each signal channel.

[0006] Furthermore, signal channels that are spatially adjacent and have similar feature parameter vectors are clustered using a self-organizing map network. Based on the clustering results, the optical fiber is divided into multiple spatially continuous independent segments, specifically including the following steps: The feature parameter vectors of each signal channel are input into the self-organizing map network for training. The self-organizing map network outputs a set of weight vectors representing feature parameter vectors of different categories. The feature parameter vectors of each signal channel are matched with the weight vectors, and the feature parameter vectors are assigned to the category corresponding to the weight vectors that are most similar to them. Each signal channel is assigned a clustering label. The signal channels with the same clustering labels and adjacent spatial locations on the optical fiber are merged and divided into independent segments. Based on the clustering results, the optical fiber is divided into multiple spatially continuous independent segments.

[0007] Furthermore, within each independent segment, the multi-mode cross-correlation function between each signal channel is calculated, and the peak sharpness index in each cross-correlation function is extracted and synthesized to generate a channel-level reliability index for each signal channel. This process includes the following steps: For any signal channel within an independent segment i Calculate signal channel i The multi-mode cross-correlation function between the signal and other signal channels within the independent segment, including the phase cross-correlation function, the amplitude cross-correlation function and the generalized cross-correlation function; Extract signal channel i The corresponding peak sharpness indices in various cross-correlation functions include peak energy ratio, peak sidelobe ratio, and peak root mean square ratio. Fusion signal channel i The corresponding peak sharpness indicators generate signal channels. i The corresponding reliability metrics.

[0008] Furthermore, fusion signal channels i The corresponding peak sharpness indices are calculated by taking the root mean square of each peak sharpness index.

[0009] Furthermore, based on the spatial distribution of reliability indicators along the optical fiber, an identification method combining global statistical features and local mutations is used to determine the risky optical fiber regions corresponding to signal channels with abnormal reliability indicators, generating a risk visualization map. This process includes the following steps: Calculate the global statistical characteristics of reliability metrics for all signal channels, including the global mean. and global standard deviation ; The reliability indices for each signal channel are normalized. The normalization formula is as follows: in, Indicates signal channel i The corresponding reliability metrics; Indicates signal channel i Normalized reliability index; Based on global statistical features, a global threshold is set. The specific calculation formula is as follows: in, This is an empirical coefficient; When the reliability index corresponding to the signal channel is less than the global threshold When this happens, the corresponding signal channel is determined to be a globally low-reliability channel; The local mutation identification method using the valley value method identifies locally low reliability channels, including the following steps: Define signal channel i Spatial variation rate of reliability index The specific mathematical expression is as follows: The signal channel is determined when the following conditions are met. i For a channel with locally low reliability, the expression for the determination condition is as follows: in, Local threshold; The fiber regions corresponding to global low-reliability channels and local low-reliability channels together constitute the risky fiber region. Based on the signal channel locations and corresponding reliability indicators of globally low-reliability channels and locally low-reliability channels in the risky fiber optic region, a risk visualization map is generated.

[0010] Furthermore, a DAS road cross-correlation risk assessment method based on SOM segmentation also includes: classifying and identifying risk events based on reliability indicators.

[0011] Furthermore, based on reliability indicators, risk events are classified and identified, specifically including the following steps: For different risk events to be identified and classified, multiple probability distributions are fitted based on the numerical distribution of reliability indicators, and the mean and variance of each fitted distribution are extracted. By analyzing the mean and variance of the numerical distribution of reliability indicators for different risk events, a preliminary identification and classification of data on different risk events can be achieved. For data on different risk events, the channel with the highest reliability in the respective reliability index curve is selected as the representative channel, and the corresponding feature parameter vector is obtained. Principal component analysis is used to reduce the dimensionality of the feature parameter vector; By using the unsupervised learning algorithm k-means to perform cluster analysis on the dimensionality-reduced feature parameter vectors, different types of risk events can be identified.

[0012] The beneficial effects of this invention are: This application introduces a self-organizing mapping network to adaptively spatially cluster DAS signal channels. Based on the similarity of signal features, the optical fiber is divided into multiple spatially continuous independent segments, effectively solving the problem of traditional methods struggling to rationally divide analysis units under uneven channel responses and complex noise environments. This provides a structured processing foundation for subsequent consistency analysis. Within each segment, a multi-mode cross-correlation function combined with multiple peak sharpness indices is used to comprehensively quantify the consistency between signal channels. This overcomes the limitations of traditional single-mode cross-correlation analysis, which suffers from inaccurate and unstable assessments under noise interference, and achieves robust extraction and refined characterization of reliability indices for each channel. Based on the spatial distribution of the extracted reliability indices along the optical fiber, and by integrating global statistical features and local mutation detection methods, channels with abnormal signal consistency and their corresponding risky optical fiber segments can be accurately identified, generating intuitive visualization maps. This significantly improves the early detection capability and positioning accuracy of risks and hidden dangers such as underground cavities and structural damage in roads, providing reliable technical support for preventive maintenance and safety early warning of road facilities. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 This is a schematic diagram of the Kalman filtering process of the present invention; Figure 3 This is a schematic diagram illustrating the calculation of the reliability index of the present invention; Figure 4 This is a schematic diagram of the risk event classification process of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] like Figures 1 to 4 As shown, a DAS road cross-correlation risk assessment method based on SOM segmentation specifically includes the following steps: It receives DAS signals along the road fiber collected by a distributed optical fiber acoustic wave sensing system and performs filtering and wave selection processing on the DAS signals. Based on the silent period data in the processed DAS signal, the feature parameter vectors of each signal channel in the distributed fiber optic acoustic wave sensing system are extracted. Signal channels that are spatially adjacent and have similar feature parameter vectors are clustered using a self-organizing map network. Based on the clustering results, the optical fiber is divided into multiple spatially continuous independent segments. Based on the effective excitation data in the DAS signal after filtering and wave selection processing, the multi-mode cross-correlation function between each signal channel is calculated in each independent segment. The peak sharpness index in each cross-correlation function is extracted and integrated to generate a channel-level reliability index for each signal channel. Based on the spatial distribution of reliability indicators along optical fibers, an identification method combining global statistical features and local mutations is used to determine the risky optical fiber regions corresponding to signal channels with abnormal reliability indicators, and generate a risk visualization map. This application introduces a self-organizing mapping network to adaptively spatially cluster DAS signal channels. Based on the similarity of signal features, the optical fiber is divided into multiple spatially continuous independent segments, effectively solving the problem of traditional methods struggling to rationally divide analysis units under uneven channel responses and complex noise environments. This provides a structured processing foundation for subsequent consistency analysis. Within each segment, a multi-mode cross-correlation function combined with multiple peak sharpness indices is used to comprehensively quantify the consistency between signal channels. This overcomes the limitations of traditional single-mode cross-correlation analysis, which suffers from inaccurate and unstable assessments under noise interference, and achieves robust extraction and refined characterization of reliability indices for each channel. Based on the spatial distribution of the extracted reliability indices along the optical fiber, and by integrating global statistical features and local mutation detection methods, channels with abnormal signal consistency and their corresponding risky optical fiber segments can be accurately identified, generating intuitive visualization maps. This significantly improves the early detection capability and positioning accuracy of risks and hidden dangers such as underground cavities and structural damage in roads, providing reliable technical support for preventive maintenance and safety early warning of road facilities.

[0017] To improve signal quality and stability while suppressing noise and preserving waveform details, this invention employs an unscented Kalman filter to perform nonlinear smoothing and state estimation on the bandpass-filtered signal, such as... Figure 2 As shown, its mathematical model is as follows: State-space model Let the system state vector be The observation vector is Then the DAS signal satisfies the following relationship: in: This is the state transition function; For observation functions; This is process noise; To observe noise; Point generation At time k-1, the state mean is known. With covariance Then generate indivual point: in For the state dimension, This is the expansion factor.

[0018] Time Update (Prediction) spread point: Predicted State and Covariance: Measurement update (calibration) spread Point to observation space: Calculation of observation and prediction: Calculate the cross-covariance and Kalman gain: Update state and covariance: By following the steps above, random noise and sudden interference in the DAS signal can be effectively removed, improving the accuracy of subsequent cross-correlation calculations while maintaining the true form of the signal.

[0019] The DAS signal undergoes wave selection processing to extract waveform segments containing valid acoustic emission events and remove invalid noise segments. The wave selection process includes: The peak detection method is used to scan the signal of each channel to obtain the peak value and its position; Signal segments with a consecutive peak count greater than a set threshold are defined as valid waveform segments. The effective waveform segment is truncated, and the corresponding channel characteristic parameters are calculated to characterize the dynamic response characteristics of the channel. These characteristic parameters include: Statistical feature parameters (uniform input for adaptive segmentation): Power spectral density (PSD): Reflects the energy distribution of a signal in the frequency domain and is used to characterize the consistency of channel energy response; Variance (Var): Describes the degree of fluctuation in signal amplitude and is used to measure the stability of the channel; Kurtosis: Characterizes the pulse characteristics and non-Gaussianity of a signal, and is used to identify abnormal channel responses; Time mean (Mean): Reflects the overall offset of the channel signal and the baseline change; Among them, let For the first Each channel in time The DAS signal has a sampling length of The features are defined as follows: Power spectral density: variance: Kuroshi: Time mean: Acoustic emission characteristic parameters (used for event identification and energy feature supplementation): Duration: energy: RA value (rise time / amplitude ratio): entropy: Zero crossing rate: Peak frequency: in, , The sampling positions are the start and end points of the waveform segment. The sampling period is For sampling signals, This represents the maximum amplitude of the waveform segment. For the rising time, This is the Fourier transform of the waveform segment.

[0020] The calculated characteristic parameters of each channel are used to form a unified time-frequency feature vector set, which serves as the input feature space for subsequent adaptive segmentation (SOM clustering analysis). This process effectively filters out invalid noise segments, extracts effective signal segments with acoustic emission response characteristics, and ensures the structural consistency and computational comparability between the selected characteristic parameters and the adaptive segmentation characteristic parameters, providing a high signal-to-noise ratio input for subsequent channel clustering and cross-correlation analysis.

[0021] Based on the silent period data in the processed DAS signal, the feature parameter vectors of each signal channel in the distributed fiber optic acoustic wave sensing system are extracted, specifically including the following steps: Steady-state characteristic parameters of each signal channel in the distributed fiber optic acoustic wave sensing system during the silent period were calculated and extracted. These steady-state characteristic parameters include power spectral density. ,variance kurtosis and time mean The specific calculation formula is as follows: in, Indicates the first Each signal channel in time DAS signal; The sampling length for the silent period; The steady-state characteristic parameters together constitute the characteristic parameter vector of the corresponding signal channel, and the specific expression is as follows: in, Indicates the first The feature parameter vector corresponding to each signal channel.

[0022] Signal channels that are spatially adjacent and have similar feature parameter vectors are clustered using a self-organizing map network. Based on the clustering results, the optical fiber is divided into multiple spatially continuous independent segments. The specific steps include: The feature parameter vectors of each signal channel are input into the self-organizing map network for training. The self-organizing map network outputs a set of weight vectors representing feature parameter vectors of different categories. The feature parameter vectors of each signal channel are matched with the weight vectors, and the feature parameter vectors are assigned to the category corresponding to the weight vectors that are most similar to them. Each signal channel is assigned a clustering label. The signal channels with the same clustering labels and adjacent spatial locations on the optical fiber are merged and divided into independent segments. Based on the clustering results, the optical fiber is divided into multiple spatially continuous independent segments. Self-Organizing Maps (SOMs) achieve unsupervised clustering of channels through competitive learning and neighborhood updates. The weight update rule for SOMs is as follows: in, For learning rate, For neighborhood functions, for The weight vector of each neuron.

[0023] Through training, clustering labels for each channel can be obtained, forming a distribution mapping of fiber optic response patterns; Based on the SOM clustering results, adjacent channels belonging to the same category are divided into the same segment interval. If the category of adjacent channels changes, that location is defined as the segment boundary. ,Right now: in For the first Cluster labels for each channel. The set of each segment interval is represented as: in This represents the total number of segments.

[0024] Within each independent segment, the multi-mode cross-correlation function between each signal channel is calculated. The peak sharpness index in each cross-correlation function is extracted and synthesized to generate a channel-level reliability index for each signal channel. The specific steps include: For any signal channel within an independent segment i Calculate signal channel i The multi-mode cross-correlation function between the signal and other signal channels within the independent segment, including the phase cross-correlation function, the amplitude cross-correlation function and the generalized cross-correlation function; Extract signal channel i The corresponding peak sharpness indices in various cross-correlation functions include peak energy ratio, peak sidelobe ratio, and peak root mean square ratio. Fusion signal channel i The corresponding peak sharpness indicators generate signal channels. i The corresponding reliability metrics; Fusion signal channel i The methods for calculating the various peak sharpness indices include the root mean square of each peak sharpness index. Let the first With the The waveforms of each channel are as follows: , Its cross-correlation function is denoted as .like Figure 3 As shown, this paper uses three types of cross-correlation: Amplitude Cross-Correlation Function (ACCF) Phase cross-correlation function (PCCF) (preserves only phase in the frequency domain) Generalized cross-correlation function (GCCF) in For Fourier transform, For conjugate, ϵ is the numerically stable term; Common weight functions such as PHAT and ROTH can be used.

[0025] Let the three types of cross-correlation The main peak lag is A symmetrical window W (the length of which is determined by the signal duration and sampling rate) is taken at the center of the main peak to calculate the local RMS.

[0026] For each type of cross-correlation (ACC / PCC / GCC), we define individual reliability using three peak sharpness metrics (PRMSR, PSR, and PCE). .

[0027] remember: PRMSR type (peak / root mean square ratio) in , That is, through a length of 2L, with its center located at The window W was used to exclude the main correlation peak. The value at that point. The original definition of PRMSR comes from the ratio of the peak value to the root mean square of the "region below 50% of the peak value". In engineering, the main peak window approximation is often used for ease of implementation and statistical robustness.

[0028] PSR type (peak / prescription difference away from peak) in It is a range far from the main peak, such as sampling at ±L / 2 or statistical variance at the far side window after decentering a certain range in the correlation length (−L,L), that is, using the far-end noise / side lobe variance to calibrate the peak "sharpness".

[0029] PCE type (peak / correlated plane energy) Continuous form is PCE is insensitive to noisy pixels, and its resolution is feasible and easy to optimize.

[0030] Applying the above sharpness definition to any cross-correlation type yields the corresponding... There are nine expressions in total.

[0031] The formal writing style can be standardized as follows: Where t∈{ACC,PCC,GCC} refers to the cross-correlation type, and m∈{PCE,PSR,PRMSR} refers to the sharpness measure. This indicates the calculation of the three corresponding peak sharpness indices.

[0032] Using the above methods, we obtain the first... A 1×N vector with 1 channel = [ , , …, [To estimate each DAS channel] Reliability, calculation The root mean square (RMS) value of a vector, and it is defined as follows: ,Right now: in In order to discard the corresponding channel autocorrelation value N is the total number of channels. Furthermore, this process is repeated sequentially for all channels, resulting in a 1×N vector. = [ , , …, This includes reliability metrics for all channels.

[0033] Based on the spatial distribution of reliability indicators along the optical fiber, an identification method combining global statistical features and local mutations is used to determine the risky optical fiber regions corresponding to signal channels with abnormal reliability indicators, and to generate a risk visualization map. The specific steps include: Calculate the global statistical characteristics of reliability metrics for all signal channels, including the global mean. and global standard deviation ; The reliability indices for each signal channel are normalized. The normalization formula is as follows: in, Indicates signal channel i The corresponding reliability metrics; Indicates signal channel i Normalized reliability index; Based on global statistical features, a global threshold is set. The specific calculation formula is as follows: in, This is an empirical coefficient, which is usually 1. When the reliability index corresponding to the signal channel is less than the global threshold When this happens, the corresponding signal channel is determined to be a globally low-reliability channel; The local mutation identification method using the valley value method identifies locally low reliability channels, including the following steps: Define signal channel i Spatial variation rate of reliability index The specific mathematical expression is as follows: The signal channel is determined when the following conditions are met. i For a channel with locally low reliability, the expression for the determination condition is as follows: in, The local threshold is set to -0.8; preferably, the local threshold is set to -0.8. The fiber regions corresponding to global low-reliability channels and local low-reliability channels together constitute the risky fiber region. Based on the signal channel locations and corresponding reliability indicators of global and local low-reliability channels in the risky fiber optic region, these are projected onto fiber optic spatial coordinates and combined with the geographical location of fiber optic deployment to map the road, generating a risk visualization map along the road. By combining global statistical features with local mutation information, adaptive identification of low-reliability channels can be achieved. This method maintains high sensitivity and robustness even in complex urban road scenarios with non-uniform fiber response and significant signal-to-noise ratio variations, thereby improving the accuracy and interpretability of identifying potential underground structural hazards.

[0034] like Figure 4 As shown, a DAS road cross-correlation risk assessment method based on SOM segmentation also includes: classifying and identifying risk events based on reliability indicators.

[0035] Based on reliability indicators, risk events are classified and identified, specifically including the following steps: For different risk events to be identified and classified, multiple probability distributions are fitted based on the numerical distribution of reliability indicators, and the mean and variance of each fitted distribution are extracted. By analyzing the mean and variance of the numerical distribution of reliability indicators for different risk events using box plots, preliminary identification and classification of data for different risk events can be achieved. For data on different risk events, the channel with the highest reliability in the respective reliability index curve is selected as the representative channel, and the corresponding feature parameter vector is obtained. Principal component analysis is used to reduce the dimensionality of the feature parameter vector; The k-means unsupervised learning algorithm is used to perform cluster analysis on the dimensionality-reduced feature parameter vector to identify different types of risk events. Various probability distributions include the normal distribution, log-normal distribution, exponential distribution, Weibull distribution, and gamma distribution; After fitting, the fitting results were verified using Kolmogorov-Smirnov (KS) hypothesis testing to obtain the mean and variance statistical parameters of each distribution. The formula for calculating the normal distribution is as follows: The formula for calculating the log-normal distribution is as follows: The formula for calculating the exponential distribution is as follows: The formula for calculating the Weibull distribution is as follows: The formula for calculating the gamma distribution is as follows: In the above formula, x The data is for fitting; μ and σ These are the mean and standard deviation, respectively. λ In the exponential distribution, the rate parameter is represented; in the Weibull distribution, the scale parameter is represented. k It is a shape parameter; gamma distribution. α These are shape parameters. β It is a scale parameter. Γ (α) is the gamma function; The unsupervised learning algorithm k-means is applied to perform cluster analysis on the dimensionality-reduced data. The k-means clustering algorithm divides the data into a specified number of clusters through iterative optimization, with each cluster centered on its centroid, thereby further realizing the classification and identification of different events. Preferably, this application also provides a comprehensive evaluation model for multiple risks of urban roads; the model comprehensively considers the spatial distribution characteristics of signal coupling consistency, adaptive segmentation results, and reliability index β, and can quantify the risk level of different sections; it identifies underground cavity hazards through threshold-valley adaptive criteria, and classifies multiple types of risk events by combining principal component analysis (PCA) and K-means clustering algorithms; finally, it generates a visual map of urban road risks, realizing an intuitive presentation of underground cavities and surface risk events, and supporting real-time monitoring and intelligent early warning of road safety status.

[0036] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0037] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for assessing the cross-correlation risk of DAS roads based on SOM segmentation, characterized in that, Specifically, the following steps are included: It receives DAS signals along the road fiber collected by a distributed optical fiber acoustic wave sensing system and performs filtering and wave selection processing on the DAS signals. Based on the silent period data in the processed DAS signal, the feature parameter vectors of each signal channel in the distributed fiber optic acoustic wave sensing system are extracted. Signal channels that are spatially adjacent and have similar feature parameter vectors are clustered using a self-organizing map network. Based on the clustering results, the optical fiber is divided into multiple spatially continuous independent segments. Based on the effective excitation data in the DAS signal after filtering and wave selection processing, the multi-mode cross-correlation function between each signal channel is calculated in each independent segment. The peak sharpness index in each cross-correlation function is extracted and integrated to generate a channel-level reliability index for each signal channel. Based on the spatial distribution of reliability indicators along optical fibers, an identification method combining global statistical features and local mutations is used to determine the risky optical fiber regions corresponding to signal channels with abnormal reliability indicators, and generate a risk visualization map.

2. The DAS road cross-correlation risk assessment method based on SOM segmentation according to claim 1, characterized in that, Based on the silent period data in the processed DAS signal, the feature parameter vectors of each signal channel in the distributed fiber optic acoustic wave sensing system are extracted, specifically including the following steps: Steady-state characteristic parameters of each signal channel in the distributed fiber optic acoustic wave sensing system during the silent period were calculated and extracted. These steady-state characteristic parameters include power spectral density. ,variance kurtosis and time mean The specific calculation formula is as follows: in, Indicates the first Each signal channel in time DAS signal; The sampling length for the silent period; The steady-state characteristic parameters together constitute the characteristic parameter vector of the corresponding signal channel, and the specific expression is as follows: in, Indicates the first The feature parameter vector corresponding to each signal channel.

3. The DAS road cross-correlation risk assessment method based on SOM segmentation according to claim 1, characterized in that, Signal channels that are spatially adjacent and have similar feature parameter vectors are clustered using a self-organizing map network. Based on the clustering results, the optical fiber is divided into multiple spatially continuous independent segments. The specific steps include: The feature parameter vectors of each signal channel are input into the self-organizing map network for training. The self-organizing map network outputs a set of weight vectors representing feature parameter vectors of different categories. The feature parameter vectors of each signal channel are matched with the weight vectors, and the feature parameter vectors are assigned to the category corresponding to the weight vectors that are most similar to them. Each signal channel is assigned a clustering label. The signal channels with the same clustering labels and adjacent spatial locations on the optical fiber are merged and divided into independent segments. Based on the clustering results, the optical fiber is divided into multiple spatially continuous independent segments.

4. The DAS road cross-correlation risk assessment method based on SOM segmentation according to claim 3, characterized in that, Within each independent segment, the multi-mode cross-correlation function between each signal channel is calculated. The peak sharpness index in each cross-correlation function is extracted and synthesized to generate a channel-level reliability index for each signal channel. The specific steps include: For any signal channel within an independent segment i Calculate signal channel i The multi-mode cross-correlation function between the signal and other signal channels within the independent segment, including the phase cross-correlation function, the amplitude cross-correlation function and the generalized cross-correlation function; Extract signal channel i The corresponding peak sharpness indices in various cross-correlation functions include peak energy ratio, peak sidelobe ratio, and peak root mean square ratio. Fusion signal channel i The corresponding peak sharpness indicators generate signal channels. i The corresponding reliability metrics.

5. The DAS road cross-correlation risk assessment method based on SOM segmentation according to claim 4, characterized in that, Fusion signal channel i The corresponding peak sharpness indices are calculated by taking the root mean square of each peak sharpness index.

6. The DAS road cross-correlation risk assessment method based on SOM segmentation according to claim 4, characterized in that, Based on the spatial distribution of reliability indicators along the optical fiber, an identification method combining global statistical features and local mutations is used to determine the risky optical fiber regions corresponding to signal channels with abnormal reliability indicators, and to generate a risk visualization map. The specific steps include: Calculate the global statistical characteristics of reliability metrics for all signal channels, including the global mean. and global standard deviation ; The reliability indices for each signal channel are normalized. The normalization formula is as follows: in, Indicates signal channel i The corresponding reliability metrics; Indicates signal channel i Normalized reliability index; Based on global statistical features, a global threshold is set. The specific calculation formula is as follows: in, This is an empirical coefficient; When the reliability index corresponding to the signal channel is less than the global threshold When this happens, the corresponding signal channel is determined to be a globally low-reliability channel; The local mutation identification method using the valley value method identifies locally low reliability channels, including the following steps: Define signal channel i Spatial variation rate of reliability index The specific mathematical expression is as follows: The signal channel is determined when the following conditions are met. i For a channel with locally low reliability, the expression for the determination condition is as follows: in, Local threshold; The fiber regions corresponding to global low-reliability channels and local low-reliability channels together constitute the risky fiber region. Based on the signal channel locations and corresponding reliability indicators of globally low-reliability channels and locally low-reliability channels in the risky fiber optic region, a risk visualization map is generated.

7. The DAS road cross-correlation risk assessment method based on SOM segmentation according to claim 1, characterized in that, Also includes: Risk events are classified and identified based on reliability indicators.

8. The DAS road cross-correlation risk assessment method based on SOM segmentation according to claim 7, characterized in that, Based on reliability indicators, risk events are classified and identified, specifically including the following steps: For different risk events to be identified and classified, multiple probability distributions are fitted based on the numerical distribution of reliability indicators, and the mean and variance of each fitted distribution are extracted. By analyzing the mean and variance of the numerical distribution of reliability indicators for different risk events, a preliminary identification and classification of data on different risk events can be achieved. For data on different risk events, the channel with the highest reliability in the respective reliability index curve is selected as the representative channel, and the corresponding feature parameter vector is obtained. Principal component analysis is used to reduce the dimensionality of the feature parameter vector; By using the unsupervised learning algorithm k-means to perform cluster analysis on the dimensionality-reduced feature parameter vectors, different types of risk events can be identified.

Images