Third-party intrusion event positioning method based on DAS pipeline monitoring system

Abstract

The application discloses a third-party intrusion event positioning method based on a DAS pipeline monitoring system and relates to the technical field of pipeline safety detection. The method comprises the following steps: preprocessing sound wave signals of each channel of the DAS pipeline monitoring system and identifying the event type and the confidence of each channel; judging whether it is an effective TPI event based on the sound wave characteristics, the event type and the confidence of a plurality of channels and obtaining the initial longitudinal position of the effective TPI event; selecting a plurality of channels around the initial longitudinal position to construct an event response channel cluster, calculating the theoretical sound wave energy and the theoretical wave arrival time of each channel based on a sound wave propagation path model, a sound wave energy attenuation model and soil wave velocity, constructing an optimization objective function in combination with the measured sound wave energy and the measured wave arrival time of each channel, performing nonlinear optimization solution on the optimization objective function with preset parameter constraints as boundary conditions, obtaining the lateral distance and the corrected longitudinal position of the effective TPI event and the optical fiber, and realizing high-precision two-dimensional positioning of the TPI event.

Description

Technical Field

[0001] This invention relates to the field of pipeline safety inspection technology, and more specifically, to a method for locating third-party intrusion events based on a DAS pipeline monitoring system. Background Technology

[0002] Distributed Acoustic Sensing (DAS) is a fully distributed sensing technology based on the Rayleigh backscattering principle of optical fibers. It uses optical cables laid along the pipeline as the sensing medium to continuously, over long distances, and in real-time acquire and sense vibrational acoustic signals along the fiber optic cable. This technology possesses inherent advantages such as passive explosion-proof capability, resistance to electromagnetic interference, corrosion resistance, and long monitoring distance. It has become the mainstream technology for monitoring third-party intrusion (TPI) in long-distance oil and gas pipelines, effectively detecting intrusion behaviors that threaten pipeline safety, such as mechanical excavation, manual drilling, and heavy vehicle running over them, thus possessing significant engineering value for ensuring pipeline operational safety.

[0003] As pipeline operation and maintenance become increasingly sophisticated, qualitative identification of TPI events alone is no longer sufficient for on-site response. Accurately determining the spatial location of TPI events is crucial for improving operational response efficiency. Currently, third-party intrusion event location technologies based on DAS systems can be categorized into two types: longitudinal location and lateral location. Longitudinal location is used to determine the event's extension along the fiber optic axis, while lateral location is used to determine the vertical distance from the event to the fiber optic cable. Existing longitudinal positioning methods mainly include channel peak positioning and interpolation fitting positioning. Channel peak positioning directly selects the fiber optic mileage corresponding to the sensing channel with the largest signal energy or amplitude as the event location. Its principle is simple and calculation speed is fast, but its positioning accuracy is limited by the spatial sampling interval of the fiber, only achieving channel-level accuracy. In long-distance monitoring scenarios with large spatial sampling intervals, the positioning error can reach several meters to tens of meters, failing to meet the requirements of high-precision operation and maintenance. Interpolation fitting positioning uses the energy values ​​of the peak channel and adjacent channels on both sides as a basis, obtaining the sub-channel level longitudinal position through Gaussian function interpolation. This method has low computational cost, is easy to implement in engineering, and has better positioning accuracy than the channel peak method. However, this method only performs one-dimensional fitting in the spatial dimension, and the fitting results are prone to shift in low signal-to-noise ratio scenarios, resulting in limited positioning accuracy. Lateral positioning methods mainly rely on the physical laws of sound wave propagation to invert the lateral distance. However, this method is highly dependent on empirical values ​​of environmental parameters such as soil sound velocity and / or energy attenuation coefficient, making it susceptible to soil environment and multipath interference. The positioning accuracy decreases significantly with increasing distance, and its adaptability and robustness in complex environments are poor. Summary of the Invention

[0004] The purpose of this invention is to provide a third-party intrusion event localization method based on a DAS pipeline monitoring system, in order to solve the problem that existing DAS-based third-party intrusion event localization methods mentioned in the background art are susceptible to noise and complex environments, resulting in insufficient localization accuracy and robustness.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The third-party intrusion event localization method based on the DAS pipeline monitoring system includes the following steps:

[0007] S1: Preprocess the acoustic signals collected by each channel of the DAS pipeline monitoring system to obtain acoustic features, and identify the event type and confidence level of the corresponding sampling point of each channel based on the acoustic features;

[0008] S2: Based on the acoustic wave characteristics of several channels, the event type and the confidence level, determine whether it is a valid TPI event. If it is, obtain the initial longitudinal position of the valid TPI event and then execute step S3. Otherwise, return to execute step S1.

[0009] S3: Select several channels around the initial longitudinal position along the longitudinal direction of the optical fiber to construct an event response channel cluster; obtain the measured acoustic energy and measured time of arrival of each channel within the event response channel cluster; based on the acoustic propagation path model, obtain the theoretical propagation distance from the sound source point of the effective TPI event to each channel within the event response channel cluster; based on the acoustic energy attenuation model and the theoretical propagation distance, calculate the theoretical acoustic energy of each channel; based on the theoretical propagation distance and the acoustic propagation speed, calculate the theoretical time of arrival of each channel.

[0010] S4: Based on the measured acoustic energy, theoretical acoustic energy, measured arrival time and theoretical arrival time of each channel in the event response channel cluster, construct an optimization objective function, use preset parameter constraints as boundary conditions, and perform nonlinear optimization to solve the optimization objective function to obtain the lateral distance and corrected longitudinal position of the effective TPI event relative to the fiber extension direction.

[0011] The principle of the third-party intrusion event localization method based on the DAS pipeline monitoring system of this invention is as follows: First, based on the original acoustic wave signals collected by each channel of the DAS system, the event type and confidence level of each sampling point are identified. Then, based on the original acoustic wave signals and the identification results, it is jointly determined whether a valid TPI event has occurred near the sampling point. If a valid TPI event is determined to have occurred, the initial longitudinal position of the valid TPI event is obtained based on the center of the channel cluster. This significantly reduces the calculation range for subsequent precise localization through longitudinal coarse localization, while filtering out single-point noise and non-intrusion interference, thus reducing the false alarm rate. Next, multiple consecutive channels are selected along the longitudinal direction of the optical fiber, centered on the initial longitudinal position, to construct an event response channel cluster. The path from the sound source point to the optical fiber is derived based on the acoustic wave propagation path model. The theoretical propagation distance of each channel is calculated by combining the acoustic energy attenuation model and the acoustic propagation speed to obtain the theoretical acoustic energy and theoretical time of arrival for each channel, establishing a forward mapping relationship between the sound source location and the channel observation values. Finally, the matching errors of the two independent physical characteristics of acoustic energy and time of arrival, measured and theoretical, are integrated to construct an optimization objective function, transforming the localization problem into a multi-constraint nonlinear optimization problem. With physical parameter constraints as boundary conditions, the minimum value of the objective function is iteratively solved through a nonlinear optimization algorithm, thereby inverting the corrected longitudinal position and lateral distance relative to the optical fiber of the intrusion event, achieving two-dimensional high-precision localization of TPI events, and effectively improving the robustness and accuracy of localization results in complex soil environments.

[0012] Preferably, the acoustic signals collected by each channel of the DAS pipeline monitoring system are preprocessed using a preprocessing method to obtain the acoustic features. The preprocessing method specifically includes: performing sliding window clipping on the acoustic signals to obtain several acoustic segment signals; performing normalization and bandpass filtering on the acoustic segment signals sequentially to obtain standard acoustic data; extracting the time-domain statistical features of the standard acoustic data, including: mean, extreme values, standard deviation, kurtosis, skewness, total energy, and zero-crossing rate; performing STFT transformation on the standard acoustic data to obtain the acoustic time spectrum; calculating the short-time frequency centroid of the acoustic time spectrum; and integrating the time-domain statistical features, the acoustic time spectrum, and the short-time frequency centroid into acoustic features. The raw acoustic signals collected by the DAS system contain background noise and have a single feature dimension, making it difficult to provide high-quality, multi-dimensional input features for subsequent TPI event recognition. By using signal filtering and time-domain and time-frequency multi-dimensional feature extraction, standardized and rich acoustic features can be obtained, improving the accuracy and anti-interference capability of subsequent event recognition.

[0013] Preferably, the event type and confidence level of each channel corresponding to the sampling point are identified based on the TPI event recognition model and the acoustic wave features. The TPI event recognition model includes an input layer, a convolutional coding layer, a temporal modeling layer, a self-attention layer, a fully connected layer, and an output layer connected in sequence. The input layer receives the input acoustic wave features, and the convolutional coding layer extracts the time-frequency features of the input features based on a convolutional neural network. The temporal modeling layer performs temporal modeling on the time-frequency features based on a bidirectional long short-term memory network to obtain temporal evolution features. The self-attention layer performs temporal adaptive weighting on the temporal evolution features based on a self-attention mechanism to obtain weighted temporal features. The fully connected layer performs nonlinear transformation and dimensionality reduction mapping on the weighted temporal features to obtain dimensionality-reduced features. The output layer outputs the event type and the corresponding confidence level based on the dimensionality-reduced features. Through the cascaded architecture of the convolutional neural network, the bidirectional long short-term memory network, and the self-attention layer, the deep features of the acoustic wave signal are mined from multiple dimensions, thereby improving the accuracy of TPI event type recognition.

[0014] Preferably, the determination of whether a TPI event is valid based on the acoustic wave characteristics, event type, and confidence level of several channels specifically includes: selecting several continuously distributed channels to form a test channel cluster; detecting whether the statistical value of the acoustic wave signal of each channel in the test channel cluster exceeds a preset statistical threshold within the same time period; if not, it is determined to be an invalid TPI event; if so, identifying the event type and confidence level of the corresponding sampling point of each channel in the test channel cluster within the same time period; determining whether the event type consistency rate of each channel exceeds a preset consistency rate threshold and whether the confidence level exceeds a preset confidence threshold; if both are satisfied, it is determined to be a valid TPI event; otherwise, it is determined to be an invalid TPI event. Through multi-dimensional joint verification combining initial screening of acoustic wave characteristics with fine screening of event type consistency rate and confidence level, valid TPI events are screened, thereby filtering out false anomaly events and single-point interference events, reducing the false alarm rate of TPI events.

[0015] Preferably, the method for obtaining the initial longitudinal position of the valid TPI event specifically includes: selecting the channel with the highest acoustic signal energy in the channel cluster to be tested as the center channel; performing continuous spatial fitting on the acoustic signal amplitude of all channels in the channel cluster to be tested within the same time period based on spatiotemporal two-dimensional Gaussian fitting to obtain an acoustic amplitude fitting surface, wherein the x-axis of the acoustic amplitude fitting surface is the longitudinal distance along the fiber extension direction, the y-axis is time, and the z-axis is the acoustic signal amplitude; and obtaining the initial longitudinal position based on the x-axis coordinate corresponding to the maximum acoustic signal amplitude in the acoustic amplitude fitting surface. By selecting the center channel based on acoustic energy and then combining it with spatiotemporal two-dimensional Gaussian fitting to obtain the initial longitudinal position, the acoustic amplitude fitting surface is constructed by fusing the channel amplitude distribution in the spatial dimension and the signal evolution information in the temporal dimension. This eliminates multi-channel false alarms caused by high-frequency disturbances or lightning reflections, achieves sub-channel level positioning accuracy, improves the stability of the fitting results under low signal-to-noise ratio conditions, and thus enhances the reliability of the preliminary longitudinal positioning results.

[0016] Preferably, the method for constructing the optimization objective function includes: constructing an energy residual term based on the sum of squared deviations between the measured acoustic energy and the theoretical acoustic energy of each channel in the event response channel cluster; constructing a time difference residual term based on the sum of squared deviations between the difference between the measured arrival time and the theoretical arrival time of every two channels; and constructing the optimization objective function based on the energy residual term and the time difference residual term. By jointly constructing the optimization objective function using both the energy residual and the time difference of arrival residual, and combining this with the physical propagation characteristics of sound waves in the soil to form physical constraints, the lateral distance and corrected longitudinal position are solved simultaneously through nonlinear optimization, avoiding the error accumulation of step-by-step cascaded solutions. This also overcomes the limitations of single-principle one-dimensional positioning methods, effectively reducing the influence of noise and multipath interference, and achieving high-precision and high-stability two-dimensional positioning of TPI events.

[0017] Preferably, the expression for the optimization objective function is:

[0018] ;

[0019] ;

[0020] ;

[0021] In the formula, , and These represent the optimization objective function, energy residual term, and time difference residual term, respectively. and These are the weights of the energy residual term and the time difference residual term, respectively. and These are all channel indexes within the event response channel cluster. and Representing channels The measured sound wave energy and the theoretical sound wave energy, and Representing channels and channels The difference between the measured arrival time and the theoretical arrival time.

[0022] Preferably, the preset parameter constraints include: constraints on the sound wave propagation velocity in the soil, constraints on the sound wave attenuation coefficient of the soil medium, and constraints on the deviation between the corrected longitudinal position and the initial longitudinal position. By limiting the reasonable range of the optimization solution through boundary constraints of multi-dimensional physical parameters, the nonlinear optimization output results are prevented from deviating from actual physical laws, while accelerating the convergence speed of the algorithm and improving the physical rationality and reliability of the positioning results.

[0023] Preferably, the method further includes:

[0024] S5: Based on the preset fiber optic route GIS calibration data and the horizontal distance and the corrected vertical position of the valid TPI event, determine the actual geographical coordinates of the occurrence of the valid TPI event.

[0025] By using fiber optic routing GIS calibration data, the two-dimensional location results of TPI events are converted into actual geographic coordinates, enabling the visualization and implementation of TPI event locations, which facilitates rapid on-site verification and handling by maintenance personnel.

[0026] Preferably, the event types include: no event, mechanical intrusion event, manual intrusion event, and pipeline leakage event. Among these, normal environmental noise in the absence of an event exhibits continuous and stable amplitude characteristics; mechanical intrusion events show high-frequency impact and strong amplitude abrupt changes; manual intrusion events exhibit slow, localized periodic changes in the mid-to-low frequency range; and pipeline leakage events exhibit high-frequency whistling and fixed-point resonance. These distinct characteristics ensure the separability and accuracy of the event category identification results.

[0027] One or more technical solutions provided by this invention have at least the following technical effects or advantages:

[0028] 1. By jointly constructing an optimization objective function using both energy residual and time-of-arrival residual, and utilizing the physical propagation characteristics of sound waves in the soil to form physical constraints, and combining nonlinear optimization to simultaneously solve the lateral distance and correct the longitudinal position, the limitations of the single-principle one-dimensional positioning method are overcome. This effectively reduces the influence of noise and multipath interference, while avoiding the accumulation of errors in step-by-step cascaded solutions, thus achieving high-precision and high-stability two-dimensional TPI event positioning.

[0029] 2. The initial longitudinal position is obtained by spatiotemporal two-dimensional Gaussian fitting. The channel amplitude distribution in the spatial dimension and the signal evolution information in the temporal dimension are fused to construct the acoustic wave amplitude fitting surface, thereby achieving sub-channel level positioning accuracy and improving the stability of the fitting results under low signal-to-noise ratio conditions, thus improving the reliability of longitudinal positioning.

[0030] 3. A TPI event recognition model architecture combining convolutional neural networks, bidirectional long short-term memory networks, and self-attention mechanisms is adopted. This architecture can simultaneously mine the local time-frequency texture features and long-term temporal evolution patterns of acoustic signals. Furthermore, by using temporal adaptive weighting to enhance the feature contribution of key event segments, the recognition accuracy of different types of intrusion events is significantly improved.

[0031] 4. By limiting the nonlinear solution process through multi-dimensional parameter boundary constraints, parameters that are greatly affected by the environment, such as soil sound velocity and attenuation coefficient, are controlled within a physically reasonable range. This reduces the dependence of existing positioning methods on fixed empirical parameters, and can be adapted to complex pipeline site environments with different soil types, humidity, and burial depths, thereby improving the robustness of the positioning method. Attached Figure Description

[0032] The accompanying drawings, which are provided to further illustrate embodiments of the invention and constitute a part of this invention, are not intended to limit the scope of the invention.

[0033] Figure 1 This is a flowchart illustrating the third-party intrusion event location method based on the DAS pipeline monitoring system of the present invention. Detailed Implementation

[0034] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, where there is no conflict, the embodiments of the present invention and the features thereof can be combined with each other.

[0035] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0036] Example 1

[0037] Please refer to Figure 1 The third-party intrusion event localization method based on the DAS pipeline monitoring system provided by this invention includes the following steps:

[0038] S1: The acoustic signals collected by each channel of the DAS pipeline monitoring system are preprocessed using a preprocessing method to obtain acoustic features. Based on the TPI event recognition model and the acoustic features, the event type and confidence level of the corresponding sampling point of each channel are identified. Specifically, the preprocessing method includes: performing sliding window clipping on the acoustic signals, with a window length of 1 second and a sliding step size of 0.5 seconds, to obtain several acoustic signal segments; performing z-score normalization and 10–500Hz bandpass filtering on the acoustic signal segments sequentially to obtain standard acoustic data; extracting the time-domain statistical features of the standard acoustic data, including: mean, extreme values, standard deviation, kurtosis, skewness, total energy, and zero-crossing rate; performing STFT transformation on the standard acoustic data to obtain the acoustic time spectrum; calculating the short-time frequency centroid of the acoustic time spectrum; integrating the time-domain statistical features, the acoustic time spectrum, and the short-time frequency centroid into acoustic features; the TPI event recognition model includes an input layer, a convolutional coding layer, and a time-series layer connected sequentially. The system comprises a modeling layer, a self-attention layer, a fully connected layer, and an output layer. The input layer receives the input acoustic wave features, and the convolutional coding layer extracts the time-frequency features of the input features based on a convolutional neural network. The temporal modeling layer performs temporal modeling on the time-frequency features based on a bidirectional long short-term memory network to obtain temporal evolution features. The self-attention layer performs temporal adaptive weighting on the temporal evolution features based on a self-attention mechanism to obtain weighted temporal features. The fully connected layer performs nonlinear transformation and dimensionality reduction mapping on the weighted temporal features to obtain dimensionality-reduced features. The output layer, coupled with a Softmax activation function, outputs the event type and the corresponding confidence level based on the dimensionality-reduced features. The training dataset for the TPI event recognition model is obtained by preprocessing historical event types of acoustic waves collected from each channel of the DAS pipeline monitoring system. The event types that the TPI event recognition model can identify include: no event, mechanical intrusion event, manual intrusion event, and pipeline leakage event.

[0039] S2: Based on the acoustic wave characteristics, event type, and confidence level of several channels, determine whether it is a valid TPI event. Specifically, this includes: selecting several continuously distributed channels to form a test channel cluster, detecting whether the statistical values ​​(including the maximum value, mean, and RMS of the acoustic wave amplitude) of each channel in the test channel cluster exceed a preset statistical threshold (the preset threshold for each statistical value can be set according to the 3σ principle, which is the sum of the mean and three times the standard deviation of each statistical value of the background noise). If not, it is determined to be an invalid TPI event. If yes, identify the event type and confidence level of the corresponding sampling point of each channel in the test channel cluster within the same time period, and determine whether the event type consistency rate of each channel exceeds a preset consistency rate threshold and whether the confidence level exceeds a preset confidence threshold. If both are satisfied, it is determined to be a valid TPI event; otherwise, it is determined to be an invalid TPI event. If it is determined to be a valid TPI event, obtain the initial vertical position of the valid TPI event and then execute step S3. If it is determined to be an invalid TPI event, return to execute step S1.

[0040] S3: Select several channels around the initial longitudinal position along the longitudinal direction of the optical fiber to construct an event response channel cluster. Obtain the measured acoustic energy and measured arrival time of each channel within the event response channel cluster. Based on the acoustic propagation path model, obtain the theoretical propagation distance from the sound source point of the effective TPI event to each channel within the event response channel cluster. Based on the acoustic energy attenuation model and the theoretical propagation distance, calculate the theoretical acoustic energy of each channel. Based on the theoretical propagation distance and the acoustic propagation speed, calculate the theoretical arrival time of each channel. The measured acoustic energy of each channel can be obtained by calculating the sum of the squares of the acoustic signal amplitudes, and the measured arrival time can be obtained using the Hilbert envelope peak arrival time extraction method.

[0041] S4: Based on the measured acoustic energy, theoretical acoustic energy, measured arrival time and theoretical arrival time of each channel in the event response channel cluster, construct an optimization objective function, use preset parameter constraints as boundary conditions, and perform nonlinear optimization to solve the optimization objective function to obtain the lateral distance and corrected longitudinal position of the effective TPI event relative to the fiber extension direction.

[0042] The method for obtaining the initial longitudinal position of a valid TPI event specifically includes: selecting the channel with the highest acoustic signal energy in the channel cluster to be tested as the central channel; performing continuous spatial fitting on the acoustic signal amplitude of all channels in the channel cluster to be tested within the same time period based on spatiotemporal two-dimensional Gaussian fitting to obtain an acoustic amplitude fitting surface, wherein the x-axis of the acoustic amplitude fitting surface is the longitudinal distance along the fiber extension direction, the y-axis is time, and the z-axis is the acoustic signal amplitude; obtaining the initial longitudinal position based on the x-axis coordinate corresponding to the maximum acoustic signal amplitude in the acoustic amplitude fitting surface; and to improve the positioning accuracy of the model in complex interference environments, events whose initial longitudinal position is located in non-sensitive areas (such as river crossing areas) can be filtered out by combining fiber optic routing GIS calibration data.

[0043] The method for constructing the optimization objective function includes: constructing an energy residual term based on the sum of squared deviations between the measured sound wave energy and the theoretical sound wave energy of each channel in the event response channel cluster; constructing a time difference residual term based on the sum of squared deviations between the difference in measured arrival time and the difference in theoretical arrival time of every two channels; and constructing the optimization objective function based on the energy residual term and the time difference residual term. The preset parameter constraints include: a sound wave propagation speed constraint in the soil (typically 300-1500 m / s), and a sound wave attenuation coefficient constraint in the soil medium (typically 0.1-1.0 m). -1 ) and the constraint of correcting the deviation between the longitudinal position and the initial longitudinal position (such as correcting the offset of the longitudinal position relative to the initial longitudinal position to not exceed a preset multiple of the DAS channel sampling interval).

[0044] The expression for the objective function to be optimized is:

[0045] ;

[0046] ;

[0047] ;

[0048] In the formula, , and These represent the optimization objective function, energy residual term, and time difference residual term, respectively. and These are the weights of the energy residual term and the time difference residual term, respectively. In this embodiment, we take... , , and These are all channel indexes within the event response channel cluster. and Representing channels The measured sound wave energy and the theoretical sound wave energy, and Representing channels and channels The difference between the measured time of arrival and the theoretical time of arrival;

[0049] The measured acoustic energy, measured time-of-arrival difference, theoretical acoustic energy, and theoretical time-of-arrival difference for each channel can be calculated using the following formulas:

[0050] ;

[0051] ;

[0052] ;

[0053] ;

[0054] ;

[0055] In the formula, This indicates the sampling point index of the acoustic signal in each channel. Indicates channel The The acoustic amplitude at each sampling point and Representing channels and channels The measured arrival time was obtained by performing a Hilbert transform on the acoustic signals of each channel to acquire the envelope energy curve. The time corresponding to the maximum envelope value was taken as the measured arrival time of each channel. The measured acoustic energy is the channel with the highest acoustic signal energy in the channel cluster under test. This represents the sound attenuation coefficient of the soil medium. and The sound source points representing valid TPI events are respectively located at the channel. and channels The theoretical propagation distance, This indicates the speed at which sound waves propagate in the soil. This represents the lateral distance of a valid TPI event relative to the fiber extension direction. Indicates channel In the longitudinal position of the fiber extension direction, This indicates the corrected longitudinal position of a valid TPI event in the fiber extension direction.

[0056] The method further includes:

[0057] S5: Based on the preset fiber optic route GIS calibration data and the horizontal distance and the corrected vertical position of the valid TPI event, determine the actual geographical coordinates of the occurrence of the valid TPI event.

[0058] Example 2

[0059] Based on Example 1, Example 2 will be described and illustrated with specific implementation cases.

[0060] This embodiment simulates a human intrusion event near a DAS monitoring system for an oil and gas pipeline to verify the two-dimensional positioning effect of the positioning method of the present invention on TPI events. The spatial sampling interval of the DAS monitoring system is configured as d = 1 meter / channel, the sampling frequency of each channel is 1 kHz, and the total length of the optical fiber is 50 km, so the total number of channels in the system is 50,000. Each channel is labeled and corresponds to the monitoring section: channel_0 corresponds to the starting point 0-1m section, channel_1 corresponds to the 1-2m section, and so on. A human pickaxe was used to dig at a location 2035.1 meters away from the starting point of the DAS monitoring optical cable and 2.6 meters laterally away from the monitoring optical cable to simulate a human intrusion event. During this period, the system detected a sudden strong vibration event. The channel cluster information corresponding to the event and the maximum amplitude of the acoustic signal of each channel are shown in Table 1.

[0061] Table 1. Channel and signal amplitude information corresponding to sudden strong vibration events monitored by the DAS system.

[0062] maximum signal amplitude 0.12 0.17 0.39 0.98 0.41 0.20 0.11

[0063] Based on the signal amplitude and the acoustic signal energy of each channel, the center channel is set to channel_2035, and the vertical position of the center channel is... In this embodiment, the preset consistency rate threshold for the event consistency rate of each channel is set to 0.85, and the preset confidence threshold is set to 80%. After identification, this sudden strong vibration event is a valid TPI event, the event type is a human intrusion event, the initial longitudinal position of the event is 2035.53m, and the initial longitudinal positioning error is 0.43m. Channels 2025 to 2045 are selected to construct an event response channel cluster. Based on the measured acoustic energy, measured arrival time, theoretical acoustic energy, and theoretical arrival time of each channel within the cluster, two-dimensional joint positioning of the valid TPI event is performed. The boundary conditions are set as follows: , , The lateral distance of this sudden strong vibration event was found to be 2.56m after nonlinear least square optimization. The corrected longitudinal position was 2035.26m, with a lateral positioning error of 0.04m and a final longitudinal positioning error of 0.16m. This represents a reduction of 0.27m from the initial longitudinal positioning error, indicating that this method can achieve high-precision two-dimensional positioning of TPI events.

[0064] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0065] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for locating third-party intrusion events based on a DAS pipeline monitoring system, characterized in that, Includes the following steps: S1: Preprocess the acoustic signals collected by each channel of the DAS pipeline monitoring system to obtain acoustic features, and identify the event type and confidence level of the corresponding sampling point of each channel based on the acoustic features; S2: Based on the acoustic wave characteristics of several channels, the event type and the confidence level, determine whether it is a valid TPI event. If it is, obtain the initial longitudinal position of the valid TPI event and then execute step S3. Otherwise, return to execute step S1. S3: Select several channels around the initial longitudinal position along the longitudinal direction of the optical fiber to construct an event response channel cluster, and obtain the measured acoustic energy and measured arrival time of each channel in the event response channel cluster. Based on the sound wave propagation path model, the theoretical propagation distance from the sound source point of the effective TPI event to each channel in the event response channel cluster is obtained; based on the sound wave energy attenuation model and the theoretical propagation distance, the theoretical sound wave energy of each channel is calculated; based on the theoretical propagation distance and the sound wave propagation speed, the theoretical time of arrival of each channel is calculated. S4: Based on the measured acoustic energy, theoretical acoustic energy, measured arrival time and theoretical arrival time of each channel in the event response channel cluster, construct an optimization objective function, use preset parameter constraints as boundary conditions, and perform nonlinear optimization to solve the optimization objective function to obtain the lateral distance and corrected longitudinal position of the effective TPI event relative to the fiber extension direction.

2. The third-party intrusion event localization method based on the DAS pipeline monitoring system according to claim 1, characterized in that, The acoustic signals collected by each channel of the DAS pipeline monitoring system are preprocessed using a preprocessing method to obtain the acoustic features. The preprocessing method specifically includes: performing sliding window clipping on the acoustic signals to obtain several acoustic segment signals; performing normalization and bandpass filtering on the acoustic segment signals sequentially to obtain standard acoustic data; extracting the time-domain statistical features of the standard acoustic data, including: mean, extreme values, standard deviation, kurtosis, skewness, total energy, and zero-crossing rate; performing STFT transformation on the standard acoustic data to obtain the acoustic time spectrum; calculating the short-time frequency centroid of the acoustic time spectrum; and integrating the time-domain statistical features, the acoustic time spectrum, and the short-time frequency centroid into the acoustic features.

3. The third-party intrusion event localization method based on the DAS pipeline monitoring system according to claim 1, characterized in that, The TPI event recognition model identifies the event type and confidence level of each sampling point corresponding to each channel based on the acoustic wave features. The TPI event recognition model includes an input layer, a convolutional coding layer, a temporal modeling layer, a self-attention layer, a fully connected layer, and an output layer connected in sequence. The input layer receives the input acoustic wave features, and the convolutional coding layer extracts the time-frequency features of the input features based on a convolutional neural network. The temporal modeling layer performs temporal modeling on the time-frequency features based on a bidirectional long short-term memory network to obtain temporal evolution features. The self-attention layer is used to perform time-adaptive weighting on the temporal evolution features based on the self-attention mechanism to obtain weighted temporal features; the fully connected layer is used to perform nonlinear transformation and dimensionality reduction mapping on the weighted temporal features to obtain dimensionality-reduced features. The output layer is used to output the event type and the confidence level corresponding to the event type based on the dimensionality reduction features.

4. The third-party intrusion event localization method based on the DAS pipeline monitoring system according to claim 3, characterized in that, The determination of whether a TPI event is valid based on the acoustic wave characteristics, event type, and confidence level of several channels specifically includes: selecting several continuously distributed channels to form a test channel cluster; detecting whether the statistical value of the acoustic wave signal of each channel in the test channel cluster exceeds a preset statistical threshold within the same time period; if not, it is determined to be an invalid TPI event; if so, identifying the event type and confidence level of the corresponding sampling point of each channel in the test channel cluster within the same time period; determining whether the event type consistency rate of each channel exceeds a preset consistency rate threshold and whether the confidence level exceeds a preset credibility threshold; if both are satisfied, it is determined to be a valid TPI event; otherwise, it is determined to be an invalid TPI event.

5. The third-party intrusion event localization method based on the DAS pipeline monitoring system according to claim 4, characterized in that, The method for obtaining the initial longitudinal position of the valid TPI event specifically includes: selecting the channel with the largest acoustic signal energy in the channel cluster to be tested as the center channel; performing continuous spatial fitting on the acoustic signal amplitude of all channels in the channel cluster to be tested within the same time period based on spatiotemporal two-dimensional Gaussian fitting to obtain an acoustic amplitude fitting surface, wherein the x-axis of the acoustic amplitude fitting surface is the longitudinal distance along the fiber extension direction, the y-axis is time, and the z-axis is the acoustic signal amplitude; and obtaining the initial longitudinal position based on the x-axis coordinate corresponding to the maximum acoustic signal amplitude in the acoustic amplitude fitting surface.

6. The method for locating third-party intrusion events based on a DAS pipeline monitoring system according to claim 1, characterized in that, The method for constructing the optimization objective function includes: constructing an energy residual term based on the sum of squared deviations between the measured acoustic energy and the theoretical acoustic energy of each channel in the event response channel cluster; constructing a time difference residual term based on the sum of squared deviations between the difference between the measured arrival time and the theoretical arrival time of every two channels; and constructing the optimization objective function based on the energy residual term and the time difference residual term.

7. The third-party intrusion event localization method based on the DAS pipeline monitoring system according to claim 6, characterized in that, The expression for the optimization objective function is: ； ； ； In the formula, , and These represent the optimization objective function, energy residual term, and time difference residual term, respectively. and These are the weights of the energy residual term and the time difference residual term, respectively. and These are all channel indexes within the event response channel cluster. and Representing channels The measured sound wave energy and the theoretical sound wave energy, and Representing channels and channels The difference between the measured arrival time and the theoretical arrival time.

8. The method for locating third-party intrusion events based on a DAS pipeline monitoring system according to claim 1, characterized in that, The preset parameter constraints include: constraints on the propagation speed of sound waves in the soil, constraints on the sound wave attenuation coefficient of the soil medium, and constraints on the deviation between the corrected longitudinal position and the initial longitudinal position.

9. The method for locating third-party intrusion events based on a DAS pipeline monitoring system according to claim 1, characterized in that, The method further includes: S5: Based on the preset fiber optic route GIS calibration data and the horizontal distance and the corrected vertical position of the valid TPI event, determine the actual geographical coordinates of the occurrence of the valid TPI event.

10. The method for locating third-party intrusion events based on a DAS pipeline monitoring system according to any one of claims 1-9, characterized in that, The event types include: no event, mechanical intrusion event, human intrusion event, and pipeline leak event.

Images