A method and system for cable external damage detection

Abstract

The application relates to the technical field of electric power operation and maintenance, and discloses a cable external damage detection method and system. The method comprises the following steps: collecting first vibration data of a cable through a first vibration sensor arranged on a first edge computing terminal; if the vibration amplitude of the first vibration data exceeds a preset amplitude threshold, a virtual reconstruction signal is generated; the virtual reconstruction signal is sent to a second edge computing terminal to trigger the second edge computing terminal to extract second vibration data of the cable; after the second vibration data is received, virtual reconstruction is performed based on the vibration data of the two, and respective corresponding reconstructed vibration data is obtained; preliminary detection is performed on the two, and a first cable external damage detection result is obtained; if the detection result is a valid event, the first reconstructed vibration data is input into a preset cable external damage detection model, and a second cable external damage detection result is output. The application improves the identification accuracy and response real-time performance of cable external damage, and guarantees the working efficiency and long-term stable operation of a cable external damage early warning system.

Description

Technical Field

[0001] This application relates to the field of power operation and maintenance technology, and more specifically, to a method and system for detecting external damage to cables. Background Technology

[0002] With the rapid expansion of urban underground cable networks, external damage to cables (such as construction excavation and drilling) has become one of the main causes of power system failures. Existing external damage monitoring technologies mostly rely on single vibration threshold alarms or simple time-frequency characteristic analysis, which generally suffer from problems such as high false alarm rates, low positioning accuracy, and delayed response.

[0003] Furthermore, traditional distributed monitoring systems mostly use Network Time Protocol (NTP) for clock synchronization, and its millisecond-level synchronization accuracy is insufficient to meet the high-precision positioning requirements based on vibration wave time difference of arrival (TDOA). At the same time, continuously uploading large amounts of raw vibration data to the cloud for processing not only consumes significant communication bandwidth but also introduces substantial transmission delays, making it difficult to achieve millisecond-level real-time early warning.

[0004] Although edge computing technology can complete some data processing locally and alleviate the pressure on the cloud, existing edge terminals are limited by computing power and storage resources, making it difficult to deploy high-precision deep learning models and lacking an effective online model adaptive update mechanism. They are also unable to cope with complex and ever-changing field environments and new external damage characteristics. There is a lack of a real-time early warning method for cable external damage that integrates high-precision time synchronization, collaborative perception mechanism and edge self-evolution capability.

[0005] There is currently no effective solution to the aforementioned technical problems. Summary of the Invention

[0006] This application provides a method and system for detecting external damage to cables, which at least solves the technical problem that the detection method for external damage to cables is singular and cannot fully reflect the external damage status of cables.

[0007] According to one aspect of the embodiments of this application, a cable external damage detection method is provided, applied to a first edge computing terminal in a cable external damage detection system. The cable external damage detection system includes a cloud server and multiple edge computing terminals, wherein the multiple edge computing terminals are disposed at multiple cable monitoring points along the cable line, including:

[0008] First vibration data of the cable is collected by a first vibration sensor installed in the first edge computing terminal;

[0009] If the vibration amplitude of the first vibration data exceeds a preset amplitude threshold, a virtual reconstruction signal is generated;

[0010] The virtual reconstruction signal is sent to a second edge computing terminal adjacent to the first edge computing terminal, so that the second edge computing terminal can extract the second vibration data of the cable based on the virtual reconstruction signal. The second vibration data is set as the vibration data of the cable collected by the second vibration sensor of the second edge computing terminal, and the second vibration data corresponds to the timestamp of the first vibration data.

[0011] Receive the second vibration data returned by the second edge computing terminal;

[0012] Virtual reconstruction is performed based on the first vibration data and the second vibration data to obtain first reconstructed vibration data and second reconstructed vibration data.

[0013] Based on the first reconstructed vibration data and the second reconstructed vibration data, external cable damage detection is performed to obtain the first external cable damage detection result.

[0014] If the first cable external damage detection result is a valid event, the first reconstructed vibration data is input into the preset cable external damage detection model for detection to obtain the second cable external damage detection result, wherein the valid event indicates that there is a cable external damage.

[0015] Optionally, before acquiring the first vibration data via the first vibration sensor located on the first edge computing terminal, the method further includes:

[0016] Based on the positioning module of the first edge computing terminal, the clock module of the first edge computing terminal is initialized;

[0017] The clock module is compensated based on the temperature data collected by the digital temperature sensor installed in the first edge computing terminal.

[0018] After acquiring the first vibration data through the first vibration sensor installed on the first edge computing terminal, the method further includes:

[0019] Based on the standard system time set in the clock module of the first edge computing terminal, a first timestamp is marked at the first sampling point of the first vibration data.

[0020] Optionally, generating a virtual reconstruction signal if the vibration amplitude of the first vibration data exceeds a preset amplitude threshold includes:

[0021] If the vibration amplitude of multiple consecutive sampling points exceeds the preset amplitude threshold, a virtual reconstructed signal is generated;

[0022] Sending the virtual reconstruction signal to a second edge computing terminal adjacent to the first edge computing terminal includes:

[0023] The first identification information of the first edge computing terminal and the virtual reconstruction signal are sent to the second edge computing terminal.

[0024] Optionally, the virtual reconstruction based on the first vibration data and the second vibration data to obtain the first reconstructed vibration data and the second reconstructed vibration data includes:

[0025] Based on the first timestamp of the first vibration data and the second timestamp of the second vibration data, determine the time deviation between the first timestamp and the second timestamp;

[0026] Calculate the sampling rate deviation ratio between the first vibration data and the second vibration data;

[0027] Based on the ratio of the time deviation to the sampling rate deviation, the second vibration data is reconstructed using a linear interpolation algorithm to obtain the second reconstructed vibration data.

[0028] The first reconstructed vibration data is determined to be the first vibration data.

[0029] Optionally, the step of performing cable external damage detection based on the first reconstructed vibration data and the second reconstructed vibration data to obtain the first cable external damage detection result includes:

[0030] The first reconstructed vibration data and the second reconstructed vibration data are subjected to coarse interference filtering to obtain the cross-correlation coefficient between the first reconstructed vibration data and the second reconstructed vibration data.

[0031] If the cross-correlation coefficient is greater than or equal to a preset coefficient threshold, then the first cable external damage detection result is determined to be a valid event;

[0032] If the cross-correlation coefficient is less than the preset coefficient threshold, then the first cable external damage detection result is determined to be local interference.

[0033] Optionally, the step of inputting the first reconstructed vibration data into a preset cable external damage detection model for detection to obtain a second cable external damage detection result includes:

[0034] The first reconstructed vibration data is input into the cable external damage detection model, and multidimensional features of the first reconstructed vibration data are extracted, wherein the multidimensional features include time domain features and frequency domain features;

[0035] The multidimensional features are concatenated to obtain a multidimensional feature vector;

[0036] Based on the multidimensional feature vector, the cable is subjected to external damage detection to obtain the external damage type and the corresponding external damage confidence level, wherein the external damage type includes at least one of excavation, drilling and impact.

[0037] Optionally, after obtaining the second cable external damage detection result, the method further includes:

[0038] Based on the clock module installed in the first edge computing terminal, the location calculation of the first reconstructed vibration data and the second reconstructed vibration data is performed to obtain the geographical location of the external failure point;

[0039] Multi-dimensional early warning data is generated based on the maximum vibration amplitude of the first vibration data, the second cable external damage detection result, and the geographical location of the external damage point, and the multi-dimensional early warning data is sent to the cloud server.

[0040] Optionally, the step of performing location calculations on the first reconstructed vibration data and the second reconstructed vibration data based on the clock module installed in the first edge computing terminal to obtain the geographical location of the external failure point includes:

[0041] Windowing and fast Fourier transform are applied to the first reconstructed vibration data and the second reconstructed vibration data to obtain first frequency domain data and second frequency domain data;

[0042] Based on the subsampling-level time delay estimation algorithm of generalized cross-correlation and phase transformation weight, the weighted cross power spectral density of the first frequency domain data and the second frequency domain data is determined;

[0043] The weighted cross-power spectral density is subjected to an inverse fast Fourier transform to obtain the generalized cross-correlation function in the time domain;

[0044] A quadratic parabolic interpolation fitting is performed on the correlation peak of the generalized cross-correlation function and its left and right adjacent points to determine the precise position of the parabola vertex and obtain the time delay estimate.

[0045] Based on the clock module, the temperature difference between the current ambient temperature and the calibrated temperature is obtained, and the propagation speed of the vibration wave in the cable is calibrated according to the temperature difference to obtain the calibrated propagation speed.

[0046] Based on the calibration propagation speed and the estimated time delay, a distance equation is constructed between the external breaking point and the first edge computing terminal and the second edge computing terminal;

[0047] Solve the distance equation to obtain the preliminary coordinates of the outer breach point, and use the Kalman filter algorithm to generate the optimal estimated coordinates of the preliminary coordinates to obtain the geographical location of the outer breach point.

[0048] Optionally, the method further includes:

[0049] Receive cable correction instructions returned by the cloud server based on the multi-dimensional early warning data;

[0050] Using the cable correction command as the label for the first reconstructed vibration data, training samples are generated;

[0051] The cable external damage detection model is updated based on the training samples.

[0052] Optionally, updating the cable external damage detection model based on the training samples includes:

[0053] The training samples are stored in the transient learning library of the first edge computing terminal;

[0054] If the number of new training samples in the transient learning library exceeds a preset threshold, or if the external damage confidence in the second cable external damage detection result is less than a preset confidence threshold, then the cable external damage detection model is self-evolved and trained using a lightweight particle swarm optimization algorithm to obtain an updated cable external damage detection model.

[0055] Optionally, if the number of new training samples in the transient learning library exceeds a preset threshold, or the external damage confidence in the second cable external damage detection result is less than a preset confidence threshold, then the cable external damage detection model is self-evolved using a lightweight particle swarm optimization algorithm to obtain an updated cable external damage detection model, including:

[0056] The hyperparameter encoding of the cable external damage detection model is used as the position vector of the lightweight particle swarm algorithm, and the fitness function of the lightweight particle swarm algorithm is determined.

[0057] A chaotic sequence is generated using logistic regression mapping, and the chaotic sequence is mapped to the solution space of the particles to obtain an initial particle swarm.

[0058] Load the candidate hyperparameters corresponding to each initial particle in the initial particle swarm, and input multiple training samples for training to obtain the candidate first cable external damage detection model;

[0059] The first fitness value corresponding to the first cable external damage detection model is calculated using the fitness function;

[0060] The initial particle swarm is updated to obtain an updated particle swarm;

[0061] Load the candidate hyperparameters corresponding to each updated particle in the updated particle swarm, and input multiple training samples for training to obtain the candidate second cable external damage detection model;

[0062] The fitness function is used to calculate the second fitness value corresponding to the second cable external damage detection model;

[0063] The particle swarm and fitness values ​​are iteratively updated to obtain the globally optimal particle;

[0064] The position vector of the globally optimal particle is decoded to obtain the optimal hyperparameter, and the candidate cable external damage detection model corresponding to the optimal hyperparameter is taken as the optimal cable external damage detection model.

[0065] According to another aspect of this embodiment, a cable external damage detection system is also provided, including a cloud server and multiple edge computing terminals. The multiple edge computing terminals are installed at multiple cable monitoring points along the cable line. The first edge detection terminal among the multiple edge computing terminals includes:

[0066] The first acquisition module is configured to acquire first vibration data of the cable through a first vibration sensor installed in the first edge computing terminal;

[0067] The generation module is configured to generate a virtual reconstruction signal if the vibration amplitude of the first vibration data exceeds a preset amplitude threshold.

[0068] The sending module is configured to send the virtual reconstruction signal to a second edge computing terminal adjacent to the first edge computing terminal, so that the second edge computing terminal can extract the second vibration data of the cable based on the virtual reconstruction signal, wherein the second vibration data is the vibration data of the cable collected by the second vibration sensor of the second edge computing terminal, and the second vibration data corresponds to the timestamp of the first vibration data;

[0069] The receiving module is configured to receive the second vibration data returned by the second edge computing terminal;

[0070] The virtual reconstruction module is configured to perform virtual reconstruction based on the first vibration data and the second vibration data to obtain first reconstructed vibration data and second reconstructed vibration data.

[0071] The first cable external damage detection module is configured to perform cable external damage detection based on the first reconstructed vibration data and the second reconstructed vibration data, and obtain the first cable external damage detection result;

[0072] The second cable external damage detection module is configured to input the first reconstructed vibration data into a preset cable external damage detection model for detection if the first cable external damage detection result is a valid event, thereby obtaining the second cable external damage detection result, wherein the valid event indicates that there is a cable external damage.

[0073] The cable external damage detection method and system provided in this application collect first vibration data of the cable through a first vibration sensor installed in a first edge computing terminal; if the vibration amplitude of the first vibration data exceeds a preset amplitude threshold, a virtual reconstruction signal is generated; the virtual reconstruction signal is sent to a second edge computing terminal, triggering it to extract second vibration data of the cable; after receiving the second vibration data, virtual reconstruction is performed based on the vibration data of both to obtain corresponding reconstructed vibration data; preliminary detection is performed on both to obtain a first cable external damage detection result; if the detection result is a valid event, the first reconstructed vibration data is input into a preset cable external damage detection model, and a second cable external damage detection result is output. This application significantly improves the accuracy and real-time response of cable external damage events, ensuring the working efficiency and long-term stable operation of the cable external damage early warning system. Attached Figure Description

[0074] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0075] Figure 1 A hardware block diagram of a computer terminal for implementing a cable external damage detection method is shown.

[0076] Figure 2 This is a schematic flowchart of the cable external damage detection method provided in the embodiments of this application;

[0077] Figure 3 This is another schematic flowchart of the cable external damage detection method provided according to the embodiments of this application;

[0078] Figure 4 This is another flowchart illustrating the cable external damage detection method provided in the embodiments of this application;

[0079] Figure 5 This is a structural block diagram of a cable external damage detection system provided according to an embodiment of this application;

[0080] Figure 6 This is another structural block diagram of the cable external damage detection system provided according to the embodiments of this application. Detailed Implementation

[0081] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0082] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0083] According to an embodiment of this application, a method for detecting external damage to cables is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0084] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal for implementing a method for detecting external cable damage is shown. Figure 1As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0085] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0086] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and an electronic program for a real-time early warning device for cable damage that integrates edge computing and BeiDou positioning.

[0087] exist Figure 1 In the electronic device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device. The electronic device calls the electronic program of the real-time early warning device for external cable damage fusion of edge computing and positioning stored in the memory 1005 through the processor 1001, and executes the real-time early warning method for external cable damage fusion of edge computing and Beidou positioning provided in the embodiment of the present invention.

[0088] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits can be implemented wholly or partially as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits can be a single, independent processing module, or wholly or partially integrated into any other element in a computer terminal. As involved in the embodiments of this application, the data processing circuit serves as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0089] The memory 1005 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the cable external damage detection method in this embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 1005, thereby implementing the cable external damage detection method of the aforementioned application. The memory 1005 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 1005 may further include memory remotely located relative to the processor, and these remote memories can be connected to a computer terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0090] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows users to interact with the user interface of the computer terminal.

[0091] Figure 2 This is a flowchart illustrating the cable external damage detection method provided in the embodiments of this application, as shown below. Figure 2 As shown, the method includes the following steps:

[0092] Step S201: First vibration data of the cable is collected by the first vibration sensor installed in the first edge computing terminal.

[0093] like Figure 5 As shown in the embodiment of this application, the cable external damage detection method is applied to the first edge computing terminal in the cable external damage detection system. The cable external damage detection system includes a cloud server and multiple edge computing terminals, which are set at multiple cable monitoring points along the cable.

[0094] Before implementing the above method, multiple monitoring points are set up along the cable route, and an edge computing terminal is deployed at each monitoring point. All edge computing terminals are connected to a cloud server to construct an edge computing sensing network, which is the cable external damage detection system. In specific implementation, a monitoring point can be set up every 500 meters to 2 kilometers, depending on the cable route and the geographical characteristics of the cable's location. At least one edge computing terminal (e.g., one) is deployed at each monitoring point. All edge computing terminals establish communication connections with the cloud server through wired fiber optic or 4G / 5G wireless networks, forming a star or bus-type edge computing sensing network to achieve distributed sensing, local processing, and cloud-based collaboration. The edge computing terminal is preferably a low-power microcontroller unit (MCU), such as the STM32L4 series, which has multiple low-power modes and high-performance computing capabilities.

[0095] Each edge computing terminal is equipped with a vibration sensor to collect cable vibration signals in real time around the monitoring points along the cable. The vibration sensor is preferably a high-sensitivity vibration sensor (such as a piezoelectric accelerometer).

[0096] In this step, the first vibration data of the cable is collected in real time by the first vibration sensor on the first edge computing terminal set at the first monitoring point along the first cable.

[0097] Step S202: If the vibration amplitude of the first vibration data exceeds a preset amplitude threshold, a virtual reconstruction signal is generated.

[0098] After acquiring the first vibration data via the first vibration sensor in step S201, the first edge computing terminal performs amplitude determination on the first vibration data. If a vibration signal exceeding a preset amplitude threshold is detected in the first vibration data, it is determined that a potential cable breakage event may exist. At this time, the first edge computing terminal generates a "virtual reconstruction signal" to initiate a collaborative analysis request to the adjacent second edge computing terminal. This signal is a lightweight control command that does not contain the original vibration data and is only used to trigger subsequent multi-terminal data synchronization and joint detection processes. The second edge computing terminal is located at a second cable-line monitoring point along the cable line, adjacent to the first cable-line monitoring point. The second edge computing terminal is the nearest neighboring node on the cable line that is physically closest to the first edge computing terminal and has a reachable communication link; it is typically the edge computing terminal of the preceding or following monitoring point.

[0099] Step S203: The virtual reconstruction signal is sent to a second edge computing terminal adjacent to the first edge computing terminal, so that the second edge computing terminal extracts the second vibration data of the cable based on the virtual reconstruction signal. The second vibration data is set as the vibration data of the cable collected by the second vibration sensor of the second edge computing terminal, and the second vibration data corresponds to the timestamp of the first vibration data.

[0100] This step involves initiating the collaborative detection request and triggering data retrieval. Based on step S202, after the first edge computing terminal generates a virtual reconstruction signal, this signal is sent via a network interface to a second edge computing terminal that is physically adjacent to the cable and has a reachable communication link. Upon receiving the virtual reconstruction signal, the second edge computing terminal identifies the source of the cable external damage detection request based on the identification information of the first edge computing terminal carried in the signal, and accordingly retrieves second vibration data from its local cache whose acquisition time window is aligned with the timestamp of the first vibration data.

[0101] The second vibration data is collected in real time by the second vibration sensor integrated in the second edge computing terminal. Its sampling start time is time-correlated with the first timestamp of the first vibration data, ensuring accurate time alignment in subsequent virtual reconstruction.

[0102] Step S204: Receive the second vibration data returned by the second edge computing terminal.

[0103] In this step, after the second edge computing terminal extracts and encapsulates the second vibration data based on the virtual reconstruction signal, the first edge computing terminal receives the second vibration data through a network interface to obtain the second vibration data with a second absolute timestamp. The second vibration data includes the original sampling sequence and corresponding time reference information, which is used for subsequent time deviation calculation, sampling rate correction, and virtual reconstruction in the first edge computing terminal.

[0104] Step S205: Based on the first vibration data and the second vibration data, perform virtual reconstruction to obtain first reconstructed vibration data and second reconstructed vibration data.

[0105] Based on step S204, after the first edge computing terminal receives the second vibration data from the second edge computing terminal, the second vibration data carries a second absolute timestamp and forms a time-correlated dual-end synchronized data pair with the previously acquired first vibration data (including the first timestamp). To eliminate time offset caused by clock drift and sampling rate differences between terminals, this step performs virtual reconstruction of the first vibration data and the second vibration data with time axis alignment, making them comparable under a unified time reference, and providing a structurally consistent input signal for subsequent cross-correlation analysis and precision detection.

[0106] Step S206: Based on the first reconstructed vibration data and the second reconstructed vibration data, perform cable external damage detection to obtain the first cable external damage detection result.

[0107] After the virtual reconstruction in step S205 is completed on the first edge computing terminal, first reconstructed vibration data and second reconstructed vibration data are obtained. The two have achieved physical-level time-domain alignment under a unified time reference, forming a dual-channel vibration signal input with consistent structure and strong comparability. In this step, the cable external damage detection model deployed locally on the first edge computing terminal is used to extract features and classify the first reconstructed vibration data, and output the first cable external damage detection result. This result includes the external damage type, such as excavation, drilling, impact, and its corresponding confidence level.

[0108] Step S207: If the first cable external damage detection result is a valid event, the first reconstructed vibration data is input into the preset cable external damage detection model for detection to obtain the second cable external damage detection result, wherein the valid event indicates that there is a cable external damage.

[0109] After completing the fine detection in step S206, the first edge computing terminal obtains the first cable external damage detection result, which includes the external damage type and confidence level. This step performs secondary verification based on this result: only when the result is determined to be a valid event is a secondary detection process with higher confidence triggered to further confirm and optimize the detection output, generating a second cable external damage detection result.

[0110] In practice, a cable external damage detection model is constructed on a cloud server using machine learning algorithms, and then deployed to all edge computing terminals connected to the cloud server. The cloud training, conducted on the cloud server, utilizes historical excavation, drilling, and impact vibration datasets, employing machine learning algorithms such as Support Vector Machine (SVM) or Long Short-Term Memory (LSTM) to train the cable external damage detection model. The model input consists of vibration signal features, and the output is the external damage type and confidence level. The edge deployment involves lightweighting the trained model (e.g., pruning and quantizing), and then remotely distributing and deploying it to the local storage of all edge computing terminals via the network.

[0111] The cable external damage detection method provided in this application embodiment collects first vibration data of the cable through a first vibration sensor installed on a first edge computing terminal; if the vibration amplitude of the first vibration data exceeds a preset amplitude threshold, a virtual reconstruction signal is generated; the virtual reconstruction signal is sent to a second edge computing terminal to trigger it to extract second vibration data of the cable; after receiving the second vibration data, virtual reconstruction is performed based on the vibration data of both to obtain corresponding reconstructed vibration data; preliminary detection of cable external damage is performed using the reconstructed vibration data of both to obtain a first cable external damage detection result; and when the first cable external damage detection result is a valid event, the first reconstructed vibration data is input into a preset cable external damage detection model for fine detection of cable external damage to obtain a second cable external damage detection result. This method can achieve cross-terminal, low-latency, and high-confidence collaborative identification of external damage events, improve the accuracy of cable external damage detection and the real-time performance of detection response, and ensure the working efficiency and long-term reliable and stable operation of the cable external damage early warning system.

[0112] As an optional embodiment, before acquiring the first vibration data via a first vibration sensor located on the first edge computing terminal, the method further includes:

[0113] S2011, Based on the positioning module of the first edge computing terminal, initialize the clock module of the first edge computing terminal;

[0114] S2012, based on the temperature data collected by the digital temperature sensor installed in the first edge computing terminal, the clock module is compensated.

[0115] Before acquiring the first vibration data through the first vibration sensor, it is necessary to establish and calibrate a high-precision time reference to ensure the consistency of the timing and positioning accuracy of subsequent vibration data.

[0116] Each edge computing terminal is equipped with a digital temperature sensor (such as DS18B20), a clock module, and a positioning module. The second pulse signal output by the positioning module is used as the trigger signal for the clock module. The positioning module preferably supports the BD2B1 frequency, such as a BeiDou positioning module or a GPS positioning module. The high-frequency clock module is preferably a high-stability temperature-compensated crystal oscillator. The second pulse signal output pin of the positioning module is directly connected to the capture / compare input pin of the clock module (such as the MCU's timer) via a circuit. An interrupt trigger is configured in the edge computing terminal to configure the MCU's internal timer to run with an internal high-frequency clock source (such as 32MHz).

[0117] Whenever the rising edge of the second pulse signal output by the positioning module arrives, an external interrupt is triggered on the MCU. In the interrupt service function, the current timer count value is immediately read and recorded, and the system timestamp of the edge computing terminal is updated. The MCU uses a software algorithm to calculate the deviation between the actual arrival interval of the Beidou second pulse signal and the theoretical 1 second, i.e., the cumulative error of the local crystal oscillator, and performs crystal oscillator discipline. When the positioning signal is normal, the edge computing terminal achieves "hard synchronization" between the local clock and the positioning module by dynamically adjusting the frequency division coefficient of the real-time clock or directly subtracting the deviation in the timestamp calculation.

[0118] Subsequently, based on the temperature data collected by the digital temperature sensor, a temperature-drift error compensation model is built on the edge computing terminal. This model is used to compensate for the system time provided by the clock module when the second pulse signal of the positioning module is lost.

[0119] During the power-on initialization phase of the edge computing terminal or when the BeiDou signal is good, data points are collected every 5 minutes, including readings from the built-in temperature sensor and the cumulative clock error at that temperature. The least squares method is used to perform a quadratic polynomial fitting on several sets of collected data points to obtain the relationship for the temperature-drift error compensation model:

[0120] ;

[0121] In the formula, This represents the error value for temperature T; is the fitting coefficient; T is the temperature data.

[0122] At each edge computing terminal, the clock module is initialized based on a temperature-drift error compensation model to obtain a standard system time. When the positioning signal is normal, calibration is primarily performed using the second pulse signal, with model fitting results used as auxiliary monitoring to determine the degree of crystal oscillator aging. When the positioning signal is lost, the MCU reads the current temperature in real time, substitutes it into the above formula to calculate the current error, and then manually adds the compensation value when calculating the system time to obtain the standard system time. This ensures that the time accuracy remains at the microsecond level even in tunnels without positioning signals.

[0123] Then, after completing the clock module initialization and temperature compensation, the clock module is initialized to obtain accurate standard system time. The first edge computing terminal activates the first vibration sensor to collect raw vibration data along the cable, which serves as the input signal for subsequent cable breakage event detection.

[0124] As an optional embodiment, after acquiring the first vibration data through a first vibration sensor installed on the first edge computing terminal, the method further includes:

[0125] S2013, based on the standard system time set in the clock module of the first edge computing terminal, mark the first timestamp at the first sampling point of the first vibration data.

[0126] In this embodiment, based on the triggered acquisition mechanism, after acquiring the first vibration data, a first timestamp is added to the first sampling point of the first vibration data based on the standard system time of the clock module. This marks the first vibration data, facilitating subsequent analysis of the second vibration data from the corresponding second edge terminal based on the first timestamp. The first timestamp is an absolute timestamp. The absolute timestamp is marked by accurately recording the time of the first sampling point of the first vibration data using the standard system time obtained in step S2012, and using it as the first absolute timestamp of the data segment. This timestamp serves as the reference for subsequent positioning calculations.

[0127] As an optional embodiment, step S202 can be implemented according to the following steps: if the vibration amplitude of the first vibration data exceeds a preset amplitude threshold, generating a virtual reconstruction signal includes:

[0128] S2021, if the vibration amplitude of multiple consecutive sampling points exceeds the preset amplitude threshold, a virtual reconstruction signal is generated.

[0129] To improve the accuracy of subsequent cable damage detection, the first edge computing terminal analyzes the vibration amplitude of the first vibration data from multiple consecutive sampling points. When the vibration amplitude of multiple consecutive sampling points exceeds the preset amplitude threshold, a potential cable damage event is identified, a virtual reconstruction signal is generated, and the process proceeds to the next step; otherwise, the first vibration data is stored locally. For example, if the vibration amplitude of multiple non-contiguous sampling points exceeds the preset amplitude threshold, this non-contiguous vibration change may be due to a sudden vibration change caused by external interference, rather than a cable damage event. In this case, there is no need to generate a virtual reconstruction signal.

[0130] The first edge computing terminal employs a sliding window mechanism to continuously analyze the first vibration data. When the vibration amplitude of a preset number of sampling points (e.g., 5) exceeds the preset amplitude threshold within a preset time window (e.g., 50 milliseconds), the vibration event is determined to be persistent and significant, and not random noise. At this time, the first edge computing terminal generates a "virtual reconstruction signal," a collaborative request instruction used to initiate a data synchronization request to adjacent terminals to verify whether the event is a cross-terminal co-occurrence of cable damage, rather than localized interference. If the vibration amplitude of multiple consecutive sampling points does not exceed the preset amplitude threshold, it is determined to be environmental noise. The first vibration data is then compressed and stored locally, overwriting older data to save storage space.

[0131] In this embodiment, when the edge computing terminal acquires vibration data, it does not continuously collect data, but is in a low-power monitoring mode. When the vibration sensor detects that the vibration amplitude exceeds the basic sensitivity threshold, it triggers the edge computing terminal to perform high sampling rate (e.g., 20kHz) data acquisition and cache the first vibration data.

[0132] As an optional embodiment, in step S203, sending the virtual reconstruction signal to a second edge computing terminal adjacent to the first edge computing terminal includes:

[0133] S2031: Send the first identification information of the first edge computing terminal and the virtual reconstruction signal to the second edge computing terminal.

[0134] After generating the virtual reconstruction signal, the first edge computing terminal encapsulates it into a collaboration request message via its network interface and sends it to a second edge computing terminal located adjacent to it along the cable route. During transmission, the first edge computing terminal encapsulates its own first identification information along with the virtual reconstruction signal as the complete content of the collaboration request and sends it to the second edge computing terminal. Upon receiving the message, the second edge computing terminal identifies the source of the request based on the first identification information and extracts the second vibration data within the time window corresponding to the first timestamp, providing a data basis for subsequent time alignment and virtual reconstruction in the first edge computing terminal. The first identification information is preferably a first IP address, which allows the second edge computing terminal to accurately send the extracted second vibration data to the first edge computing terminal, achieving precise data transmission.

[0135] As an optional embodiment, step S205 can be implemented according to the following steps: the virtual reconstruction based on the first vibration data and the second vibration data to obtain the first reconstructed vibration data and the second reconstructed vibration data includes:

[0136] S2051, determine the time deviation between the first timestamp and the second timestamp based on the first timestamp of the first vibration data and the second timestamp of the second vibration data;

[0137] S2052, Calculate the sampling rate deviation ratio between the first vibration data and the second vibration data;

[0138] S2053, Based on the ratio of the time deviation to the sampling rate deviation, the second vibration data is reconstructed using a linear interpolation algorithm to obtain the second reconstructed vibration data;

[0139] S2054, determine the first reconstructed vibration data as the first vibration data.

[0140] Steps S2051 to S2054 are the process of performing virtual reconstruction based on the first vibration data and the second vibration data to obtain the first reconstructed vibration data and the second reconstructed vibration data.

[0141] First, based on the first timestamp marked by the first edge computing terminal for the first vibration data in step S2013, and the second absolute timestamp marked synchronously by the second edge computing terminal when collecting the second vibration data, the time deviation between the first timestamp and the second timestamp is calculated. This time deviation reflects the system time difference between the two terminals at the start of the collection, and its sources include crystal oscillator frequency drift, network transmission delay, or BeiDou signal synchronization error.

[0142] Next, the sampling rate deviation ratio of the first vibration data and the second vibration data is calculated using the following formula:

[0143] ;

[0144] in, This is due to time deviation; The first absolute timestamp and the second absolute timestamp; The first sampling rate for the first vibration data and the second sampling rate for the second vibration data; This represents the sampling rate deviation ratio.

[0145] Then, based on the ratio of time deviation to sampling rate deviation, a linear interpolation algorithm is used to reconstruct the second vibration data, obtaining the second reconstructed vibration data. The first vibration data is then used as the first reconstructed vibration data. The formula is as follows:

[0146] ;

[0147] in, For the first n Second reconstructed vibration data from the sampling points; For the first Second vibration data from the sampling point; n, This is the sampling point indication value.

[0148] This step physically aligns the first and second vibration data on the time axis and reconstructs the second vibration data. The first edge computing terminal determines the originally acquired first vibration data as the first reconstructed vibration data.

[0149] As an optional embodiment, such as Figure 3 As shown, step S206 can be implemented according to the following steps: The cable external damage detection is performed based on the first reconstructed vibration data and the second reconstructed vibration data to obtain the first cable external damage detection result, including:

[0150] S2061, perform interference coarse filtering on the first reconstructed vibration data and the second reconstructed vibration data to obtain the cross-correlation coefficient between the first reconstructed vibration data and the second reconstructed vibration data;

[0151] S2062, if the cross-correlation coefficient is greater than or equal to the preset coefficient threshold, then the first cable external damage detection result is determined to be a valid event;

[0152] S2063, if the cross-correlation coefficient is less than the preset coefficient threshold, then the first cable external damage detection result is determined to be local interference.

[0153] In steps S2061 to S2063, firstly, interference coarse filtering is performed on the reconstructed first and second reconstructed vibration data to obtain the cross-correlation coefficient, the formula of which is:

[0154] ;

[0155] in, The cross-correlation coefficient; These are the first and second reconstructed vibration data after normalization. n This is the indicator value for the sampling point; W This represents the total number of sampling points within the time window.

[0156] If the cross-correlation coefficient is greater than or equal to a preset coefficient threshold (e.g., 0.3), it indicates that the signal is simultaneously captured by two nodes and has a high correlation, suggesting a high probability of a cable breakage event. In this case, the coarse detection result is output as a valid event, and the process proceeds to the fine detection step. The coarse detection refers to a rapid, low-computing-power, and highly robust preliminary screening process at the edge based on the cross-correlation of signals from multiple terminals. Its goal is to filter noise and local interference, retaining only potential breakage events with multi-point collaborative characteristics. The fine detection, after passing the coarse detection, inputs the vibration data of the valid event into the cable breakage detection model deployed at the edge, performs multi-dimensional feature extraction and deep classification, and outputs the specific breakage type and confidence level. Conversely, if the cross-correlation coefficient is less than the preset coefficient threshold, it indicates that the signals from the two nodes have extremely low similarity. In this case, the coarse detection result is output as local interference, such as wind blowing the terminal or a vehicle directly running over the sensor housing. The first and second reconstructed vibration data are discarded, only ordinary recording is performed, no alarm is triggered, and the process returns to the data acquisition step.

[0157] As an optional embodiment, such as Figure 4 As shown, step S207 can be implemented according to the following steps: inputting the first reconstructed vibration data into a preset cable external damage detection model for detection, and obtaining the second cable external damage detection result, includes:

[0158] S2071, The first reconstructed vibration data is input into the cable external damage detection model, and the multidimensional features of the first reconstructed vibration data are extracted, wherein the multidimensional features include time domain features and frequency domain features;

[0159] S2072, The multidimensional features are concatenated to obtain a multidimensional feature vector;

[0160] S2073, Based on the multidimensional feature vector, perform external damage detection on the cable to obtain the external damage type and the corresponding external damage confidence level of the cable.

[0161] In steps S2071 to S2073, firstly, after confirming that the first cable external damage detection result (coarse detection result) is a valid event, the first edge computing terminal re-calls the cable external damage detection model deployed locally, and uses the first reconstructed vibration data obtained in step S205 as input. The cable external damage detection model performs feature extraction on the input signal to obtain multidimensional features of the first reconstructed vibration data. The multidimensional features include time-domain features, such as peak value, peak factor, impulse factor, margin factor, and kurtosis; and frequency-domain features, which are calculated by fast Fourier transform (FFT) to extract the dominant frequency, spectral centroid, and energy ratio of the 0~50Hz and 50~150Hz frequency bands.

[0162] Secondly, the first edge computing terminal concatenates all the time-domain and frequency-domain features extracted in step S2071 into vectors in a fixed order to form a multi-dimensional feature vector with a consistent structure. This vector dimension is consistent with the feature vectors used in the model training phase, ensuring input format compatibility.

[0163] Finally, the first edge computing terminal inputs the multidimensional feature vector generated in step S2072 into the classifier layer of the cable external damage detection model, performs secondary classification inference, and outputs the second cable external damage detection result (precise detection result). The second cable external damage detection result includes the cable external damage type and the corresponding external damage confidence level. The external damage type includes excavation, drilling, and impact, etc., and the external damage confidence level refers to the re-evaluation probability value output by the model, which is used to compare the change in confidence level of the first cable external damage detection result.

[0164] As an optional embodiment, after obtaining the second cable external damage detection result through step S207, the method further includes:

[0165] S301, based on the clock module set in the first edge computing terminal, the first reconstructed vibration data and the second reconstructed vibration data are used to perform positioning calculations to obtain the geographical location of the external failure point;

[0166] S302, based on the maximum vibration amplitude of the first vibration data, the second cable external damage detection result and the geographical location of the external damage point, multi-dimensional early warning data is generated and sent to the cloud server.

[0167] After obtaining the second cable external damage detection result based on steps S2071 to S2073, in this step, the first edge computing terminal calls the locally deployed clock module. Using the microsecond-level high-precision system time provided by the clock module as a reference, it performs time delay estimation and spatial positioning calculation on the first and second reconstructed vibration data to determine the specific geographical location of the external damage event along the cable line. After obtaining the geographical location of the external damage point, the first edge computing terminal integrates three types of key information: the maximum vibration amplitude of the first vibration data, reflecting the damage intensity; the second cable external damage detection result, i.e., the external damage type and external damage confidence level, reflecting the nature of the event; and the geographical location of the external damage point, reflecting the spatial location of the event. Based on this, structured multi-dimensional early warning data is generated. This data includes fields such as time, location, type, confidence level, and amplitude, conforming to the system's custom communication protocol, such as the CoAP / MQTT payload format. Subsequently, this early warning data is uploaded to the cloud server in real time via NB-IoT or 4GCat.1 network to trigger maintenance work orders, map mapping, and historical data analysis.

[0168] As an optional embodiment, step S301 can be implemented according to the following steps: The step of performing location calculations on the first reconstructed vibration data and the second reconstructed vibration data based on the clock module installed in the first edge computing terminal to obtain the geographical location of the external failure point includes:

[0169] S3011, Windowing and Fast Fourier Transform are applied to the first reconstructed vibration data and the second reconstructed vibration data to obtain the first frequency domain data and the second frequency domain data;

[0170] S3012, based on the subsampling-level time delay estimation algorithm of generalized cross-correlation and phase transformation weight, determine the weighted cross power spectral density of the first frequency domain data and the second frequency domain data;

[0171] S3013, Perform an inverse fast Fourier transform on the weighted cross-power spectral density to obtain the generalized cross-correlation function in the time domain;

[0172] S3014, Perform quadratic parabolic interpolation fitting on the correlation peak of the generalized cross-correlation function and its left and right adjacent points to determine the precise position of the parabola vertex and obtain the time delay estimate.

[0173] S3015, Based on the clock module, obtain the temperature difference between the current ambient temperature and the calibrated temperature, and calibrate the propagation speed of the vibration wave in the cable according to the temperature difference to obtain the calibrated propagation speed;

[0174] S3016, Based on the calibration propagation speed and the estimated time delay, construct the distance equation between the external breaking point and the first edge computing terminal and the second edge computing terminal;

[0175] S3017, Solve the distance equation to obtain the preliminary coordinates of the outer breach point, and use the Kalman filter algorithm to generate the optimal estimated coordinates of the preliminary coordinates to obtain the geographical location of the outer breach point.

[0176] Steps S3011 to S3017 are the process of the first edge computing terminal performing high-precision positioning calculation based on the dual-terminal collaborative vibration signal. This process takes the time reference provided by the clock module as the core, and achieves accurate spatial positioning of the cable external break point through generalized cross-correlation subsampling level time delay estimation, temperature compensation wave velocity calibration and Kalman filtering optimization.

[0177] First, in order to eliminate spectral leakage and improve frequency resolution, the system first applies a Hamming window to the first reconstructed vibration data and the second reconstructed vibration data respectively for windowing processing. Then, it performs a Fast Fourier Transform (FFT) to convert the time-domain vibration signal to the frequency domain, thereby obtaining the first frequency domain data and the second frequency domain data respectively.

[0178] Secondly, based on the subsampling-level time delay estimation algorithm of generalized cross-correlation, phase transformation weights are introduced to calculate the weighted cross-power spectral density of the first and second frequency domain data, as shown in the formula:

[0179] ;

[0180] in, The weighted cross-power spectral density; This is the data in the first frequency domain; For second frequency domain data The complex conjugate; This is the phase transformation weighting factor; For regularization parameters; This is a frequency indication value.

[0181] Next, an inverse fast Fourier transform is performed on the weighted cross-power spectral density to obtain the generalized cross-correlation function in the time domain. Then, a quadratic parabolic interpolation fitting is performed between the correlation peak of the generalized cross-correlation function and its left and right adjacent points to fit the parabolic equation passing through these three points. The precise position of the parabola vertex is calculated to obtain the time delay estimate.

[0182] Then, based on the temperature-drift error compensation model of the clock module, the temperature difference between the current ambient temperature and the calibration temperature is obtained. Based on this temperature difference, the propagation speed of the vibration wave in the cable is calibrated to obtain the calibrated propagation speed, as shown in the formula:

[0183] ;

[0184] in, For the calibrated propagation speed; Standard wave speed (e.g., 2800 m / s); Temperature-wave velocity coefficient of the cable-soil system (must be pre-calibrated); This represents the temperature difference.

[0185] Next, based on the calibrated propagation speed and time delay estimates, the distance equations between the external break point and the current edge computing terminal and adjacent edge computing terminals are constructed, as follows:

[0186] ;

[0187] in, This refers to the distance between the external breach point and the current edge computing terminal, and the distance between the external breach point and the adjacent edge computing terminal. It represents the distance difference; This is an estimate of the time delay.

[0188] Finally, combining the known geographic coordinates of the first and second edge computing terminals, the distance equation is solved using the least squares method to obtain the preliminary coordinates of the outer breach point. Since this solution is affected by noise, minor clock jitter, and wave speed fluctuations, the preliminary coordinates contain some error. To improve accuracy and stability, the system further introduces a Kalman filter algorithm, using historical positioning trajectories or model predictions as priors, to perform state estimation and error suppression on the current preliminary coordinates, generating the optimal estimated coordinates of the preliminary coordinates, and thus obtaining the geographic location of the outer breach point.

[0189] As an optional embodiment, the cable external damage detection method can also be implemented according to the following steps:

[0190] S401, Receive the cable correction instruction returned by the cloud server based on the multi-dimensional early warning data;

[0191] S402, using the cable correction instruction as the label for the first reconstructed vibration data, generate training samples;

[0192] S403, Update the cable external damage detection model based on the training samples.

[0193] Steps S401 to S403 constitute the closed-loop feedback mechanism for the self-evolution training of the cable external damage detection model. This mechanism can use the first edge computing terminal as the core execution subject, and utilize the locally stored reconstructed vibration data and the correction labels sent from the cloud to achieve lightweight, real-time, privacy-secure self-evolution of the model without uploading the original data or relying on cloud computing power.

[0194] First, the first edge computing terminal continuously monitors downlink communication from the cloud server via its network interface. When maintenance personnel manually confirm the multi-dimensional early warning data reported by the first edge computing terminal on the cloud platform, the cloud server generates a cable correction instruction. This instruction is a structured data packet that clearly indicates the true status of the early warning event, such as "confirmed external damage" or "false alarm." The multi-dimensional early warning data includes the type of external damage, confidence level, maximum vibration amplitude, and geographical location. Upon receiving the cable correction instruction, the first edge computing terminal matches it with the locally cached first reconstructed vibration data based on the early warning identifiers it carries, such as timestamp, terminal IP, and sample ID, preparing for subsequent sample generation.

[0195] Secondly, the first edge computing terminal uses the real label contained in the cable correction instruction as a supervision signal and binds it to the locally stored first reconstructed vibration data to form a labeled training sample with clear semantics. This training sample includes: input data (first reconstructed vibration data); label data (the real external damage category or state indicated by the cable correction instruction); and metadata (acquisition time, terminal ID, original confidence level, and temperature compensation coefficient (used for model analysis)). This training sample is written into the local transient learning library and stored in the edge terminal's non-volatile memory as a data source for subsequent model self-evolution training.

[0196] Finally, the training samples are used to continuously update and iterate the cable damage detection model. By iteratively searching for the optimal combination of hyperparameters, such as the penalty factor, learning rate, and kernel function parameters, the classification accuracy of the model on the local real sample set is maximized.

[0197] As an optional embodiment, step S4031 can be implemented according to the following steps: updating the cable external damage detection model based on the training samples includes:

[0198] S4031, Store the training samples in the transient learning library of the first edge computing terminal;

[0199] S4032, if the number of new training samples in the transient learning library exceeds a preset threshold, or the confidence level in the second cable external damage detection result is less than a preset confidence threshold, then the cable external damage detection model is self-evolved and trained using a lightweight particle swarm optimization algorithm to obtain an updated cable external damage detection model.

[0200] In this step, the process of updating the cable external damage detection model includes: after generating the training sample, the first edge computing terminal does not upload it to the cloud, but stores it in local non-volatile memory (such as SPIFlash or eMMC), forming a transient learning library specifically for model self-evolution training. The first edge computing terminal continuously monitors the accumulated number of new training samples in the transient learning library, and records the confidence level of the second cable external damage detection results output in the most recent few fine detections. When any of the following triggering conditions are met, the system automatically starts the self-evolution training process: quantity condition, the number of new samples in the transient learning library is greater than or equal to a preset quantity threshold; confidence condition, in the most recent N fine detections, the average confidence level is less than a preset confidence threshold, indicating that the uncertainty of the model's discrimination of external damage features in the current environment has increased and its performance has decreased.

[0201] Once triggered, the first edge computing terminal calls the locally deployed lightweight particle swarm optimization algorithm to perform hyperparameter optimization on the currently running cable external damage detection model, rather than model structure reconstruction, to ensure efficient execution even under MCU resource constraints.

[0202] As an optional embodiment, step S4032 can be implemented according to the following steps: if the number of new training samples in the transient learning library exceeds a preset threshold, or the confidence level in the second cable external damage detection result is less than a preset confidence threshold, then the cable external damage detection model is self-evolved through a lightweight particle swarm optimization algorithm to obtain an updated cable external damage detection model, including:

[0203] S501, the hyperparameter encoding of the cable external damage detection model is used as the position vector of the lightweight particle swarm algorithm, and the fitness function of the lightweight particle swarm algorithm is determined.

[0204] S502, Use logistic regression mapping to generate a chaotic sequence, and map the chaotic sequence to the solution space of the particles to obtain an initial particle swarm;

[0205] S503, load the candidate hyperparameters corresponding to each initial particle in the initial particle swarm, and input multiple training samples for training to obtain the candidate first cable external damage detection model;

[0206] S504, using the fitness function, calculate the first fitness value corresponding to the first cable external damage detection model;

[0207] S505, Update the position of the initial particle swarm to obtain an updated particle swarm;

[0208] S506, Load the candidate hyperparameters corresponding to each updated particle in the updated particle swarm, and input multiple training samples for training to obtain the candidate second cable external damage detection model.

[0209] S507, using the fitness function, calculate the second fitness value corresponding to the second cable external damage detection model;

[0210] S508 iteratively updates the particle swarm and fitness values ​​to obtain the globally optimal particle;

[0211] S509, decode the position vector of the globally optimal particle to obtain the optimal hyperparameter, and take the candidate cable external damage detection model corresponding to the optimal hyperparameter as the optimal cable external damage detection model.

[0212] Steps S501 to S509 are the process of using a lightweight particle swarm optimization algorithm to perform self-evolution training on the cable external damage detection model to obtain an updated cable external damage detection model.

[0213] First, the hyperparameters of the cable external damage detection model, such as the penalty factor and kernel function parameters of SVM, and the learning rate and number of layers of LSTM, are encoded into position vectors for a lightweight particle swarm optimization (PSO) algorithm, and a fitness function for the lightweight PSO algorithm is defined. This fitness function describes the model's classification accuracy on several training samples in a transient learning library.

[0214] Secondly, to avoid local convergence and uneven distribution caused by traditional random initialization, the system uses logistic regression to generate a chaotic sequence, and then maps the chaotic sequence to the solution space of the particles to obtain the initial particle swarm; the formula is:

[0215] (1) ;

[0216] in, Let n+1 and n be the chaotic variables; The stability coefficient is, for example, 4. This sequence has ergodicity and randomness, which can ensure that the initial particle swarm is uniformly distributed in the solution space, avoiding getting trapped in local optima, which is better than traditional random initialization. n It is a chaotic variable indicator.

[0217] (2) ;

[0218] in, For the initial particle swarm, the first i An initial particle; For the first i One chaotic variable; These are the upper and lower bounds of the search space; This is a particle indicator.

[0219] (3) ;

[0220] in, For the first i The initial velocity of the initial particle; It is a random number in the range (-1, 1).

[0221] Then, the hyperparameter combination represented by each particle in the initial particle swarm is loaded one by one and temporarily loaded into the locally deployed cable external damage detection model to form a candidate model, namely the first candidate cable external damage detection model.

[0222] Then, the fitness function defined in step S501 is used to evaluate the first cable external damage detection model: multiple identical training samples (e.g., 20) from the transient learning library are input; the predicted label is output and compared with the true label; the classification accuracy is calculated as the first fitness value of the particle. The first fitness value reflects the performance of the set of hyperparameters on the local data and is used for subsequent particle ranking.

[0223] Next, the system updates the positions of the initial particle swarm to obtain the updated particle swarm; the formula is:

[0224] (1) ;

[0225] in, The update rate of the i-th particle at iteration number t+1; The update velocity of the i-th particle in iteration number t is the initial velocity in the first iteration; The improved inertia weight is the convergence factor for the iteration number t; The globally optimal particle for iteration number t; This is the i-th updated particle in iteration t, and the initial particle in the first iteration; The position of the i-th particle at iteration number t is the historical best position; t is the current iteration number. The cooperation coefficient; is a random number in (0, 1); i is the particle indicator.

[0226] (2) ;

[0227] in, These represent the maximum and minimum values ​​of the inertia weight; This is the threshold for the number of iterations; , To adjust the parameters; It is a hyperbolic tangent function; this design results in a larger weight in the early stage, which is beneficial for global search; and a smaller weight and a gradual change in the later stage, which is beneficial for fine-grained local mining.

[0228] (3) ;

[0229] in, This refers to the i-th updated particle at iteration number t+1.

[0230] Then, the new hyperparameter combination represented by each particle in the particle swarm is loaded and temporarily added to the model. Lightweight training is performed using the same number of training samples to generate a second cable breakage detection model, which is the set of candidate models for this iteration. Similarly, the fitness function is used to evaluate each second candidate model to obtain a second fitness value, which is used to compare with the previous generation and update the individual optimum and the global optimum.

[0231] Finally, the position vector of the globally optimal particle, i.e., the optimal hyperparameter combination, is decoded into actual parameter values ​​and loaded into the local cable external damage detection model, replacing the hyperparameter configuration of the original model. This model is the optimal cable external damage detection model and is immediately used for the next fine inspection, achieving immediate effect upon update.

[0232] In summary, the cable external damage detection method provided in this application achieves microsecond-level clock consistency between terminals by constructing a second-pulse hard-synchronized edge sensing network; effectively distinguishes between local interference and real external damage events by combining a collaborative coarse detection mechanism based on virtual reconstruction signals; introduces a temperature-compensated calibration generalized cross-correlation subsampling positioning algorithm to achieve meter-level or even sub-meter-level accurate positioning of external damage points even in environments without positioning signals; and continuously optimizes the model locally on the edge terminal through a transient learning library and a lightweight particle swarm self-evolutionary training architecture, without requiring cloud retraining or uploading of original data. This solution constructs an end-side intelligent closed loop of perception, collaboration, positioning, and self-evolution without relying on manually labeled big data, significantly improving the real-time performance, accuracy, and environmental adaptability of cable external damage early warning. It solves the technical problems of low synchronization accuracy, high false alarm rate, inaccurate positioning, and model rigidity in traditional cable external damage detection, providing a deployable, auditable, and scalable industrial-grade solution for high-reliability, low-maintenance, and highly secure intelligent operation and maintenance of urban underground cable networks, strongly supporting the construction of digital twin power grids and new power systems.

[0233] Specifically, compared with the prior art, this application has the following beneficial effects:

[0234] (1) By introducing a high-precision time synchronization source and an ambient temperature drift compensation mechanism, microsecond-level time consistency between edge computing terminals is achieved, effectively overcoming the synchronization deviation caused by temperature fluctuations and signal obstruction in the clock module, and providing stable and reliable time base support for subsequent high-precision positioning based on time difference.

[0235] (2) By constructing a virtual reconstructed signal collaborative detection mechanism, the vibration data is linked to adjacent nodes after single-point triggering to perform temporal alignment and cross-correlation analysis, thereby achieving efficient identification and filtering of local interference. Only events with spatial correlation are sent to the subsequent fine detection process, which significantly reduces the false alarm rate.

[0236] (3) A time delay estimation algorithm combining generalized cross-correlation and subsampling interpolation is adopted, and an environmental parameter dynamic calibration model is integrated to adaptively correct the signal velocity in the propagation medium, which greatly improves the resolution and robustness of time delay estimation, thereby achieving meter-level or even sub-meter-level spatial positioning accuracy of external damage events.

[0237] (4) By deploying the entire process of data collection, feature extraction, coarse detection and fine detection on the edge terminal, local real-time response is achieved, which greatly reduces the data transmission bandwidth requirements and cloud computing load, and improves the overall response efficiency and reliability.

[0238] (5) By constructing an edge transient learning library and combining it with a lightweight swarm intelligence optimization algorithm, the online self-evolution update of the power grid external damage detection model can be achieved without relying on cloud training or uploading original data. This enables the model to continuously adapt to the evolution of external damage characteristics caused by different geographical environments, construction types and seasonal changes, effectively solving the problems of traditional models being easy to become outdated and having poor generalization.

[0239] According to an embodiment of this application, a cable external damage detection system for implementing the above-described cable external damage detection method is also provided. Figure 5 and Figure 6 This is a structural block diagram of a cable external damage detection system provided according to an embodiment of this application, such as... Figure 5 and Figure 6 As shown, the cable external damage detection system includes a cloud server and multiple edge computing terminals. These edge computing terminals are installed at multiple monitoring points along the cable route. The first edge detection terminal among the multiple edge computing terminals includes:

[0240] The first acquisition module 100 is configured to acquire first vibration data of the cable through a first vibration sensor installed in the first edge computing terminal.

[0241] The generation module 200 is configured to generate a virtual reconstruction signal if the vibration amplitude of the first vibration data exceeds a preset amplitude threshold.

[0242] The sending module 300 is configured to send the virtual reconstruction signal to a second edge computing terminal adjacent to the first edge computing terminal, so that the second edge computing terminal can extract the second vibration data of the cable based on the virtual reconstruction signal. The second vibration data is the vibration data of the cable collected by the second vibration sensor of the second edge computing terminal, and the second vibration data corresponds to the timestamp of the first vibration data.

[0243] The receiving module 400 is configured to receive the second vibration data returned by the second edge computing terminal;

[0244] The virtual reconstruction module 500 is configured to perform virtual reconstruction based on the first vibration data and the second vibration data to obtain first reconstructed vibration data and second reconstructed vibration data.

[0245] The first cable external damage detection module 600 is configured to perform cable external damage detection based on the first reconstructed vibration data and the second reconstructed vibration data to obtain a first cable external damage detection result; the second cable external damage detection module 700 is configured to input the first reconstructed vibration data into a preset cable external damage detection model for detection if the first cable external damage detection result is a valid event to obtain a second cable external damage detection result, wherein the valid event indicates that there is a cable external damage.

[0246] It should be noted that the first acquisition module 100, generation module 200, transmission module 300, receiving module 400, virtual reconstruction module 500, first cable external damage detection module 600, and second cable external damage detection module 700 correspond to steps S201 to S207 in the embodiments. Multiple modules and their corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in the above embodiments. It should also be noted that the above modules, as part of the cable external damage detection system, can run on the computer terminal provided in the embodiments.

[0247] Embodiments of this application may provide a computer device. Optionally, in this embodiment, the computer device may be located in at least one of a plurality of network devices in a computer network. The computer device includes a memory and a processor.

[0248] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the cable external damage detection method and system in this embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned cable external damage detection method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to a computer terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0249] The processor can call the information and application program stored in the memory through the transmission device to perform the following steps: acquiring the first vibration data of the cable through the first vibration sensor set in the first edge computing terminal; if the vibration amplitude of the first vibration data exceeds the preset amplitude threshold, generating a virtual reconstruction signal; sending the virtual reconstruction signal to the second edge computing terminal to trigger it to extract the second vibration data of the cable; after receiving the second vibration data, performing virtual reconstruction based on the vibration data of both to obtain the corresponding reconstructed vibration data; performing preliminary detection on both to obtain the first cable external damage detection result; if the detection result is a valid event, inputting the first reconstructed vibration data into the preset cable external damage detection model and outputting the second cable external damage detection result.

[0250] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a non-volatile storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0251] Embodiments of this application also provide a non-volatile storage medium. Optionally, in this embodiment, the aforementioned non-volatile storage medium can be used to store the program code executed by the cable external damage detection method provided in the above embodiments.

[0252] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0253] Embodiments of this application also provide a computer program product, including a computer program. Optionally, in this embodiment, the computer program, when executed by a processor, can implement:

[0254] The first vibration data of the cable is collected by a first vibration sensor installed in the first edge computing terminal; if the vibration amplitude of the first vibration data exceeds a preset amplitude threshold, a virtual reconstruction signal is generated; the virtual reconstruction signal is sent to the second edge computing terminal to trigger it to extract the second vibration data of the cable; after receiving the second vibration data, virtual reconstruction is performed based on the vibration data of the two to obtain the corresponding reconstructed vibration data; the two are initially detected to obtain the first cable external damage detection result; if the detection result is a valid event, the first reconstructed vibration data is input into the preset cable external damage detection model and the second cable external damage detection result is output.

[0255] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0256] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0257] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0258] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0259] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0260] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0261] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for detecting external damage to cables, characterized in that, A first edge computing terminal is applied to a cable external damage detection system, the cable external damage detection system including a cloud server and multiple edge computing terminals, the multiple edge computing terminals being set at multiple cable monitoring points along the cable line, the method comprising: First vibration data of the cable is collected by a first vibration sensor installed in the first edge computing terminal; If the vibration amplitude of the first vibration data exceeds a preset amplitude threshold, a virtual reconstruction signal is generated; The virtual reconstruction signal is sent to a second edge computing terminal adjacent to the first edge computing terminal, so that the second edge computing terminal can extract the second vibration data of the cable based on the virtual reconstruction signal. The second vibration data is set as the vibration data of the cable collected by the second vibration sensor of the second edge computing terminal, and the second vibration data corresponds to the timestamp of the first vibration data. Receive the second vibration data returned by the second edge computing terminal; Virtual reconstruction is performed based on the first vibration data and the second vibration data to obtain first reconstructed vibration data and second reconstructed vibration data. Based on the first reconstructed vibration data and the second reconstructed vibration data, external cable damage detection is performed to obtain the first external cable damage detection result. If the first cable external damage detection result is a valid event, the first reconstructed vibration data is input into the preset cable external damage detection model for detection to obtain the second cable external damage detection result, wherein the valid event indicates that there is a cable external damage.

2. The method according to claim 1, characterized in that, Before acquiring the first vibration data through the first vibration sensor installed on the first edge computing terminal, the method further includes: Based on the positioning module of the first edge computing terminal, the clock module of the first edge computing terminal is initialized; The clock module is compensated based on the temperature data collected by the digital temperature sensor installed in the first edge computing terminal. After acquiring the first vibration data through the first vibration sensor installed on the first edge computing terminal, the method further includes: Based on the standard system time set in the clock module of the first edge computing terminal, a first timestamp is marked at the first sampling point of the first vibration data.

3. The method according to claim 1, characterized in that, If the vibration amplitude of the first vibration data exceeds a preset amplitude threshold, a virtual reconstruction signal is generated, including: If the vibration amplitude of multiple consecutive sampling points exceeds the preset amplitude threshold, a virtual reconstructed signal is generated; Sending the virtual reconstruction signal to a second edge computing terminal adjacent to the first edge computing terminal includes: The first identification information of the first edge computing terminal and the virtual reconstruction signal are sent to the second edge computing terminal.

4. The method according to claim 2, characterized in that, The virtual reconstruction based on the first vibration data and the second vibration data to obtain first reconstructed vibration data and second reconstructed vibration data includes: Based on the first timestamp of the first vibration data and the second timestamp of the second vibration data, determine the time deviation between the first timestamp and the second timestamp; Calculate the sampling rate deviation ratio between the first vibration data and the second vibration data; Based on the ratio of the time deviation to the sampling rate deviation, the second vibration data is reconstructed using a linear interpolation algorithm to obtain the second reconstructed vibration data. The first reconstructed vibration data is determined to be the first vibration data.

5. The method according to claim 1, characterized in that, The method of detecting cable external damage based on the first reconstructed vibration data and the second reconstructed vibration data to obtain the first cable external damage detection result includes: The first reconstructed vibration data and the second reconstructed vibration data are subjected to coarse interference filtering to obtain the cross-correlation coefficient between the first reconstructed vibration data and the second reconstructed vibration data. If the cross-correlation coefficient is greater than or equal to a preset coefficient threshold, then the first cable external damage detection result is determined to be a valid event; If the cross-correlation coefficient is less than the preset coefficient threshold, then the first cable external damage detection result is determined to be local interference.

6. The method according to claim 1, characterized in that, The step of inputting the first reconstructed vibration data into a preset cable external damage detection model for detection to obtain a second cable external damage detection result includes: The first reconstructed vibration data is input into the cable external damage detection model, and multidimensional features of the first reconstructed vibration data are extracted, wherein the multidimensional features include time domain features and frequency domain features; The multidimensional features are concatenated to obtain a multidimensional feature vector; Based on the multidimensional feature vector, the cable is subjected to external damage detection to obtain the external damage type and the corresponding external damage confidence level, wherein the external damage type includes at least one of excavation, drilling and impact.

7. The method according to claim 6, characterized in that, After obtaining the second cable external damage detection result, the method further includes: Based on the clock module installed in the first edge computing terminal, the location calculation of the first reconstructed vibration data and the second reconstructed vibration data is performed to obtain the geographical location of the external failure point; Multi-dimensional early warning data is generated based on the maximum vibration amplitude of the first vibration data, the second cable external damage detection result, and the geographical location of the external damage point, and the multi-dimensional early warning data is sent to the cloud server.

8. The method according to claim 7, characterized in that, The step of using a clock module located in the first edge computing terminal to perform location calculations on the first and second reconstructed vibration data to obtain the geographical location of the external failure point includes: Windowing and fast Fourier transform are applied to the first reconstructed vibration data and the second reconstructed vibration data to obtain first frequency domain data and second frequency domain data; Based on the subsampling-level time delay estimation algorithm of generalized cross-correlation and phase transformation weight, the weighted cross power spectral density of the first frequency domain data and the second frequency domain data is determined; The weighted cross-power spectral density is subjected to an inverse fast Fourier transform to obtain the generalized cross-correlation function in the time domain; A quadratic parabolic interpolation fitting is performed on the correlation peak of the generalized cross-correlation function and its left and right adjacent points to determine the precise position of the parabola vertex and obtain the time delay estimate. Based on the clock module, the temperature difference between the current ambient temperature and the calibrated temperature is obtained, and the propagation speed of the vibration wave in the cable is calibrated according to the temperature difference to obtain the calibrated propagation speed. Based on the calibration propagation speed and the estimated time delay, a distance equation is constructed between the external breaking point and the first edge computing terminal and the second edge computing terminal; Solve the distance equation to obtain the preliminary coordinates of the outer breach point, and use the Kalman filter algorithm to generate the optimal estimated coordinates of the preliminary coordinates to obtain the geographical location of the outer breach point.

9. The method according to claim 7, characterized in that, The method further includes: Receive cable correction instructions returned by the cloud server based on the multi-dimensional early warning data; Using the cable correction command as the label for the first reconstructed vibration data, training samples are generated; The cable external damage detection model is updated based on the training samples.

10. The method according to claim 9, characterized in that, The update of the cable external damage detection model based on the training samples includes: The training samples are stored in the transient learning library of the first edge computing terminal; If the number of new training samples in the transient learning library exceeds a preset threshold, or if the external damage confidence in the second cable external damage detection result is less than a preset confidence threshold, then the cable external damage detection model is self-evolved and trained using a lightweight particle swarm optimization algorithm to obtain an updated cable external damage detection model.

11. The method according to claim 10, characterized in that, If the number of new training samples in the transient learning library exceeds a preset threshold, or the external damage confidence score in the second cable external damage detection result is less than a preset confidence threshold, then the cable external damage detection model is self-evolved using a lightweight particle swarm optimization algorithm to obtain an updated cable external damage detection model, including: The hyperparameter encoding of the cable external damage detection model is used as the position vector of the lightweight particle swarm algorithm, and the fitness function of the lightweight particle swarm algorithm is determined. A chaotic sequence is generated using logistic regression mapping, and the chaotic sequence is mapped to the solution space of the particles to obtain an initial particle swarm. Load the candidate hyperparameters corresponding to each initial particle in the initial particle swarm, and input multiple training samples for training to obtain the candidate first cable external damage detection model; The first fitness value corresponding to the first cable external damage detection model is calculated using the fitness function; The initial particle swarm is updated to obtain an updated particle swarm; Load the candidate hyperparameters corresponding to each updated particle in the updated particle swarm, and input multiple training samples for training to obtain the candidate second cable external damage detection model; The fitness function is used to calculate the second fitness value corresponding to the second cable external damage detection model; The particle swarm and fitness values ​​are iteratively updated to obtain the globally optimal particle; The position vector of the globally optimal particle is decoded to obtain the optimal hyperparameter, and the candidate cable external damage detection model corresponding to the optimal hyperparameter is taken as the optimal cable external damage detection model.

12. A cable external damage detection system, characterized in that, It includes a cloud server and multiple edge computing terminals, which are installed at multiple cable monitoring points along the cable line. The first edge detection terminal among the multiple edge computing terminals includes: The first acquisition module is configured to acquire first vibration data of the cable through a first vibration sensor installed in the first edge computing terminal; The generation module is configured to generate a virtual reconstruction signal if the vibration amplitude of the first vibration data exceeds a preset amplitude threshold. The sending module is configured to send the virtual reconstruction signal to a second edge computing terminal adjacent to the first edge computing terminal, so that the second edge computing terminal can extract the second vibration data of the cable based on the virtual reconstruction signal, wherein the second vibration data is the vibration data of the cable collected by the second vibration sensor of the second edge computing terminal, and the second vibration data corresponds to the timestamp of the first vibration data; The receiving module is configured to receive the second vibration data returned by the second edge computing terminal; The virtual reconstruction module is configured to perform virtual reconstruction based on the first vibration data and the second vibration data to obtain first reconstructed vibration data and second reconstructed vibration data. The first cable external damage detection module is configured to perform cable external damage detection based on the first reconstructed vibration data and the second reconstructed vibration data, and obtain the first cable external damage detection result; The second cable external damage detection module is configured to input the first reconstructed vibration data into a preset cable external damage detection model for detection if the first cable external damage detection result is a valid event, thereby obtaining the second cable external damage detection result, wherein the valid event indicates that there is a cable external damage.

Images