Cable damage detection method and system based on wavelet scattering network

Abstract

The invention relates to a cable damage detection method and system based on a wavelet scattering network. The method comprises the following steps: collecting signal data generated in the working process of a to-be-detected cable, carrying out data preprocessing, and carrying out signal segmentation processing to obtain a plurality of signal segments; multi-scale features are extracted based on a wavelet scattering network, feature vectors are formed and input into an LSTM model for multi-scale feature classification, a classification result is obtained, and then a damage detection result is obtained. Data feature extraction is carried out through a wavelet scattering network, so that the influence of environmental noise and signal interference on cable damage detection can be reduced; by designing the LSTM-based classification model and classifying the multi-scale features extracted by the wavelet scattering network, the classification problem in the cable damage detection task is solved, the recognition precision of the damaged cable is improved, a solid foundation is provided for further cable state monitoring and maintenance, accurate detection of cable damage is realized, and the detection efficiency is improved. The method has high reliability and precision.

Description

Technical Field

[0001] The present invention relates to the technical field of cable detection, and in particular to a cable damage detection method and system based on wavelet scattering network. Background Art

[0002] Cable damage detection is an important part of maintaining the safety and reliability of cable systems. In the long-term use of cables, they may be damaged or fail due to environmental factors, mechanical stress, aging and other external interference. If these problems are not discovered and handled in time, they may cause serious safety accidents and economic losses. In China, cable damage detection technology has received widespread attention and rapid development. Traditional cable damage detection methods include DC high-voltage testing, AC high-voltage testing, reflection wave analysis (Time Domain Reflectometry, TDR), bridge method, infrared thermal imaging and acoustic detection. With the development of artificial intelligence technology, cable damage detection methods based on machine learning and deep learning have received widespread attention and application. For example, by training long short-term memory networks (LSTM) and recurrent neural networks (RNN), accurate identification and positioning of cable faults can be achieved. These methods use massive data and powerful computing power to achieve more efficient and accurate fault detection. Machine learning and deep learning methods can adapt to complex and changing environmental conditions by automatically learning features and patterns in data without human intervention, thereby improving the reliability and accuracy of detection.

[0003] However, traditional cable damage detection methods often have a lot of noise when identifying and locating cable faults, such as environmental noise, electromagnetic interference, etc., resulting in low accuracy and reliability of cable damage detection. Summary of the invention

[0004] Based on this, in order to solve the above technical problems, a cable damage detection method and system based on wavelet scattering network are provided, which can improve the detection accuracy and reliability of cable damage detection.

[0005] A cable damage detection method for a wavelet scattering network, the method comprising:

[0006] The sensor monitors the cable to be detected in real time and collects signal data generated during the operation of the cable to be detected;

[0007] Performing data preprocessing on the signal data, and performing signal segmentation processing on the preprocessed signal data to obtain a plurality of signal segments;

[0008] Performing signal analysis on the signal fragment based on a wavelet scattering network, extracting multi-scale features, and forming a feature vector according to the multi-scale features;

[0009] The feature vector is input into the LSTM model, and multi-scale feature classification is performed through the LSTM model to obtain a classification result;

[0010] A damage detection result of the cable to be detected is obtained based on the classification result.

[0011] In one embodiment, the real-time monitoring of the cable to be detected by a sensor and the collection of signal data generated during the operation of the cable to be detected include:

[0012] The vibration sensor is used to collect the vibration signal generated during the operation of the cable to be detected;

[0013] The acoustic wave sensor is used to collect the acoustic wave signal generated during the operation of the cable to be detected;

[0014] The electromagnetic field change signal generated during the operation of the cable to be detected is collected by an electromagnetic sensor.

[0015] In one embodiment, the vibration sensor, the acoustic wave sensor, and the electromagnetic sensor are respectively connected to a data acquisition device; the method further comprises:

[0016] Determine the data sampling frequency, sampling time window, and sampling interval by means of the data acquisition device;

[0017] The signal data generated during the working process of the cable to be detected is collected based on the data sampling frequency, the sampling time window and the sampling interval.

[0018] In one embodiment, the performing data preprocessing operation on the signal data includes:

[0019] Performing missing value processing on the signal data, and performing noise processing on the data after the missing value processing to obtain processed data;

[0020] A linear normalization operation is performed on the processed data to complete data preprocessing.

[0021] In one embodiment, the preprocessed signal data is subjected to signal segmentation processing to obtain a plurality of signal segments, including:

[0022] determining the frequency content and rate of change of the signal based on the preprocessed signal data;

[0023] Determine the signal segment length value and segment overlap rate according to the frequency component and the change rate;

[0024] The preprocessed signal data is segmented into signal segments of fixed length based on the signal segment length value and the segment overlap rate.

[0025] In one embodiment, the signal analysis is performed on the signal segment based on a wavelet scattering network to extract multi-scale features, and a feature vector is formed according to the multi-scale features, including:

[0026] Decomposing the signal segments layer by layer into sub-signals of different frequency bands;

[0027] Use wavelet transform, nonlinear transform and downsampling to transform and calculate the wavelet coefficients and scattering coefficients in the wavelet scattering network;

[0028] Extracting multi-scale features from each of the sub-signals based on the calculated wavelet coefficients and scattering coefficients;

[0029] The calculated backscattering coefficients are combined to form a feature vector according to the multi-scale features.

[0030] In one embodiment, the use of wavelet transform, nonlinear transform and downsampling to transform and calculate wavelet coefficients and scattering coefficients in a wavelet scattering network includes:

[0031] Perform modulus transformation on the wavelet coefficients, take the absolute value, and complete the nonlinear transformation;

[0032] Downsampling the wavelet coefficients after nonlinear transformation to remove redundant information and obtain the transformed wavelet coefficients;

[0033] The signal segments are subjected to wavelet transformation, nonlinear transformation and downsampling for several times to obtain scattering coefficients at various levels.

[0034] In one embodiment, the feature vector is input into an LSTM model, and multi-scale feature classification is performed by the LSTM model to obtain a classification result, including:

[0035] The feature vector passes through the input layer of the LSTM model and is captured in time series in the LSTM layer and the Dropout layer;

[0036] Feature classification is performed through the fully connected layer of the LSTM model, and the activation function is used to output the classification result.

[0037] In one of the embodiments, the method further includes: displaying the classification result.

[0038] A cable damage detection system based on a wavelet scattering network, the system comprising:

[0039] A data acquisition module, used to monitor the cable to be detected in real time through a sensor, and to collect signal data generated during the operation of the cable to be detected;

[0040] A data processing module, used for performing data preprocessing operation on the signal data, and performing signal segmentation processing on the preprocessed signal data to obtain a plurality of signal segments;

[0041] A feature extraction module, used to perform signal analysis on the signal fragment based on a wavelet scattering network, extract multi-scale features, and form a feature vector according to the multi-scale features;

[0042] A feature classification module is used to input the feature vector into the LSTM model, perform multi-scale feature classification through the LSTM model, and obtain a classification result;

[0043] A damage detection module is used to obtain a damage detection result of the cable to be detected based on the classification result.

[0044] The above-mentioned cable damage detection method and system based on wavelet scattering network extracts data features through wavelet scattering network, solves and reduces the influence of environmental noise and signal interference on cable damage detection, and solves the problem of insufficient feature extraction of traditional methods; by designing a classification model based on LSTM, the multi-scale features extracted by the wavelet scattering network are classified, which solves the classification problem in the cable damage detection task, and improves the recognition accuracy of damaged cables while ensuring the detection accuracy, providing a solid foundation for further cable status monitoring and maintenance, and realizing accurate detection of cable damage with strong reliability and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 FIG. 1 is an application environment diagram of a cable damage detection method based on a wavelet scattering network in an embodiment;

[0046] Figure 2 1 is a flow chart of a cable damage detection method based on a wavelet scattering network in one embodiment;

[0047] Figure 3 A schematic diagram of a network structure of wavelet transformation in one embodiment;

[0048] Figure 4 This is a schematic diagram of the change in training accuracy with the number of iterations during the experiment;

[0049] Figure 5 This is a diagram of the confusion matrix showing the classification effect on the test set during the experiment;

[0050] Figure 6 is a structural block diagram of a cable damage detection system based on a wavelet scattering network in one embodiment;

[0051] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] The cable damage detection method based on wavelet scattering network provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Figure 1 As shown, the application environment includes a computer device 110. The computer device 110 can monitor the cable to be detected in real time through sensors, and collect signal data generated during the working process of the cable to be detected; the computer device 110 performs data preprocessing operations on the signal data, and performs signal segmentation processing on the preprocessed signal data to obtain a number of signal segments; the computer device 110 performs signal analysis on the signal segments based on a wavelet scattering network, extracts multi-scale features, and forms a feature vector based on the multi-scale features; the computer device 110 inputs the feature vector into the LSTM model, and performs multi-scale feature classification through the LSTM model to obtain a classification result; the computer device 110 obtains the damage detection result of the cable to be detected based on the classification result. Among them, the computer device 110 can be, but is not limited to, various personal computers, laptops, robots, unmanned aerial vehicles and other devices.

[0054] In one embodiment, Figure 2 As shown, a cable damage detection method based on wavelet scattering network is provided, comprising the following steps:

[0055] Step 202: monitor the cable to be detected in real time through sensors, and collect signal data generated during the operation of the cable to be detected.

[0056] Computer equipment can use a variety of sensors to monitor the cable in real time and collect various signal data during the operation of the cable, which reflects the working status and potential fault information of the cable. Sensors can be installed at key locations of the cable, such as joints, terminals and places prone to mechanical stress concentration, to ensure the comprehensiveness and accuracy of the data.

[0057] Among them, the sensor may include a vibration sensor, an acoustic wave sensor, an electromagnetic sensor, etc.

[0058] In one embodiment, a cable damage detection method based on a wavelet scattering network is provided, which may also include a process of collecting data through sensors, and the specific process includes: collecting vibration signals generated during the operation of the cable to be detected through a vibration sensor; collecting sound wave signals generated during the operation of the cable to be detected through an acoustic wave sensor; and collecting electromagnetic field change signals generated during the operation of the cable to be detected through an electromagnetic sensor.

[0059] Specifically, in this embodiment, the vibration sensor is used to detect the mechanical vibration of the cable, and can capture the vibration signal caused by external impact or internal structural changes; the acoustic wave sensor is used to capture the acoustic wave signal inside the cable, which can reflect the defects or damage points inside the cable; the electromagnetic sensor is used to monitor the electromagnetic field changes of the cable, and can detect the damage of the cable insulation layer and current leakage. Among them, the sensor should be installed at the key position of the cable to ensure that it covers the entire operating area of ​​the cable and avoid interference of the installation position on signal acquisition.

[0060] In one embodiment, a provided cable damage detection method of a wavelet scattering network may also include a process of collecting data through a data acquisition device, and the specific process includes: connecting a vibration sensor, an acoustic wave sensor, and an electromagnetic sensor to the data acquisition device respectively; determining a data sampling frequency, a sampling time window, and a sampling interval through the data acquisition device; and collecting signal data generated during the working process of the cable to be detected based on the data sampling frequency, the sampling time window, and the sampling interval.

[0061] Specifically, in this embodiment, a data acquisition system may be provided on the computer device, and the vibration sensor, the acoustic wave sensor, and the electromagnetic sensor are respectively connected to the data acquisition device, and the data acquisition device may be a data recorder or a data acquisition card and other devices, which can realize real-time signal acquisition. Among them, the data acquisition device should have a high sampling rate and high resolution to ensure that the collected data has sufficient details and accuracy; the data acquisition system should also have data storage and transmission functions, and can store the collected data locally or transmit it to the data processing center in the computer device through the network.

[0062] In this embodiment, the data acquisition device can set an appropriate sampling frequency to capture all important features in the signal. The sampling frequency should be at least twice the highest frequency component in the signal (Nyquist sampling theorem); the time window and interval of data acquisition can also be determined to ensure the continuity and timeliness of the data. For a real-time monitoring system, a shorter time window and acquisition interval should be set to capture the state changes of the cable in a timely manner.

[0063] Step 204, performing data preprocessing on the signal data, and performing signal segmentation processing on the preprocessed signal data to obtain a plurality of signal segments.

[0064] When performing data preprocessing operations on signal data, operations such as data cleaning and normalization processing may be specifically included. Specifically, in one embodiment, a cable damage detection method of a wavelet scattering network provided may also include a process of performing data preprocessing operations, and the specific process may include: performing missing value processing on signal data, and performing noise processing on the data after the missing value processing to obtain processed data; performing linear normalization operations on the processed data to complete data preprocessing.

[0065] Among them, missing value processing refers to the process of data collection, which may cause some data to be missing due to sensor failure or signal transmission problems. Common missing value processing methods include: deleting missing values, filling missing values, and noise processing. Deleting missing values ​​means that for a small number of missing values, the records containing missing values ​​can be directly deleted; filling missing values ​​means that for a large number of missing values, interpolation methods are used to fill in and restore the integrity of the data; noise processing means that various noises may be mixed in the signal data, such as environmental noise, electromagnetic interference, etc. Wavelet transform is used to decompose the signal into components of different scales, retain the main features of the signal, and remove the noise components.

[0066] Linear normalization is to scale the signal data to a fixed range, such as 0 to 1 or -1 to 1, to eliminate the impact between different dimensions. The formula for linear normalization is: , X is the original data, Xmin and Xmax are the minimum and maximum values ​​of the data respectively.

[0067] In one embodiment, a provided cable damage detection method for a wavelet scattering network may also include a process of performing signal segmentation processing, the specific process including: determining the frequency component and change rate of the signal based on the preprocessed signal data; determining the signal segment length value and the segment overlap rate based on the frequency component and the change rate; and segmenting the preprocessed signal data into signal segments of fixed length based on the signal segment length value and the segment overlap rate.

[0068] Signal segmentation is to segment long continuous signals into short time segments of fixed length to facilitate subsequent feature extraction and analysis. When segmenting, it should be ensured that each segment contains complete feature information. Appropriate segment length and overlap rate can be set through computer equipment to ensure signal continuity and feature integrity. Generally speaking, the segment length should be determined based on the frequency component and change rate of the signal, and the overlap rate can be set to 50% to increase the amount of data and improve the stability of feature extraction. Through data collection and preprocessing steps, reliable quality and feature-rich cable operation data can be obtained, laying the foundation for subsequent feature extraction and cable damage detection.

[0069] Step 206 , performing signal analysis on the signal segment based on the wavelet scattering network, extracting multi-scale features, and forming a feature vector according to the multi-scale features.

[0070] Among them, using wavelet scattering network to extract multi-scale features of pre-processed signal data is an important step in cable damage detection. Wavelet scattering network can extract multi-scale features of signals through wavelet transform, nonlinear transformation and downsampling to form stable and noise-resistant feature vectors.

[0071] Wavelet transform is a powerful tool for time-frequency analysis of signals. It can decompose signals into components of different scales and frequency bands and extract multi-scale features of signals. The network structure of wavelet transform is as follows: Figure 3 As shown, represents Gaussian low-pass filtering, represents wavelet filtering (bandpass filtering), f represents input data, S[.] and U[.] represent scattering output operator and scattering propagation operator respectively, represents complex modulus, represents the empty set symbol, * represents linear convolution, represents Gaussian low-pass filtering, (i = 1, 2, ..., w) represents some specific complex wavelet, p = ( ) is the frequency reduction path.

[0072] In one embodiment, a cable damage detection method for a wavelet scattering network is provided, which may also include a process of extracting multi-scale features based on the wavelet scattering network, the specific process including: decomposing signal segments layer by layer into sub-signals of different frequency bands; transforming and calculating the wavelet coefficients and scattering coefficients in the wavelet scattering network using wavelet transform, nonlinear transform and downsampling; extracting multi-scale features from each sub-signal based on the calculated wavelet coefficients and scattering coefficients; combining the calculated scattering coefficients to form a feature vector based on the multi-scale features.

[0073] Specifically, in this embodiment, the specific process of wavelet transform may include: selecting wavelet basis function, multi-scale decomposition, and time-frequency positioning. In the process of selecting wavelet basis function, commonly used wavelet basis functions include Haar wavelet, Daubechies wavelet, Meyer wavelet, etc. Selecting wavelet basis function suitable for signal characteristics is the first step of wavelet transform; in the process of multi-scale decomposition, the signal is decomposed layer by layer into sub-signals of different frequency bands, and each layer of decomposition contains low-frequency components (approximate parts) and high-frequency components (detail parts). The low-frequency components retain the main characteristics of the signal, and the high-frequency components reflect the detailed changes of the signal; in the process of time-frequency positioning, wavelet transform can not only provide high-resolution analysis in the frequency domain, but also accurately locate the mutation point of the signal in the time domain, which is very useful for capturing the instantaneous changes of cable damage.

[0074] Assume that the original signal is x(t), and the result of wavelet transform is a set of wavelet coefficients Wj,k, where j represents the scale and k represents the position. When performing wavelet transform, the specific formula used is: ,in, is the wavelet basis function.

[0075] In one embodiment, a cable damage detection method for a wavelet scattering network is provided, which may also include a transformation calculation process, the specific process including: performing a modulus transformation on the wavelet coefficients, taking the absolute value, and completing a nonlinear transformation; performing downsampling processing on the wavelet coefficients after the nonlinear transformation, removing redundant information, and obtaining the transformed wavelet coefficients; performing a number of wavelet transformations, nonlinear transformations, and downsampling on the signal segments to obtain scattering coefficients at various levels.

[0076] In this embodiment, in order to improve the stability and noise resistance of the feature, the wavelet coefficients are nonlinearly transformed and downsampled. Specifically, the nonlinear transformation is to transform the wavelet coefficients modulus and take the absolute value to ensure the stability of the feature. The nonlinear transformation helps to suppress noise and make the feature insensitive to small changes in the signal. The nonlinear transformation formula is: .

[0077] Downsampling is to downsample the coefficients after nonlinear transformation to reduce the amount of data and retain important features. Downsampling can remove redundant information, reduce computational complexity, and retain the main features of the signal. The downsampling formula is: , where d represents the downsampling factor.

[0078] In this embodiment, the multi-layer wavelet scattering transform is to extract more in-depth features of the signal through multi-level convolution and nonlinear transformation. The specific steps are:

[0079] First-order scattering transform: The original signal is subjected to wavelet transform, nonlinear transform and downsampling to obtain the first-order scattering coefficient. The calculation formula of the first-order scattering coefficient S1 is: ;

[0080] Multi-level scattering transform: Take the first-level scattering coefficient as input, perform further wavelet transform, nonlinear transform and downsampling to obtain a higher level of scattering coefficient. The calculation formula of the second-level scattering coefficient S2 is: ; Among them, ∗ represents the convolution operation. Through multi-level scattering transformation, the deep features of the signal can be extracted to form a stable and noise-resistant feature vector;

[0081] Feature vector formation: Combine the multi-level scattering coefficients to form the final feature vector, which can be used as the input of the deep learning model for cable damage detection. The final feature vector S is calculated as: ; where n represents the number of layers of scattering transformation.

[0082] Through the steps of wavelet scattering transform, the multi-scale features of the signal can be extracted and a stable and noise-resistant feature vector can be formed, providing high-quality feature data for subsequent deep learning model training and cable damage detection.

[0083] Step 208: input the feature vector into the LSTM model, perform multi-scale feature classification through the LSTM model, and obtain a classification result.

[0084] In one embodiment, a provided cable damage detection method of a wavelet scattering network may also include a process of obtaining a classification result through an LSTM model, and the specific process includes: the feature vector passes through the input layer of the LSTM model, and time series is captured in the LSTM layer and the Dropout layer; feature classification is performed through the fully connected layer of the LSTM model, and the activation function output is used to obtain the classification result.

[0085] Among them, the detailed structure of the LSTM model and the parameter configuration of each layer are expressed as follows: Table 1: Detailed structure of the LSTM model and parameter configuration of each layer

[0086] Specifically, the LSTM model implements the classification of time series data through a multi-layer structure. First, the output shape of the input layer is (None, 1730, 1), which means that the time step of the input data is 1730, each time step has one feature, and None means that the batch size is not fixed.

[0087] Next, the first LSTM layer contains 64 units and the output shape is (None, 1730, 64), which means that a 64-dimensional feature vector is returned at each time step, and this layer has 16896 parameters. To prevent overfitting, a Dropout layer is added with an output shape of (None, 1730, 64) and a Dropout rate of 0.5, which means that 50% of the neurons are randomly discarded.

[0088] The second LSTM layer contains 32 units, the output shape is (None, 32), only the output of the last time step is returned, and the number of parameters is 12416. Add a Dropout layer again, the output shape is (None, 32), and the Dropout rate is also 0.5.

[0089] Finally, classification is performed through a fully connected layer (Dense layer), which contains 1 neuron, uses the Sigmoid activation function, and has an output shape of (None, 1), which is suitable for binary classification tasks. The total number of parameters is 33, and the total number of parameters of the entire model is 29345. Through this multi-layer structure, the LSTM model can effectively capture the long-term dependencies in time series data and improve classification performance.

[0090] Step 210: obtaining a damage detection result of the cable to be detected based on the classification result.

[0091] In one embodiment, a cable damage detection method using a wavelet scattering network may further include displaying a classification result.

[0092] In one embodiment, a process of experimental verification of a cable damage detection method using a wavelet scattering network is also provided. In this experiment, Figure 4 , Figure 5 As shown in the figure, wavelet scattering network and LSTM model are used to identify cable damage, where 1 represents damage. Figure 4 It shows the accuracy and loss changes of the model during training. Figure 4 The blue curve in the figure shows the change of training accuracy with the number of iterations. The accuracy rises rapidly in the early stage of training and stabilizes when it approaches 100%, indicating that the model performs very well on the training data and can almost correctly classify all samples. Figure 4 The orange curve in the figure shows the change of training loss with the number of iterations. The loss drops rapidly in the early stage of training and gradually stabilizes, indicating that the optimization process of the model is effective and the loss function is significantly minimized during the iteration process.

[0093] Figure 5 The confusion matrix shown further demonstrates the classification effect of the model on the test set. The overall accuracy is 97%, indicating that the overall classification effect of the model is good. Specifically, among the negative samples (i.e., no damage, category 0), 98.6% of the samples are correctly classified, and only 1.4% of the samples are misclassified as damaged; while among the positive samples (i.e., damaged, category 1), 86.0% of the samples are correctly classified, and 14.0% of the samples are misclassified as not damaged.

[0094] Therefore, the overall performance of the cable damage detection method based on a wavelet scattering network provided in this application on the training and test data is very good, with an overall accuracy of 94%.

[0095] The cable damage detection method of a wavelet scattering network provided in this application shows an accuracy rate of up to 97%. The accuracy rate increases rapidly and stabilizes during the training process, and the loss function decreases significantly and remains stable, indicating that the model performs very well on the training data. At the same time, the confusion matrix shows that the model has a classification accuracy of 98.6% for non-damaged samples (category 0) and a classification accuracy of 86.0% for damaged samples (category 1). Although the classification effect of damaged samples is relatively weak, the overall performance is still superior. By further optimizing the feature extraction method and balancing the number of data set samples, it can be expected that the classification effect of a cable damage detection method of a wavelet scattering network provided in this application in each category will be more balanced, and the overall performance will be further improved, showing the practical application potential of cable damage detection in dynamic and complex environments.

[0096] It should be understood that, although the various steps in the above-mentioned flow chart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above-mentioned flow chart may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0097] In one embodiment, Figure 6 As shown, a cable damage detection system of a wavelet scattering network is provided, comprising: a data acquisition module 610, a data processing module 620, a feature extraction module 630, a feature classification module 640 and a damage detection module 650, wherein:

[0098] The data acquisition module 610 is used to monitor the cable to be detected in real time through sensors, and to collect signal data generated during the operation of the cable to be detected;

[0099] The data processing module 620 is used to perform data preprocessing operations on the signal data, and perform signal segmentation processing on the preprocessed signal data to obtain a plurality of signal segments;

[0100] A feature extraction module 630 is used to perform signal analysis on the signal segment based on a wavelet scattering network, extract multi-scale features, and form a feature vector based on the multi-scale features;

[0101] The feature classification module 640 is used to input the feature vector into the LSTM model, perform multi-scale feature classification through the LSTM model, and obtain a classification result;

[0102] The damage detection module 650 is used to obtain a damage detection result of the cable to be detected based on the classification result.

[0103] In one embodiment, the data acquisition module 610 is also used to collect vibration signals generated during the operation of the cable to be detected through a vibration sensor; to collect sound wave signals generated during the operation of the cable to be detected through an sound wave sensor; and to collect electromagnetic field change signals generated during the operation of the cable to be detected through an electromagnetic sensor.

[0104] In one embodiment, the vibration sensor, the acoustic wave sensor, and the electromagnetic sensor are respectively connected to the data acquisition device; the data acquisition module 610 is also used to determine the data sampling frequency, the sampling time window, and the sampling interval through the data acquisition device; based on the data sampling frequency, the sampling time window, and the sampling interval, the signal data generated during the working process of the cable to be detected is collected.

[0105] In one embodiment, the data processing module 620 is also used to perform missing value processing on the signal data, and perform noise processing on the data after the missing value processing to obtain processed data; perform linear normalization operation on the processed data to complete data preprocessing.

[0106] In one embodiment, the data processing module 620 is also used to determine the frequency component and change rate of the signal based on the preprocessed signal data; determine the signal segment length value and segment overlap rate according to the frequency component and change rate; and segment the preprocessed signal data into signal segments of fixed length based on the signal segment length value and segment overlap rate.

[0107] In one embodiment, the feature extraction module 630 is also used to decompose the signal segments layer by layer into sub-signals of different frequency bands; use wavelet transform, nonlinear transform and downsampling to transform and calculate the wavelet coefficients and scattering coefficients in the wavelet scattering network; extract multi-scale features from each sub-signal based on the calculated wavelet coefficients and scattering coefficients; combine the calculated scattering coefficients to form a feature vector based on the multi-scale features.

[0108] In one embodiment, the feature extraction module 630 is also used to perform modulus transformation on the wavelet coefficients, take absolute values, and complete nonlinear transformation; downsample the wavelet coefficients after nonlinear transformation to remove redundant information and obtain transformed wavelet coefficients; perform wavelet transformation, nonlinear transformation and downsampling on the signal segment several times to obtain scattering coefficients at various levels.

[0109] In one embodiment, the feature classification module 640 is also used to pass the feature vector through the input layer of the LSTM model, capture the time series in the LSTM layer and the Dropout layer; perform feature classification through the fully connected layer of the LSTM model, and use the activation function output to obtain the classification result.

[0110] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a cable damage detection method based on a wavelet scattering network is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse, etc.

[0111] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0112] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of a cable damage detection method based on a wavelet scattering network when executing the computer program.

[0113] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of a cable damage detection method based on a wavelet scattering network are implemented.

[0114] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0115] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0116] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A cable damage detection method based on wavelet scattering network, characterized in that: include: The sensor monitors the cable to be detected in real time and collects signal data generated during the operation of the cable to be detected; Performing data preprocessing on the signal data, and performing signal segmentation processing on the preprocessed signal data to obtain a plurality of signal segments; Performing signal analysis on the signal fragment based on a wavelet scattering network, extracting multi-scale features, and forming a feature vector according to the multi-scale features; The feature vector is input into the LSTM model, and multi-scale feature classification is performed through the LSTM model to obtain a classification result; A damage detection result of the cable to be detected is obtained based on the classification result.

2. The cable damage detection method of wavelet scattering network according to claim 1 is characterized in that: The real-time monitoring of the cable to be detected by the sensor and the collection of signal data generated during the operation of the cable to be detected include: The vibration sensor is used to collect the vibration signal generated during the operation of the cable to be detected; The acoustic wave sensor is used to collect the acoustic wave signal generated during the operation of the cable to be detected; The electromagnetic field change signal generated during the operation of the cable to be detected is collected by an electromagnetic sensor.

3. The cable damage detection method of wavelet scattering network according to claim 2 is characterized in that: The vibration sensor, the acoustic wave sensor, and the electromagnetic sensor are respectively connected to a data acquisition device; the method further comprises: Determine the data sampling frequency, sampling time window, and sampling interval by means of the data acquisition device; The signal data generated during the working process of the cable to be detected is collected based on the data sampling frequency, the sampling time window and the sampling interval.

4. The cable damage detection method of wavelet scattering network according to claim 1, characterized in that: The performing of data preprocessing operation on the signal data comprises: Performing missing value processing on the signal data, and performing noise processing on the data after the missing value processing to obtain processed data; A linear normalization operation is performed on the processed data to complete data preprocessing.

5. The cable damage detection method of wavelet scattering network according to claim 1, characterized in that: The signal segmentation processing is performed on the pre-processed signal data to obtain a plurality of signal segments, including: determining the frequency content and rate of change of the signal based on the preprocessed signal data; Determine the signal segment length value and segment overlap rate according to the frequency component and the change rate; The preprocessed signal data is segmented into signal segments of fixed length based on the signal segment length value and the segment overlap rate.

6. The cable damage detection method of wavelet scattering network according to claim 1, characterized in that: The signal analysis is performed on the signal segment based on the wavelet scattering network to extract multi-scale features, and a feature vector is formed according to the multi-scale features, including: Decomposing the signal segments layer by layer into sub-signals of different frequency bands; Use wavelet transform, nonlinear transform and downsampling to transform and calculate the wavelet coefficients and scattering coefficients in the wavelet scattering network; Extracting multi-scale features from each of the sub-signals based on the calculated wavelet coefficients and scattering coefficients; The calculated backscattering coefficients are combined to form a feature vector according to the multi-scale features.

7. The cable damage detection method of wavelet scattering network according to claim 6, characterized in that: The method of using wavelet transform, nonlinear transform and downsampling to transform and calculate wavelet coefficients and scattering coefficients in the wavelet scattering network includes: Perform modulus transformation on the wavelet coefficients, take the absolute value, and complete the nonlinear transformation; Downsampling the wavelet coefficients after nonlinear transformation to remove redundant information and obtain the transformed wavelet coefficients; The signal segments are subjected to wavelet transformation, nonlinear transformation and downsampling for several times to obtain scattering coefficients at various levels.

8. The cable damage detection method of wavelet scattering network according to claim 1, characterized in that: The feature vector is input into the LSTM model, and multi-scale feature classification is performed by the LSTM model to obtain a classification result, including: The feature vector passes through the input layer of the LSTM model and is captured in time series in the LSTM layer and the Dropout layer; Feature classification is performed through the fully connected layer of the LSTM model, and the activation function is used to output the classification result.

9. The cable damage detection method of wavelet scattering network according to claim 1, characterized in that: The method also includes: displaying the classification result.

10. A cable damage detection system based on wavelet scattering network, characterized in that: The system comprises: A data acquisition module, used to monitor the cable to be detected in real time through a sensor, and to collect signal data generated during the operation of the cable to be detected; A data processing module, used for performing data preprocessing operation on the signal data, and performing signal segmentation processing on the preprocessed signal data to obtain a plurality of signal segments; A feature extraction module, used to perform signal analysis on the signal fragment based on a wavelet scattering network, extract multi-scale features, and form a feature vector according to the multi-scale features; A feature classification module is used to input the feature vector into the LSTM model, perform multi-scale feature classification through the LSTM model, and obtain a classification result; A damage detection module is used to obtain a damage detection result of the cable to be detected based on the classification result.

Images