A cable joint abnormality early warning method and system

Abstract

The present application relates to the technical field of cable joint diagnosis, and particularly relates to a cable joint abnormality early warning method and system, which comprises feature extraction on infrared thermal image data and acoustic vibration data to obtain a multimodal feature dataset; according to the multimodal feature dataset, a coupling correlation degree between a thermal spot diffusion gradient and a vibration frequency component is quantified to obtain a regional thermal-vibration coupling abnormality index; a regional defect risk score is calculated according to a regional defect type distribution; a modal fusion weight distribution value is obtained based on the regional defect risk score and the regional thermal-vibration coupling abnormality index; the infrared thermal image feature data and the acoustic vibration feature data are weighted and fused according to the modal fusion weight distribution value to obtain a regional defect coupling feature vector; a cable joint defect probability is obtained according to the regional defect coupling feature vector, and a hierarchical early warning is performed according to the cable joint defect probability. The present application realizes accurate early warning of cable joint abnormalities by fusing infrared thermal image data and acoustic vibration data.

Description

Technical Field

[0001] This invention relates to the field of cable joint diagnostic technology, and in particular to a method and system for early warning of cable joint abnormalities. Background Technology

[0002] As a core component of power transmission systems, cable joints are directly related to the safety and stability of the power grid. Especially in complex operating environments with high voltage and high load, the internal connection interface of cable joints is prone to failure due to poor contact or material aging. If these failures are not detected and dealt with in time, they may lead to serious power accidents and pose a huge threat to the normal operation of the power system. Therefore, how to effectively improve the identification accuracy of cable joint faults has become a key issue to ensure the reliable operation of the power system.

[0003] However, most current methods for monitoring the condition of cable joints rely on the analysis of a single data source, such as using only infrared thermography or acoustic vibration data for fault diagnosis. This single-data-source approach often fails to fully reflect the complex internal operating conditions of the joint. Due to the complex internal structure of cable joints and the diverse forms of fault manifestations, a single data source often cannot comprehensively and accurately reflect the true internal state of the joint. In practical applications, different areas of a cable joint have different structural and functional characteristics, such as the crimping area, the insulation layer covering area, and the outer sheath overlap area. When a fault occurs inside the joint, the characteristic responses of different areas also differ. For example, abnormal temperature distribution may manifest as irregular diffusion of hot spots on the joint surface. This diffusion further affects the axial temperature gradient change inside the crimping sleeve, exhibiting complex nonlinear characteristics. Acoustic vibration data may also show different characteristics depending on the location and type of the fault. A single data source cannot simultaneously capture these complex changing characteristics. In the identification of defects in multiple areas and with multiple structural types, misjudgments or omissions are prone to occur, thus limiting the accuracy and reliability of fault diagnosis.

[0004] In summary, existing cable joint condition monitoring technologies have many shortcomings in defect identification. Therefore, there is an urgent need to provide an intelligent diagnostic cable joint early warning method that can accurately capture the correlation characteristics of abnormal internal temperature and mechanical vibration in the joint, so as to meet the high requirements of power systems for the reliable operation of cable joints. Summary of the Invention

[0005] To address the above technical problems, this invention provides a method and system for early warning of abnormal cable joints.

[0006] In a first aspect, the present invention provides a method for early warning of abnormal cable joints, the method comprising the following steps:

[0007] Infrared thermal image data of the cable joint surface and acoustic vibration data generated by the cable joint vibration are collected in real time, and feature extraction is performed on the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset.

[0008] Based on the multimodal feature dataset, the coupling correlation between the hot spot diffusion gradient and vibration frequency components in different directions of the current infrared temperature distribution is quantified to obtain the regional thermal vibration coupling anomaly index of the cable joint.

[0009] Based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint, the distribution of regional defect types is identified, and the regional defect risk score of the cable joint in different regions is calculated based on the regional defect type distribution.

[0010] Based on the regional defect risk score and the regional thermal vibration coupling anomaly index, multimodal weight allocation is performed to obtain the modal fusion weight allocation value.

[0011] The infrared thermal image feature data and acoustic vibration feature data are weighted and fused according to the modal fusion weight allocation value to obtain the regional defect coupling feature vector of the cable joint.

[0012] The probability of cable joint defects is predicted based on the regional defect coupling feature vector, and a graded early warning is performed based on the probability of cable joint defects.

[0013] In a further implementation, the step of extracting features from the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset includes:

[0014] The infrared thermal image data is filtered to obtain infrared thermal image filtered data, and the gray values ​​in the infrared thermal image filtered data are converted into actual temperature values ​​to obtain the surface temperature field matrix of the cable joint.

[0015] The first-order gradient components of the surface temperature field matrix of the cable joint in the horizontal and vertical directions are calculated using the edge detection operator to obtain the temperature gradient components.

[0016] Temperature anomaly regions are identified based on the temperature gradient components, and thermal spatial feature vectors of the temperature anomaly regions are extracted.

[0017] The acoustic vibration data is subjected to a fast Fourier transform to obtain an acoustic energy spectrum, and the acoustic vibration frequency components whose amplitudes exceed a preset characteristic peak threshold are identified from the acoustic energy spectrum.

[0018] The thermal spatial feature vector and the acoustic vibration frequency component are aligned using timestamps, and the aligned thermal spatial feature vector and acoustic vibration frequency component are arranged in time sequence to form a multimodal feature dataset.

[0019] In a further embodiment, the thermal space feature vector includes at least the geometric center coordinates, boundary perimeter, area, and temperature peak of the temperature anomaly region.

[0020] In a further implementation, the step of quantifying the coupling correlation between the hot spot diffusion gradient and the vibration frequency component in different directions based on the multimodal feature dataset to obtain the regional thermal vibration coupling anomaly index of the cable joint includes:

[0021] Based on the analysis of the multimodal feature dataset, temperature gradient abrupt change points are identified, and the boundary contours of hot spot regions are determined.

[0022] The hotspot evolution is calculated based on the boundary contours of the hotspot regions at adjacent time points, and a hotspot diffusion gradient vector is constructed based on the hotspot evolution; the hotspot evolution includes at least the hotspot area change rate and the hotspot centroid displacement velocity;

[0023] The location of the hot spot centroid is determined based on the hot spot diffusion gradient vector, and the amplitude of the acoustic vibration frequency component corresponding to the location of the hot spot centroid is extracted from the multimodal feature dataset.

[0024] The Pearson correlation coefficient between the hot spot diffusion gradient vector and the amplitude of the acoustic vibration frequency component is calculated to obtain the temperature vibration correlation coefficient matrix.

[0025] Using the temperature vibration correlation coefficient matrix as the correlation weight, the hot spot diffusion gradient vector and the amplitude of the acoustic vibration frequency component are weighted and fused to obtain cross-modal fusion features.

[0026] Based on the cross-modal fusion characteristics, the spatial thermal-vibration coupling anomaly distribution is obtained by using the least squares linear regression algorithm, and the thermal-vibration coupling horizontal gradient component, thermal-vibration coupling vertical gradient component, and thermal-vibration coupling gradient amplitude are calculated according to the thermal-vibration coupling anomaly distribution between adjacent time points.

[0027] The regional thermal vibration coupling anomaly index of the cable joint is obtained by weighted summing of the horizontal gradient component of thermal vibration coupling, the vertical gradient component of thermal vibration coupling, and the magnitude of thermal vibration coupling.

[0028] In a further embodiment, the hot spot diffusion gradient vector includes a diffusion rate scalar value and a diffusion direction unit vector; wherein, the diffusion rate scalar value is the vector magnitude of the hot spot area change rate and the hot spot centroid displacement velocity; and the diffusion direction unit vector is a normalized vector of the hot spot centroid displacement direction.

[0029] In a further embodiment, the step of identifying the distribution of regional defect types based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint includes:

[0030] The spatial boundary coordinates of the conductor connection area, insulation layer area, and outer sheath area are divided according to the regional structural characteristic parameters of the cable joint.

[0031] The regional thermal vibration coupling anomaly index is mapped to the spatial boundary coordinates of the region according to its spatial location, and the arithmetic mean and maximum value of the regional thermal vibration coupling anomaly index in each region of the cable joint are calculated to identify the defective region of the cable joint.

[0032] Based on the mechanical resonance characteristics of the material, the characteristic frequency band corresponding to the defect area of ​​the cable joint is determined, and the characteristic frequency band of the defect area of ​​the cable joint is obtained.

[0033] Calculate the relative rate of change of frequency band energy of the characteristic frequency band of each cable joint defect area relative to the reference value to obtain the rate of change of vibration energy in the cable joint defect area.

[0034] Based on the rate of change of vibration energy, the regional defect types of each cable joint defect area are identified, forming a regional defect type distribution.

[0035] In a further embodiment, the step of calculating the regional defect risk score of the cable joint in different regions based on the regional defect type distribution includes:

[0036] Extract the regional defect types corresponding to conductor connection area, insulation layer area and outer sheath area from the regional defect type distribution, and obtain the defect duration and defect over-limit ratio corresponding to each regional defect type;

[0037] The proportion of the time the defect continues to exceed the limit to the total monitoring time is calculated, a time enhancement factor is generated, and the time enhancement factor is used to correct the proportion of the defect exceeding the limit to obtain an effective value exceeding the limit.

[0038] The regional defect score is calculated based on the effective excess value and the preset defect type basic weight coefficient.

[0039] When defects exist simultaneously in adjacent areas of a cable joint, the defect score of the area is corrected using a preset cross-area coupling coefficient to obtain a corrected area defect score.

[0040] The corrected regional defect scores are normalized to obtain the regional defect risk scores of cable joints in each region.

[0041] In a further implementation, the modal fusion weight allocation value includes a temperature modal weight allocation value and an acoustic modal weight allocation value. The step of performing multimodal weight allocation based on the regional defect risk score and the regional thermal vibration coupling anomaly index to obtain the modal fusion weight allocation value includes:

[0042] The scope of important monitoring areas for cable joints in each region is determined based on the regional defect risk score.

[0043] The information entropy of the temperature anomaly gradient and the amplitude of the acoustic vibration frequency component within the important monitoring area is calculated using the regional thermal vibration coupling anomaly index, and the corresponding temperature mode weight allocation value and acoustic mode weight allocation value are obtained.

[0044] In a further embodiment, the step of predicting the cable joint defect probability based on the regional defect coupling feature vector includes:

[0045] The region defect coupling feature vector is input into a random forest classifier that integrates multiple decision trees. Each decision tree classifies and judges the region defect coupling feature vector to obtain the defect probability votes of each decision tree.

[0046] The defect probability of the cable joint is obtained by selecting the highest number of votes for the defect probability from the defect probability votes of each decision tree, calculating the proportion of the highest number of votes for the defect to the total number of decision trees.

[0047] Secondly, the present invention provides a cable joint abnormality early warning system, the system comprising:

[0048] The feature extraction module is used to collect infrared thermal image data of the cable joint surface and acoustic vibration data generated by the cable joint vibration in real time, and to extract features from the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset.

[0049] The temperature analysis module is used to quantify the coupling correlation between the hot spot diffusion gradient and vibration frequency components in different directions of the current infrared temperature distribution based on the multimodal feature dataset, and to obtain the regional thermal vibration coupling anomaly index of the cable joint.

[0050] The defect analysis module is used to identify the distribution of regional defect types based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint, and to calculate the regional defect risk score of the cable joint in different regions based on the regional defect type distribution.

[0051] The weight allocation module is used to perform multimodal weight allocation based on the regional defect risk score and the regional thermal vibration coupling anomaly index to obtain the modal fusion weight allocation value.

[0052] The modal fusion module is used to perform weighted fusion of infrared thermal image feature data and acoustic vibration feature data according to the modal fusion weight allocation value to obtain the regional defect coupling feature vector of the cable joint;

[0053] The defect early warning module is used to predict the probability of cable joint defects based on the regional defect coupling feature vector, and to provide graded early warnings based on the probability of cable joint defects.

[0054] This invention provides a method and system for early warning of cable joint anomalies. The method involves real-time acquisition of infrared thermographic data of the cable joint surface and acoustic vibration data generated by the cable joint vibration, and feature extraction of the infrared thermographic data and acoustic vibration data to obtain a multimodal feature dataset. Based on the multimodal feature dataset, the coupling correlation between the hot spot diffusion gradient and vibration frequency components of the current infrared temperature distribution in different directions is quantified to obtain a regional thermal-vibration coupling anomaly index of the cable joint. The distribution of regional defect types is identified based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint, and a regional defect risk score for the cable joint in different regions is calculated based on the regional defect type distribution. Multimodal weight allocation is performed based on the regional defect risk score to obtain a modal fusion weight allocation value. The infrared thermographic feature data and acoustic vibration feature data are weighted and fused according to the modal fusion weight allocation value to obtain a regional defect coupling feature vector of the cable joint. The probability of cable joint defects is predicted based on the regional defect coupling feature vector, and a graded early warning is performed based on the cable joint defect probability. Compared with existing technologies, this method achieves multimodal feature weighted fusion and defect probability prediction by fusing infrared thermography and acoustic vibration data, thus enabling accurate early warning of cable joint anomalies and effectively improving the accuracy of cable joint anomaly detection and the timeliness of early warning. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the cable joint abnormality early warning method provided in an embodiment of the present invention;

[0056] Figure 2 This is a block diagram of the cable joint abnormality early warning system provided in an embodiment of the present invention.

[0057] Figure labeling: 101, Feature extraction module; 102, Temperature analysis module; 103, Defect analysis module; 104, Weight allocation module; 105, Modal fusion module; 106, Defect early warning module. Detailed Implementation

[0058] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. The embodiments are given for illustrative purposes only and should not be construed as limiting the present invention. The accompanying drawings are for reference and illustration only and do not constitute a limitation on the scope of patent protection of the present invention, because many changes can be made to the present invention without departing from the spirit and scope of the present invention.

[0059] refer to Figure 1 This invention provides a method for early warning of cable joint abnormalities, such as... Figure 1 As shown, the method includes the following steps:

[0060] S1. Real-time acquisition of infrared thermal image data of the cable joint surface and acoustic vibration data generated by the cable joint vibration, and feature extraction of the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset.

[0061] In some embodiments, the step of extracting features from the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset includes:

[0062] The infrared thermal image data is filtered to obtain infrared thermal image filtered data, and the gray values ​​in the infrared thermal image filtered data are converted into actual temperature values ​​to obtain the surface temperature field matrix of the cable joint.

[0063] The first-order gradient components of the surface temperature field matrix of the cable joint in the horizontal and vertical directions are calculated using the edge detection operator to obtain the temperature gradient components.

[0064] Temperature anomaly regions are identified based on the temperature gradient components, and thermal spatial feature vectors of the temperature anomaly regions are extracted.

[0065] The acoustic vibration data is subjected to a fast Fourier transform to obtain an acoustic energy spectrum, and the acoustic vibration frequency components whose amplitudes exceed a preset characteristic peak threshold are identified from the acoustic energy spectrum.

[0066] The thermal spatial feature vector and the acoustic vibration frequency component are aligned using timestamps, and the aligned thermal spatial feature vector and acoustic vibration frequency component are arranged in time sequence to form a multimodal feature dataset.

[0067] Specifically, this embodiment uses a high-resolution infrared thermal imager to scan the surface of the cable joint in real time. The resolution of the infrared thermal imager can be set to 640×480 pixels, and the scanning frequency is 30Hz to ensure accurate capture of minute temperature changes on the surface of the cable joint. In this embodiment, the infrared thermal imager is installed close to the cable joint to ensure that its field of view can completely cover the cable joint and a certain area around it, avoiding data loss due to obstruction of the view. The infrared thermal imager collects infrared radiation data from the surface of the cable joint in a non-contact manner and converts it into digital signals, which are then stored as raw infrared thermal image data. At the same time, this embodiment uses a high-sensitivity acoustic sensor installed on the outer shell of the cable joint or on the structure of the equipment connected to the cable joint. The frequency response range of the high-sensitivity acoustic sensor should cover the main frequency range of the cable joint vibration to ensure stable and reliable acquisition of the acoustic vibration signals generated by the cable joint vibration. The sampling frequency of the acoustic sensor can be set to 20kHz to ensure capture of high-frequency vibration signals. The collected acoustic vibration signals are converted into digital signals by a data acquisition card and then stored as raw acoustic vibration data.

[0068] This embodiment uses a Gaussian filtering algorithm to denoise the raw infrared thermal image data, eliminating the influence of environmental noise and sensor noise, resulting in filtered infrared thermal image data. The Gaussian filter kernel size can be set to 5×5 pixels with a standard deviation of 1.5. Then, based on the calibration curve of the infrared thermal imager (pre-calibrated using a blackbody radiation source), the grayscale value of each pixel in the filtered infrared thermal image data is converted into an actual temperature value. The converted data forms a temperature field matrix for the cable joint surface. The dimension of the temperature field matrix is ​​consistent with the image size of the infrared thermal image data. Each element of the temperature field matrix corresponds to the temperature value of a pixel on the cable joint surface. This embodiment uses the Sobel edge detection operator to perform convolution operations on the temperature field matrix for the cable joint surface, calculating the first-order values ​​in the horizontal (x-direction) and vertical (y-direction) directions respectively. The gradient components are obtained by calculating the horizontal and vertical gradient components, and then taking the square root of the sum of the squares of the horizontal and vertical gradient components. This yields the temperature gradient amplitude, which represents the degree of temperature change on the surface of the cable joint. In this embodiment, the temperature gradient amplitude is used as the temperature gradient component, and a connected component analysis algorithm is used to segment the temperature gradient component. Regions with temperature gradient amplitudes exceeding a preset threshold are marked as temperature anomaly regions. For each temperature anomaly region, thermal space features such as geometric center coordinates, boundary perimeter, peak temperature, average temperature, mean temperature gradient amplitude, variance, and area (number of pixels) are extracted to form a thermal space feature vector.

[0069] Simultaneously, this embodiment applies Hanning windowing to the acoustic vibration data, converts the acoustic vibration data into a frequency domain signal using Fast Fourier Transform (FFT), and calculates the squared amplitude of the FFT result to obtain an acoustic energy spectrum. The acoustic energy spectrum represents the energy distribution of different frequency components. Then, it identifies frequency points in the acoustic energy spectrum whose amplitude exceeds a preset characteristic peak threshold. The vibration center frequency and normalized vibration energy ratio corresponding to these frequency points are used as acoustic vibration characteristic frequency components. In this embodiment, the preset characteristic peak threshold can be set to 50% of the maximum amplitude of the acoustic energy spectrum. Finally, this embodiment adds a timestamp to each thermal space feature vector and acoustic vibration frequency component, and uses nearest neighbor interpolation to align the thermal space feature vector and acoustic vibration frequency component to the same time reference. The aligned thermal space feature vector and acoustic vibration frequency component are arranged in chronological order to form a multimodal feature dataset.

[0070] S2. Based on the multimodal feature dataset, quantify the coupling correlation between the hot spot diffusion gradient and vibration frequency components in different directions of the current infrared temperature distribution to obtain the regional thermal vibration coupling anomaly index of the cable joint.

[0071] In some implementations, the step of quantifying the coupling correlation between the hot spot diffusion gradient and the vibration frequency components in different directions based on the multimodal feature dataset to obtain the regional thermal vibration coupling anomaly index of the cable joint includes:

[0072] Based on the analysis of the multimodal feature dataset, temperature gradient abrupt change points are identified, and the boundary contours of hot spot regions are determined.

[0073] The hotspot evolution is calculated based on the boundary contours of the hotspot regions at adjacent time points, and a hotspot diffusion gradient vector is constructed based on the hotspot evolution; the hotspot evolution includes at least the hotspot area change rate and the hotspot centroid displacement velocity;

[0074] The location of the hot spot centroid is determined based on the hot spot diffusion gradient vector, and the amplitude of the acoustic vibration frequency component corresponding to the location of the hot spot centroid is extracted from the multimodal feature dataset.

[0075] The Pearson correlation coefficient between the hot spot diffusion gradient vector and the amplitude of the acoustic vibration frequency component is calculated to obtain the temperature vibration correlation coefficient matrix.

[0076] Using the temperature vibration correlation coefficient matrix as the correlation weight, the hot spot diffusion gradient vector and the amplitude of the acoustic vibration frequency component are weighted and fused to obtain cross-modal fusion features.

[0077] Based on the cross-modal fusion characteristics, the spatial thermal-vibration coupling anomaly distribution is obtained by using the least squares linear regression algorithm, and the thermal-vibration coupling horizontal gradient component, thermal-vibration coupling vertical gradient component, and thermal-vibration coupling gradient amplitude are calculated according to the thermal-vibration coupling anomaly distribution between adjacent time points.

[0078] The regional thermal vibration coupling anomaly index of the cable joint is obtained by weighted summing of the horizontal gradient component of thermal vibration coupling, the vertical gradient component of thermal vibration coupling, and the magnitude of thermal vibration coupling.

[0079] Specifically, this embodiment extracts the temperature gradient component at each time point from the multimodal feature dataset and performs second-order difference operations on the temperature gradient components. Points exceeding a preset second-order difference threshold are marked as temperature gradient abrupt change points. These temperature gradient abrupt change points reflect abnormal changes at the edge of the hot spot or on the heat conduction path. Then, this embodiment uses a region growing algorithm, with the temperature gradient abrupt change points as seed points, to grow regions based on the similarity of temperature gradient values. Starting from the seed points, according to certain similarity criteria (such as temperature gradient changes within a certain range), adjacent pixels are gradually grouped into the same hot spot region until no more pixels that meet the conditions can be found. The Canny edge detection algorithm is then used to perform edge detection on the grown region to obtain the hot spot. For each hotspot region, this embodiment obtains the area at adjacent time points by counting the number of pixels within the boundary contour of the hotspot region. The area change rate is obtained by calculating the ratio of the area difference between the current and next time points to the area of ​​the hotspot region at the current time point. Simultaneously, the displacement velocity of the hotspot centroid is calculated based on the ratio of the displacement distance of the hotspot centroid to the time interval between adjacent time points. The position of the hotspot centroid can be obtained by averaging the coordinates of all pixels within the hotspot region. This embodiment obtains the diffusion rate scalar value by calculating the vector magnitude of the hotspot area change rate and the hotspot centroid displacement velocity, and determines the direction of the hotspot centroid displacement, using this direction as the diffusion direction. The displacement vector is normalized by dividing it by its magnitude to obtain a unit vector for the diffusion direction. The diffusion rate scalar value and the unit vector for the diffusion direction are combined to form the hotspot diffusion gradient vector, which describes the diffusion of the hotspot in space.

[0080] This embodiment updates the centroid coordinates of the hot spot based on the unit vector of the diffusion direction and the scalar value of the diffusion rate in the hot spot diffusion gradient vector. Within each time step (e.g., 1 second), the current centroid coordinates are moved along the direction of the unit vector of diffusion by a distance corresponding to the scalar value of the diffusion rate, resulting in new centroid coordinates. Then, the acoustic vibration frequency component corresponding to the current time point is found from the multimodal feature dataset, and the amplitude of the acoustic vibration frequency component closest to or corresponding to the new hot spot centroid is extracted from the acoustic vibration frequency component. For statistical purposes, this embodiment can average the amplitudes of the acoustic vibration frequency components within a small region to obtain the amplitude representing the acoustic vibration frequency component corresponding to the hot spot centroid position. It should be noted that the Pearson correlation coefficient is used to measure the linear correlation between two variables. For the hot spot diffusion gradient vector and the amplitude of the acoustic vibration frequency component, this embodiment considers the hot spot diffusion separately. The correlation between the diffusion rate scalar value and the diffusion direction unit vector of the gradient vector and the amplitude of the acoustic vibration frequency component is as follows: For example, for the diffusion rate scalar value sequence and the acoustic vibration frequency component amplitude sequence, this embodiment first calculates the mean of the two sequences, and then calculates the covariance and standard deviation based on the mean of the two sequences. Finally, the covariance of the diffusion rate scalar value and the acoustic vibration frequency component amplitude is divided by the product of their standard deviations to obtain the Pearson correlation coefficient between the diffusion rate scalar value sequence and the acoustic vibration frequency component amplitude sequence. Repeating the above steps, this embodiment calculates the Pearson correlation coefficients between the diffusion rate scalar value and the acoustic vibration frequency component amplitude, the horizontal component of the diffusion direction and the acoustic vibration frequency component amplitude, and the vertical component of the diffusion direction and the acoustic vibration frequency component amplitude, respectively, thereby obtaining the temperature vibration correlation coefficient matrix.

[0081] For each hot spot, this embodiment uses elements in the temperature-vibration correlation coefficient matrix as correlation weights to perform a weighted summation of the hot spot diffusion gradient vector and the acoustic vibration frequency component amplitude. This embodiment combines the hot spot diffusion gradient information and the acoustic vibration frequency component information through weighted fusion to obtain a cross-modal fusion feature. This cross-modal fusion feature reflects the coupling relationship between hot spot diffusion and vibration. Next, this embodiment performs spatial gridding of the cross-modal fusion feature, with each grid point corresponding to a cross-modal fusion feature vector. The least squares linear regression algorithm is used to fit the cross-modal fusion feature at the grid point to obtain the spatial thermal-vibration coupling anomaly distribution. This embodiment focuses on the spatial thermal-vibration coupling anomaly distribution with respect to the horizontal direction. The partial derivatives (in the x-direction) are calculated to obtain the horizontal gradient component of thermal vibration coupling. Simultaneously, the partial derivatives of the spatial thermal vibration coupling anomaly distribution with respect to the vertical direction (y-direction) are calculated to obtain the vertical gradient component of thermal vibration coupling. The square root of the sum of the squares of the horizontal and vertical gradient components of thermal vibration coupling is then calculated, i.e., the square root of the sum of the squares of the horizontal and vertical gradient components of thermal vibration coupling, to obtain the amplitude of the thermal vibration coupling gradient. Finally, in this embodiment, a weighted summation method is used to integrate the horizontal, vertical, and amplitudes of the thermal vibration coupling to obtain the regional thermal vibration coupling anomaly index of the cable joint. The regional thermal vibration coupling anomaly index can be used to assess the degree of anomaly in the regional thermal vibration coupling of the cable joint.

[0082] S3. Identify the distribution of regional defect types based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint, and calculate the regional defect risk score of the cable joint in different regions based on the regional defect type distribution.

[0083] In some embodiments, the step of identifying the distribution of regional defect types based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint includes:

[0084] The spatial boundary coordinates of the conductor connection area, insulation layer area, and outer sheath area are divided according to the regional structural characteristic parameters of the cable joint.

[0085] The regional thermal vibration coupling anomaly index is mapped to the spatial boundary coordinates of the region according to its spatial location, and the arithmetic mean and maximum value of the regional thermal vibration coupling anomaly index in each region of the cable joint are calculated to identify the defective region of the cable joint.

[0086] Based on the mechanical resonance characteristics of the material, the characteristic frequency band corresponding to the defect area of ​​the cable joint is determined, and the characteristic frequency band of the defect area of ​​the cable joint is obtained.

[0087] Calculate the relative rate of change of frequency band energy of the characteristic frequency band of each cable joint defect area relative to the reference value to obtain the rate of change of vibration energy in the cable joint defect area.

[0088] Based on the rate of change of vibration energy, the regional defect types of each cable joint defect area are identified, forming a regional defect type distribution.

[0089] Specifically, this embodiment obtains the regional structural characteristic parameters of the cable joint from the structural parameter dataset of the cable joint. These parameters may include conductor cross-sectional area, insulation layer thickness, and outer sheath material density. Based on these parameters, the surface of the cable joint is divided into a conductor connection region, an insulation layer region, and an outer sheath region. The conductor connection region corresponds to the cable core crimping area, and its spatial boundary is determined by the coordinates of the metal shielding layer. The insulation layer region covers the surface of the insulating medium, and its boundary is the junction line between the outer semiconducting layer and the sheath. The outer sheath region is the outermost protective structure, and its boundary is defined by the geometric dimensions of the joint shell. This determines the regional spatial boundary coordinates of the conductor connection region, insulation layer region, and outer sheath region. In this embodiment, the cable joint is spatially gridded, with each grid point corresponding to a spatial coordinate. The grid points are then divided according to the regional boundaries. The system is divided into conductor connection area, insulation layer area, and outer sheath area. In this embodiment, the regional thermal vibration coupling anomaly index is mapped to the regional spatial boundary coordinates according to its pixel spatial position coordinates. For example, for each grid point, its regional thermal vibration coupling anomaly index value is assigned to the corresponding area. The regional thermal vibration coupling anomaly index in each area (conductor connection area, insulation layer area, and outer sheath area) is statistically analyzed, and the arithmetic mean and maximum value of the regional thermal vibration coupling anomaly index in each area are calculated to obtain the regional average value and regional maximum value of the regional thermal vibration coupling anomaly index in each area. Based on the regional average value and regional maximum value, the cable joint defect area is identified. For example, in this embodiment, a coupling anomaly threshold can be set. When the regional average value or regional maximum value of the regional thermal vibration coupling anomaly index of a certain area exceeds the coupling anomaly threshold, the area is considered to be a defect area.

[0090] Simultaneously, this embodiment, based on the material type of the candidate defect area (such as copper conductor, XLPE insulation, or PVC sheath), calls a preset material mechanical resonance characteristic database to obtain the material mechanical resonance characteristic parameters of each region of the cable joint (conductor connection region, insulation layer region, and outer sheath region) from the database. These parameters include natural frequency, damping ratio, etc. Based on the material mechanical resonance characteristics, the characteristic frequency band corresponding to the cable joint defect area is determined. For example, the material mechanical resonance characteristics of the defect area may change, causing its natural frequency to shift. By analyzing the material mechanical resonance characteristics of the cable joint defect area, its characteristic frequency band can be determined. Assuming the natural frequency of the normal region is... The inherent frequency shift of the defect region is Then the characteristic frequency band is This embodiment performs spectral analysis on the acoustic vibration signal of the cable joint, integrating or summing the frequency components in the characteristic frequency band to obtain the frequency band energy of each defect area. This embodiment calculates the relative rate of change of the frequency band energy of the defect area's characteristic frequency band compared to the baseline energy value of the same area under historical normal conditions, obtaining the vibration energy change rate. Based on the magnitude of the vibration energy change rate, the defect type of the defect area is identified. For example, this embodiment sets different vibration energy change rate ranges corresponding to different defect types. A vibration energy change rate greater than 200% in the conductor connection area is identified as a poor contact defect; a vibration energy change rate greater than 150% in the insulation layer area is identified as an insulation degradation defect; and a vibration energy change rate greater than 120% in the outer sheath area is identified as a mechanical loosening defect. This embodiment maps the identified defect types onto the spatial area of ​​the cable joint and outputs them in the form of spatial coordinate mapping, forming a regional defect type distribution containing defect type labels. For areas that simultaneously meet multiple defect criteria, they are classified according to the type corresponding to the maximum energy change rate.

[0091] In some embodiments, the step of calculating the regional defect risk score of the cable joint in different regions based on the regional defect type distribution includes:

[0092] Extract the regional defect types corresponding to conductor connection area, insulation layer area and outer sheath area from the regional defect type distribution, and obtain the defect duration and defect over-limit ratio corresponding to each regional defect type;

[0093] The proportion of the time the defect continues to exceed the limit to the total monitoring time is calculated, a time enhancement factor is generated, and the time enhancement factor is used to correct the proportion of the defect exceeding the limit to obtain an effective value exceeding the limit.

[0094] The regional defect score is calculated based on the effective excess value and the preset defect type basic weight coefficient.

[0095] When defects exist simultaneously in adjacent areas of a cable joint, the defect score of the area is corrected using a preset cross-area coupling coefficient to obtain a corrected area defect score.

[0096] The corrected regional defect scores are normalized to obtain the regional defect risk scores of cable joints in each region.

[0097] Specifically, this embodiment extracts the regional defect types corresponding to conductor connection areas, insulation layer areas, and outer sheath areas from the regional defect type distribution dataset. For each area, the duration of continuous existence of its defect state exceeding a preset time threshold is counted to obtain the defect duration exceeding the limit for each regional defect type. Simultaneously, this embodiment calculates the defect exceedance ratio for each area. Assuming that each area is monitored N times within the monitoring period, and n times defect exceedance is detected, the defect exceedance ratio is... For example, if the conductor connection area is monitored 100 times and the number of times defects exceeding the limit are detected is 30 times, then the defect exceeding the limit ratio is 0.3.

[0098] For each region, this embodiment calculates the square of the ratio of its defect duration exceeding the limit to the total monitoring time, generating a time enhancement factor. The time enhancement factor reflects the relative importance of the defect duration; the longer the duration, the larger the time enhancement factor. The time enhancement factor is used to correct the defect exceeding the limit ratio. Multiplying the time enhancement factor and the defect exceeding the limit ratio yields the effective exceeding value. When the effective exceeding value exceeds 100%, it is forcibly set to 100%. Based on historical experience and the assessment of defect severity in each region of the cable joint, this embodiment sets basic weight coefficients for different defect types. The effective exceeding value of each region is multiplied by the corresponding basic weight coefficient for the defect type to calculate the regional defect score for each region. Simultaneously, this embodiment checks for adjacency relationships between regions of the cable joint and determines whether adjacent regions have defects. For example, the conductor connection region and... When adjacent insulation regions have defects, and defects also exist in adjacent areas of the cable joint, a cross-regional coupling coefficient is set based on the degree of coupling between the regions. For example, the cross-regional coupling coefficient between the conductor connection region and the insulation region is 0.5; the cross-regional coupling coefficient between the insulation region and the outer sheath region is 0.4. For regions with adjacent defects, the regional defect score is corrected using the cross-regional coupling coefficient. The regional defect score is calculated by adding the product of the cross-regional coupling coefficient and the regional defect score to obtain the corrected regional defect score. The cross-regional coupling coefficient reflects the degree of mutual influence between defects in adjacent regions. The larger the coupling coefficient, the greater the correction magnitude. The corrected regional defect score is then normalized to a range between [0, 1], where 0 represents no risk and 1 represents extremely high risk.

[0099] S4. Based on the regional defect risk score and the regional thermal vibration coupling anomaly index, perform multimodal weight allocation to obtain the modal fusion weight allocation value.

[0100] In some implementations, the modal fusion weight allocation value includes a temperature modal weight allocation value and an acoustic modal weight allocation value. The step of performing multimodal weight allocation based on the regional defect risk score and the regional thermal vibration coupling anomaly index to obtain the modal fusion weight allocation value includes:

[0101] The scope of important monitoring areas for cable joints in each region is determined based on the regional defect risk score.

[0102] The information entropy of the temperature anomaly gradient and the amplitude of the acoustic vibration frequency component within the important monitoring area is calculated using the regional thermal vibration coupling anomaly index, and the corresponding temperature mode weight allocation value and acoustic mode weight allocation value are obtained.

[0103] Specifically, this embodiment traverses all areas of the cable joint and determines the important monitoring area range for each area based on the regional defect risk score. For example, this embodiment sets a risk score threshold. When the regional defect risk score of a certain area exceeds the risk score threshold, the area is considered an important monitoring area. Furthermore, when the regional thermal vibration coupling anomaly index of a certain area exceeds the upper limit of the historical normal fluctuation range, the coordinates of that area are added to the important monitoring area. This embodiment compresses the regional thermal vibration coupling anomaly index of all pixels within the key monitoring area to a range of 0 to 1. The minimum regional thermal vibration coupling anomaly index is mapped to 0, the maximum regional thermal vibration coupling anomaly index is mapped to 1, and the values ​​of the intermediate regional thermal vibration coupling anomaly index are linearly scaled to obtain a normalized regional thermal vibration coupling anomaly index. This embodiment calculates the arithmetic mean of the normalized regional thermal vibration coupling anomaly index as the global coupling strength coefficient. This embodiment takes 1 minus the global coupling strength coefficient as the value. The entropy correction coefficient is used to suppress the influence of entropy values. In this embodiment, the distribution of temperature anomaly gradient directions of all pixels within a key monitoring area is statistically analyzed. The proportion of pixels in each directional partition to the total number of pixels is calculated to obtain the probability distribution of temperature anomaly gradients in different directions. Based on this probability distribution, the original temperature gradient information entropy of the conductor connection area is calculated. The original temperature gradient information entropy is equal to the sum of the probability distribution multiplied by the negative of the base-2 logarithm of the corresponding probability distribution. A larger temperature gradient information entropy indicates a more disordered gradient direction. The original temperature gradient information entropy is then multiplied by the entropy correction coefficient to correct it, resulting in a corrected temperature entropy value. In this embodiment, the initial temperature mode weight allocation value is calculated based on the regional defect risk score and the corrected temperature entropy value. The formula for calculating the initial temperature mode weight allocation value is as follows:

[0104]

[0105] In the formula, Assign values ​​to the initial temperature mode weights; Assess regional defect risk. This is the corrected temperature entropy value.

[0106] Simultaneously, this embodiment calculates the proportion distribution between the amplitude of the acoustic vibration frequency components and the total spectral energy of all vibration measurement points within the important monitoring area. Based on the proportion distribution, it calculates the original acoustic frequency amplitude information entropy of the acoustic vibration frequency components in the conductor connection area. For each important monitoring area, this embodiment multiplies the original acoustic frequency amplitude information entropy with an entropy correction coefficient, and corrects the original acoustic frequency amplitude information entropy using the entropy correction coefficient to obtain the corrected acoustic entropy value. This embodiment calculates the initial acoustic modal weight allocation value based on the regional thermal-vibration coupling anomaly index and the corrected acoustic entropy value. The calculation formula for the initial acoustic modal weight allocation value is as follows:

[0107]

[0108] In the formula, Assign values ​​to the initial acoustic modal weights; This refers to the regional thermal vibration coupling anomaly index; This is the corrected acoustic entropy value.

[0109] In this embodiment, the sum of the initial temperature modal weight allocation value and the initial acoustic modal weight allocation value is calculated, and the ratio of the initial temperature modal weight allocation value to the sum of the modal weights is calculated to obtain the final temperature modal weight allocation value. The ratio of the initial acoustic modal weight allocation value to the sum of the modal weights is also calculated to obtain the final acoustic modal weight allocation value.

[0110] S5. The infrared thermal image feature data and acoustic vibration feature data are weighted and fused according to the modal fusion weight allocation value to obtain the regional defect coupling feature vector of the cable joint.

[0111] S6. The probability of cable joint defects is predicted based on the regional defect coupling feature vector, and a graded early warning is performed based on the probability of cable joint defects.

[0112] In some embodiments, the step of predicting the cable joint defect probability based on the regional defect coupling feature vector includes:

[0113] The region defect coupling feature vector is input into a random forest classifier that integrates multiple decision trees. Each decision tree classifies and judges the region defect coupling feature vector to obtain the defect probability votes of each decision tree.

[0114] The defect probability of the cable joint is obtained by selecting the highest number of votes for the defect probability from the defect probability votes of each decision tree, calculating the proportion of the highest number of votes for the defect to the total number of decision trees.

[0115] Specifically, the infrared thermographic feature data and acoustic vibration feature data are weighted and fused using modal fusion weight allocation values. Specifically, the infrared thermographic feature data is multiplied by the temperature modal weight allocation value to obtain an infrared feature vector, and the acoustic vibration feature data is multiplied by the acoustic modal weight allocation value to obtain an acoustic feature vector. The infrared and acoustic feature vectors are then concatenated end-to-end to generate a regional defect coupling feature vector. The dimension of the regional defect coupling feature vector is the sum of the dimensions of the infrared thermographic feature data and the acoustic vibration feature data. The weighted fused regional defect coupling feature vector integrates information from infrared thermography and acoustic vibration, and can more comprehensively reflect the defect status of the cable joint. Simultaneously, this embodiment constructs a random forest classifier integrating multiple decision trees. Each decision tree is trained through random sampling and random feature selection. In this embodiment, the regional defect coupling feature vector is input into the pre-trained random forest classifier, and the classifier traverses from the root node to the leaf node. Each decision tree classifies the regional defect coupling feature vector, and the leaf node outputs the decision tree. The defect type prediction results are as follows: For example, each decision tree determines whether a cable joint has a defect based on the feature value of the regional defect coupling feature vector and outputs the defect probability vote count. The defect probability vote count of each decision tree for the regional defect coupling feature vector is counted. The highest defect vote count is selected from the defect probability vote counts of each decision tree, and the proportion of the highest defect vote count to the total number of decision trees is calculated to obtain the cable joint defect probability. The cable joint defect probability reflects the possibility that the cable joint has a defect. The higher the probability, the greater the defect risk. This embodiment performs graded early warning based on the cable joint defect probability. For example, this embodiment can set the graded standard as follows: when the defect probability is less than 0.3, a level 1 warning (normal) is determined; when the defect probability is between 0.3 and 0.7, a level 2 warning (minor defect) is determined; and when the defect probability is greater than or equal to 0.7, a level 3 warning (serious defect) is determined. This embodiment outputs corresponding early warning information based on the graded results of the cable joint defect probability, indicating that there is a serious defect and that immediate inspection and repair are required.

[0116] This invention provides a method for early warning of cable joint anomalies. The method involves real-time acquisition of infrared thermographic data of the cable joint surface and acoustic vibration data generated by the cable joint vibration. Feature extraction is performed on the infrared thermographic data and the acoustic vibration data to obtain a multimodal feature dataset. Based on the multimodal feature dataset, the coupling correlation between the hotspot diffusion gradient and vibration frequency components of the current infrared temperature distribution in different directions is quantified to obtain a regional thermal-vibration coupling anomaly index for the cable joint. The distribution of regional defect types is identified based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint, and a regional defect risk score for the cable joint in different regions is calculated based on the regional defect type distribution. Multimodal weight allocation is performed based on the regional defect risk score to obtain a modal fusion weight allocation value. The infrared thermographic feature data and acoustic vibration feature data are weighted and fused according to the modal fusion weight allocation value to obtain a regional defect coupling feature vector for the cable joint. The probability of cable joint defects is predicted based on the regional defect coupling feature vector, and a graded early warning is performed based on the cable joint defect probability. Compared with existing technologies, this method achieves multimodal feature weighted fusion and defect probability prediction by fusing infrared thermography and acoustic vibration data, thus enabling accurate early warning of cable joint anomalies and effectively improving the accuracy of cable joint anomaly detection and the timeliness of early warning.

[0117] It should be noted that the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0118] In one embodiment, such as Figure 2 As shown, this embodiment of the invention provides a cable joint abnormality early warning system, the system comprising:

[0119] The feature extraction module 101 is used to collect infrared thermal image data of the cable joint surface and acoustic vibration data generated by the cable joint vibration in real time, and to extract features from the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset.

[0120] Temperature analysis module 102 is used to quantify the coupling correlation between the hot spot diffusion gradient and vibration frequency component in different directions of the current infrared temperature distribution based on the multimodal feature dataset, and obtain the regional thermal vibration coupling anomaly index of the cable joint.

[0121] The defect analysis module 103 is used to identify the distribution of regional defect types based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint, and to calculate the regional defect risk score of the cable joint in different regions based on the regional defect type distribution.

[0122] The weight allocation module 104 is used to perform multimodal weight allocation based on the regional defect risk score and the regional thermal vibration coupling anomaly index to obtain the modal fusion weight allocation value.

[0123] Modal fusion module 105 is used to perform weighted fusion of infrared thermal image feature data and acoustic vibration feature data according to the modal fusion weight allocation value to obtain the regional defect coupling feature vector of the cable joint;

[0124] The defect early warning module 106 is used to predict the probability of cable joint defects based on the regional defect coupling feature vector, and to provide graded early warning based on the probability of cable joint defects.

[0125] For specific limitations regarding a cable joint anomaly early warning system, please refer to the above-described limitations regarding a cable joint anomaly early warning method, which will not be repeated here. Those skilled in the art will recognize that the various modules and steps described in conjunction with the embodiments disclosed in this application can be implemented in hardware, software, or a combination of both. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0126] This invention provides a cable joint anomaly early warning system. The system uses a feature extraction module to collect real-time infrared thermal image data of the cable joint surface and acoustic vibration data generated by the cable joint vibration. It then extracts features from the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset. A temperature analysis module quantifies the coupling correlation between the current infrared temperature distribution in different directions and the vibration frequency components based on the multimodal feature dataset, obtaining a regional thermal-vibration coupling anomaly index for the cable joint. A defect analysis module identifies the regional defect type distribution based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint, and calculates the regional defect risk score of the cable joint in different regions based on the regional defect type distribution. A weight allocation module performs multimodal weight allocation based on the regional defect risk score to obtain a modal fusion weight allocation value. A modal fusion module performs weighted fusion of the infrared thermal image feature data and acoustic vibration feature data based on the modal fusion weight allocation value to obtain a regional defect coupling feature vector for the cable joint. A defect early warning module predicts the cable joint defect probability based on the regional defect coupling feature vector and performs graded early warning based on the cable joint defect probability. Compared with existing technologies, this system achieves multimodal feature weighted fusion and defect probability prediction by fusing infrared thermal imaging and acoustic vibration data, enabling accurate early warning of cable joint anomalies and effectively improving the accuracy of cable joint anomaly detection and the timeliness of early warning.

[0127] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the technical principles of this invention, and these improvements and substitutions should also be considered within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the scope of the claims.

Claims

1. A method for early warning of abnormal cable joints, characterized in that, Includes the following steps: Infrared thermal image data of the cable joint surface and acoustic vibration data generated by the cable joint vibration are collected in real time, and feature extraction is performed on the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset. Based on the multimodal feature dataset, the coupling correlation between the hot spot diffusion gradient and vibration frequency components in different directions of the current infrared temperature distribution is quantified to obtain the regional thermal vibration coupling anomaly index of the cable joint. Based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint, the distribution of regional defect types is identified, and the regional defect risk score of the cable joint in different regions is calculated based on the regional defect type distribution. Based on the regional defect risk score and the regional thermal vibration coupling anomaly index, multimodal weight allocation is performed to obtain the modal fusion weight allocation value. The infrared thermal image feature data and acoustic vibration feature data are weighted and fused according to the modal fusion weight allocation value to obtain the regional defect coupling feature vector of the cable joint. The probability of cable joint defects is predicted based on the regional defect coupling feature vector, and a graded early warning is performed based on the probability of cable joint defects. The step of identifying the distribution of regional defect types based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint includes: The spatial boundary coordinates of the conductor connection area, insulation layer area, and outer sheath area are divided according to the regional structural characteristic parameters of the cable joint. The regional thermal vibration coupling anomaly index is mapped to the spatial boundary coordinates of the region according to its spatial location, and the arithmetic mean and maximum value of the regional thermal vibration coupling anomaly index in each region of the cable joint are calculated to identify the defective region of the cable joint. Based on the mechanical resonance characteristics of the material, the characteristic frequency band corresponding to the defect area of ​​the cable joint is determined, and the characteristic frequency band of the defect area of ​​the cable joint is obtained. Calculate the relative rate of change of frequency band energy of the characteristic frequency band of each cable joint defect area relative to the reference value to obtain the rate of change of vibration energy in the cable joint defect area. Based on the rate of change of vibration energy, the regional defect types of each cable joint defect area are identified, forming a regional defect type distribution.

2. The cable joint abnormality early warning method as described in claim 1, characterized in that, The step of extracting features from the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset includes: The infrared thermal image data is filtered to obtain infrared thermal image filtered data, and the gray values ​​in the infrared thermal image filtered data are converted into actual temperature values ​​to obtain the surface temperature field matrix of the cable joint. The first-order gradient components of the surface temperature field matrix of the cable joint in the horizontal and vertical directions are calculated using the edge detection operator to obtain the temperature gradient components. Temperature anomaly regions are identified based on the temperature gradient components, and thermal spatial feature vectors of the temperature anomaly regions are extracted. The acoustic vibration data is subjected to a fast Fourier transform to obtain an acoustic energy spectrum, and the acoustic vibration frequency components whose amplitudes exceed a preset characteristic peak threshold are identified from the acoustic energy spectrum. The thermal spatial feature vector and the acoustic vibration frequency component are aligned using timestamps, and the aligned thermal spatial feature vector and acoustic vibration frequency component are arranged in time sequence to form a multimodal feature dataset.

3. The cable joint abnormality early warning method as described in claim 2, characterized in that: The thermal space feature vector includes at least the geometric center coordinates, boundary perimeter, area, and temperature peak of the temperature anomaly region.

4. The cable joint abnormality early warning method as described in claim 1, characterized in that, The step of quantifying the coupling correlation between the hot spot diffusion gradient and vibration frequency components in different directions based on the multimodal feature dataset to obtain the regional thermal vibration coupling anomaly index of the cable joint includes: Based on the analysis of the multimodal feature dataset, temperature gradient abrupt change points are identified, and the boundary contours of hot spot regions are determined. The hotspot evolution is calculated based on the boundary contours of the hotspot regions at adjacent time points, and a hotspot diffusion gradient vector is constructed based on the hotspot evolution; the hotspot evolution includes at least the hotspot area change rate and the hotspot centroid displacement velocity; The location of the hot spot centroid is determined based on the hot spot diffusion gradient vector, and the amplitude of the acoustic vibration frequency component corresponding to the location of the hot spot centroid is extracted from the multimodal feature dataset. The Pearson correlation coefficient between the hot spot diffusion gradient vector and the amplitude of the acoustic vibration frequency component is calculated to obtain the temperature vibration correlation coefficient matrix. Using the temperature vibration correlation coefficient matrix as the correlation weight, the hot spot diffusion gradient vector and the amplitude of the acoustic vibration frequency component are weighted and fused to obtain cross-modal fusion features. Based on the cross-modal fusion characteristics, the spatial thermal-vibration coupling anomaly distribution is obtained by using the least squares linear regression algorithm, and the thermal-vibration coupling horizontal gradient component, thermal-vibration coupling vertical gradient component, and thermal-vibration coupling gradient amplitude are calculated according to the thermal-vibration coupling anomaly distribution between adjacent time points. The regional thermal vibration coupling anomaly index of the cable joint is obtained by weighted summing of the horizontal gradient component of thermal vibration coupling, the vertical gradient component of thermal vibration coupling, and the magnitude of thermal vibration coupling.

5. The cable joint abnormality early warning method as described in claim 4, characterized in that: The hot spot diffusion gradient vector includes a diffusion rate scalar value and a diffusion direction unit vector; wherein, the diffusion rate scalar value is the vector magnitude of the hot spot area change rate and the hot spot centroid displacement velocity; and the diffusion direction unit vector is a normalized vector of the hot spot centroid displacement direction.

6. The cable joint abnormality early warning method as described in claim 1, characterized in that, The step of calculating the regional defect risk score of cable joints in different regions based on the regional defect type distribution includes: Extract the regional defect types corresponding to conductor connection area, insulation layer area and outer sheath area from the regional defect type distribution, and obtain the defect duration and defect over-limit ratio corresponding to each regional defect type; The proportion of the time the defect continues to exceed the limit to the total monitoring time is calculated, a time enhancement factor is generated, and the time enhancement factor is used to correct the proportion of the defect exceeding the limit to obtain an effective value exceeding the limit. The regional defect score is calculated based on the effective excess value and the preset defect type basic weight coefficient. When defects exist simultaneously in adjacent areas of a cable joint, the defect score of the area is corrected using a preset cross-area coupling coefficient to obtain a corrected area defect score. The corrected regional defect scores are normalized to obtain the regional defect risk scores of cable joints in each region.

7. The cable joint abnormality early warning method as described in claim 4, characterized in that, The modal fusion weight allocation value includes a temperature modal weight allocation value and an acoustic modal weight allocation value. The step of performing multimodal weight allocation based on the regional defect risk score and the regional thermal vibration coupling anomaly index to obtain the modal fusion weight allocation value includes: The scope of important monitoring areas for cable joints in each region is determined based on the regional defect risk score. The information entropy of the temperature anomaly gradient and the amplitude of the acoustic vibration frequency component within the important monitoring area is calculated using the regional thermal vibration coupling anomaly index, and the corresponding temperature mode weight allocation value and acoustic mode weight allocation value are obtained.

8. The cable joint abnormality early warning method as described in claim 1, characterized in that, The step of predicting the cable joint defect probability based on the regional defect coupling feature vector includes: The region defect coupling feature vector is input into a random forest classifier that integrates multiple decision trees. Each decision tree classifies and judges the region defect coupling feature vector to obtain the defect probability votes of each decision tree. The defect probability of the cable joint is obtained by selecting the highest number of votes for the defect probability from the defect probability votes of each decision tree, calculating the proportion of the highest number of votes for the defect to the total number of decision trees.

9. A cable joint abnormality early warning system, characterized in that, For implementing the cable joint abnormality early warning method as described in any one of claims 1-8, the system comprises: The feature extraction module is used to collect infrared thermal image data of the cable joint surface and acoustic vibration data generated by the cable joint vibration in real time, and to extract features from the infrared thermal image data and the acoustic vibration data to obtain a multimodal feature dataset. The temperature analysis module is used to quantify the coupling correlation between the hot spot diffusion gradient and vibration frequency components in different directions of the current infrared temperature distribution based on the multimodal feature dataset, and to obtain the regional thermal vibration coupling anomaly index of the cable joint. The defect analysis module is used to identify the distribution of regional defect types based on the regional structural characteristic parameters and material mechanical resonance characteristics of the cable joint, and to calculate the regional defect risk score of the cable joint in different regions based on the regional defect type distribution. The weight allocation module is used to perform multimodal weight allocation based on the regional defect risk score and the regional thermal vibration coupling anomaly index to obtain the modal fusion weight allocation value. The modal fusion module is used to perform weighted fusion of infrared thermal image feature data and acoustic vibration feature data according to the modal fusion weight allocation value to obtain the regional defect coupling feature vector of the cable joint; The defect early warning module is used to predict the probability of cable joint defects based on the regional defect coupling feature vector, and to provide graded early warnings based on the probability of cable joint defects.

Images