Isolation switch defect detection method, system and equipment based on vibration signal and medium
By combining denoising and multimodal signal decomposition methods, multidimensional features of disconnecting switches are extracted, solving the problem of poor fault feature extraction in the background of noise in the existing technology, and realizing accurate identification of disconnecting switch faults and high sensitivity detection of early minor faults.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-28
- Publication Date
- 2026-03-31
AI Technical Summary
Existing technologies struggle to effectively extract fault characteristics from disconnect switches in strong noise environments and fail to establish a comprehensive multi-dimensional feature system that fully reflects the mechanical state, resulting in inaccurate separation of fault components, especially insufficient sensitivity to early minor faults.
A combined denoising and multimodal signal decomposition method is adopted. Noise is removed by adaptive thresholding and matrix decomposition. Combined with adaptive noise addition and constraint variational problem solving, multidimensional features such as correlation, complexity, energy distribution and spectral features are extracted to construct a multidimensional feature system. Specific faults are identified by the energy ratio of the fundamental band and third harmonic band.
It effectively preserves weak fault characteristics in high-noise environments, achieves accurate separation of fault components and multi-dimensional feature extraction, and improves the detection capability of early minor faults and high-risk serious faults of disconnecting switches, especially the accurate identification of specific faults such as contact arcing and mechanical resonance.
Smart Images
Figure CN121763068A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of condition monitoring and fault diagnosis technology, specifically to a method, system, device, and medium for detecting defects in disconnect switches based on vibration signals. Background Technology
[0002] With the deepening of smart grid construction, gas-insulated metal-enclosed switchgear, as a key piece of equipment in the power system, directly affects grid security due to its operational reliability. Disconnect switches, as the most frequently operated mechanical components in GIS (Gas-Insulated Switchgear), are subjected to multiple stresses such as electrodynamics and mechanical shocks over long periods, making them prone to typical faults such as contact wear and mechanism jamming. Existing technologies mainly focus on signal acquisition optimization and feature extraction algorithm improvement. Time-frequency analysis techniques, such as wavelet transform and empirical mode decomposition, are widely used in vibration signal processing to achieve preliminary extraction of fault features through multi-scale decomposition of the signal. In terms of feature representation, researchers have proposed various feature parameters, including time-domain statistical features, frequency-domain features, and time-frequency-domain features, attempting to establish a mapping relationship between vibration signals and equipment states. In terms of diagnostic strategies, pattern recognition and machine learning methods are mainly used to achieve state classification.
[0003] Existing technologies still have significant limitations in practical applications: First, in the signal preprocessing stage, traditional single denoising methods are difficult to effectively address the coexistence of strong noise interference and weak fault features in the complex electromagnetic environment of substations. While wavelet threshold denoising can retain some high-frequency components, it easily leads to the loss of effective information. Singular value decomposition denoising is not adaptable enough to the non-stationary characteristics of signals, resulting in poor fault feature extraction under strong noise backgrounds. Second, in terms of signal decomposition, traditional empirical mode decomposition exhibits significant mode aliasing, making it difficult to achieve accurate separation of fault components. Although variational mode decomposition can alleviate this problem, its parameter settings rely on prior knowledge and have poor adaptability to various fault types of disconnecting switches. More importantly, existing feature extraction methods are mostly limited to a single dimension, or focus on time-domain statistical features while ignoring frequency-domain information, or focus on energy features while lacking complexity representation. They have failed to establish a multi-dimensional feature system that comprehensively reflects the mechanical state of disconnecting switches, especially lacking sensitivity to early minor faults. Summary of the Invention
[0004] In view of the above-mentioned existing problems, the present invention provides a method, system, device and medium for detecting defects in disconnecting switches based on vibration signals, in order to solve the problems in the prior art that the fault feature extraction effect is not good under strong noise background, it is difficult to achieve accurate separation of fault components, and it fails to establish a multi-dimensional feature system that fully reflects the mechanical state of disconnecting switches.
[0005] To address the aforementioned technical problems, a method for detecting defects in disconnecting switches based on vibration signals is proposed, including: Vibration signals are collected by vibration sensors mounted on the disconnector switch housing. The vibration signals are then subjected to joint denoising processing to remove environmental noise and interference while retaining fault characteristics. The denoised vibration signals are then subjected to multimodal signal decomposition to obtain signal components. Multimodal signal decomposition includes adding adaptive noise and iteratively decomposing the signal to obtain multiple first signal components, and decomposing the first signal components to obtain second signal components by solving a constrained variational problem. Multidimensional features are extracted from the second signal components, and defects in the disconnector switch are detected based on these multidimensional features.
[0006] As a preferred embodiment of the vibration signal-based disconnector defect detection method of the present invention, the joint denoising process includes: a first denoising method decomposing the vibration signal into frequency sub-bands and performing adaptive thresholding on each frequency sub-band to remove noise; a second denoising method constructing a signal matrix from the denoised signal, performing matrix decomposition, extracting the main components, and reconstructing the denoised signal based on the main components; wherein the adaptive thresholding includes dynamically adjusting the threshold value according to the signal characteristics to reconstruct the denoised signal.
[0007] As a preferred embodiment of the vibration signal-based disconnector defect detection method of the present invention, the multimodal signal decomposition includes: a first decomposition adding adaptive noise to the denoised signal and obtaining a first signal component through iterative calculation; and a second decomposition constructing a constrained variational problem and obtaining a second signal component through frequency domain iterative solution.
[0008] As a preferred embodiment of the vibration signal-based disconnector defect detection method of the present invention, the extraction of multi-dimensional features includes correlation feature extraction, which involves calculating the correlation between each signal component and the denoised vibration signal, evaluating the signal correlation degree, and screening sensitive signal components based on the correlation magnitude. Complexity feature extraction assesses signal randomness by analyzing the complexity of each signal component under coarse-grained processing and permutation pattern statistics. Energy feature extraction involves calculating the energy distribution of each signal component and assessing energy symmetry. Spectral feature extraction assesses the significance of impulse components by analyzing the envelope spectrum of each signal component.
[0009] As a preferred embodiment of the vibration signal-based disconnector defect detection method of the present invention, the correlation feature extraction includes: calculating the mean of each signal component and the mean of the denoised vibration signal, calculating the covariance of each signal component and the denoised vibration signal, calculating the standard deviation of each signal component and the standard deviation of the denoised vibration signal respectively, obtaining the correlation coefficient value by the ratio of the product of the covariance and the two standard deviations, and selecting signal components with a correlation coefficient higher than the preset correlation coefficient threshold as sensitive signal components based on the preset correlation coefficient threshold. The formula for calculating the correlation coefficient is as follows: in, Let be the correlation coefficient between the i-th intrinsic mode function component and the denoised vibration signal. Let be the signal value at time t in the i-th eigenmode function. Let i be the average value of the i-th intrinsic mode function component over all time points. Let be the value of the vibration signal at time t after joint denoising processing. The noise-reduced vibration signal is the average value at all time points. The total length of the vibration signal. For time indexing; The complexity feature extraction includes coarse-graining each sensitive signal component under different scale factors, obtaining a coarse-grained sequence by segmenting the signal and calculating the average value of each segment, reconstructing each coarse-grained sequence into a multi-dimensional vector, obtaining a symbol sequence by arranging the vector elements, and counting the frequency of each symbol sequence, calculating the permutation entropy value based on the frequency, taking the permutation entropy values at multiple scales, and constructing a multi-scale permutation entropy feature vector to characterize the complexity change of the signal at different time scales. The formula for coarsening is expressed as follows: in, Let be the value of the j-th coarse-grained sequence under the scale factor τ. As a scale factor, For indexing coarse-grained sequences, This is the length of the coarsened sequence, rounded down. The permutation entropy value is calculated as follows: in, Let m be the reconstructed m-dimensional vector, representing m consecutive data points starting from the l-th point in the coarse-grained sequence. For the embedding dimension, The values for the l-th to l+m-1-th coarse-grained sequences, This is the starting index for reconstructing the vector. This represents the total number of reconstructed vectors; The formula for the permutation entropy value under multiple scales is expressed as follows: in, For the arrangement pattern, For the occurrence pattern Number of times, Arrangement mode The probability of occurrence Let be the permutation entropy value under the scale factor τ; The energy feature extraction includes: summing the squares of the amplitudes of each signal component to obtain the energy value; calculating the total energy of all signal components; calculating the ratio of the energy of each signal component to the total energy; calculating the mean of the energy distribution based on the ratio; and obtaining the energy distribution skewness by calculating the ratio of the third moment of the difference between the energy ratio and the mean to the cube of the standard deviation. The formula for calculating the total energy of all signal components is expressed as: in, Let be the energy of the i-th eigenmode function component. The total energy of all intrinsic mode function components. The total number of intrinsic mode function components. The index of the intrinsic mode function component; The formula for calculating the mean of the energy distribution is expressed as: in, Energy distribution skewness characterizes the asymmetry of energy distribution. Let be the normalized energy of the i-th eigenmode function component. This represents the average value of the normalized energy. The spectral feature extraction includes performing a Hilbert transform on each signal component to obtain an analytical signal, calculating the amplitude of the analytical signal as an envelope signal, performing a Fourier transform on the envelope signal to obtain an envelope spectrum, and calculating the maximum value and root mean square value of the envelope spectrum, and obtaining the peak factor by the ratio of the maximum value to the root mean square value of the envelope spectrum. The formula for calculating the envelope signal is expressed as: in, Let i be the envelope signal of the i-th intrinsic mode function component at time t. The result of the Hilbert transform of the i-th intrinsic mode function component at time t; The formula for obtaining the peak factor is expressed as: in, Let be the peak factor of the envelope spectrum of the i-th intrinsic mode function component. Let i be the envelope spectrum of the i-th intrinsic mode function component. The number of frequency points in the envelope spectrum. is the root mean square value of the envelope spectrum, representing the overall energy level of the envelope spectrum.
[0010] As a preferred embodiment of the vibration signal-based disconnector defect detection method of the present invention, the extraction of multi-dimensional features further includes: calculating the power spectral density of the vibration signal, integrating the power spectral density in the fundamental frequency band to obtain the fundamental frequency band energy, integrating the power spectral density in the range from zero to the Nyquist frequency to obtain the total energy, the ratio of the fundamental frequency band energy to the total energy is the fundamental frequency band energy ratio, and integrating the power spectral density in the third harmonic band to obtain the third harmonic band energy, the ratio of the third harmonic band energy to the total energy is the third harmonic band energy ratio; The formula for the fundamental band energy ratio is expressed as: in, The fundamental frequency band energy ratio, The power spectral density of the signal. For frequency, The sampling frequency of the signal. The Nyquist frequency; The formula for the third harmonic band energy ratio is expressed as follows: in, It has a three-fold energy ratio. It is the integral variable.
[0011] As a preferred embodiment of the vibration signal-based disconnector defect detection method of the present invention, the detection of disconnector defects includes: combining multi-scale permutation entropy feature vectors, energy distribution skewness, peak factor, fundamental band energy ratio, and third harmonic band energy ratio into a feature vector set; inputting the current feature vector set into a trained classifier for pattern recognition; and comparing it with a threshold set based on historical data; when the feature value exceeds the threshold range, it is determined to be a specific fault type, and automatic detection of disconnector defect is performed.
[0012] The beneficial effects of this preferred technical solution are as follows: By calculating the fundamental frequency band energy ratio and the third harmonic band energy ratio, targeted monitoring of the specific fault mechanism of the disconnecting switch is realized. The fundamental frequency band energy ratio focuses on the fundamental frequency component of the disconnecting switch operation, effectively reflecting the state changes of the mechanical transmission system. The third harmonic band energy ratio is highly sensitive to specific faults such as contact arc and mechanical resonance, achieving the effect of accurate identification of specific fault types and making up for the shortcomings of general features in specific fault diagnosis.
[0013] As a preferred embodiment of the vibration signal-based disconnector defect detection system of the present invention, it is characterized by comprising a signal acquisition module, a signal preprocessing module, a signal decomposition module, and a feature extraction module.
[0014] The signal acquisition module is used to construct a monitoring network covering the key areas of the disconnect switch by arranging vibration sensors at different locations on the GIS shell, and to collect vibration signals generated during operation in real time.
[0015] The signal preprocessing module is used to process the original vibration signal using a joint denoising strategy, employing both wavelet packet denoising and singular value decomposition denoising methods, to suppress environmental noise and electromagnetic interference while retaining fault-related impact components.
[0016] The signal decomposition module is used to adopt a multimodal decomposition strategy to adaptively decompose the denoised signal into intrinsic mode functions through CEEMDAN decomposition, and to perform VMD secondary decomposition on the selected high-frequency IMF components.
[0017] The feature extraction module is used to extract multi-dimensional features such as correlation coefficient, multi-scale permutation entropy, energy entropy, envelope spectrum peak factor, and fault-sensitive frequency band energy ratio from the decomposed signal components.
[0018] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a method for detecting defects in disconnect switches based on vibration signals.
[0019] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for detecting defects in disconnecting switches based on vibration signals.
[0020] The beneficial effects of this invention are as follows: This invention employs joint denoising processing, combining frequency band subdivision with adaptive threshold processing to effectively preserve weak fault characteristics in noisy environments, providing a high-quality signal foundation for subsequent analysis. Through multimodal signal decomposition, combined with adaptive noise addition and constrained variational problem solving, it achieves precise separation of fault components, overcoming mode aliasing problems. By constructing a multi-dimensional feature system, it comprehensively characterizes equipment status from multiple perspectives, including correlation, complexity, energy distribution, and spectral characteristics. Correlation coefficient screening ensures feature specificity, multi-scale permutation entropy captures random changes in signals, energy distribution skewness reflects energy redistribution phenomena, and envelope spectrum peak factor highlights impact components, forming a detection capability with high sensitivity to both early minor faults and high-risk severe faults. Furthermore, through targeted analysis of the energy ratio between the fundamental band and third harmonic band, it achieves accurate identification of specific faults such as contact arcing and mechanical resonance. With the help of feature vector combination and intelligent classifier decision-making, a multi-level diagnostic mechanism is established, providing reliable technical support for predictive maintenance of equipment status. Attached Figure Description
[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0022] Figure 1 The above is a flowchart of a method for detecting defects in disconnecting switches based on vibration signals, provided in one embodiment of the present invention.
[0023] Figure 2 The flowchart shows a system scheme for a vibration signal-based disconnector defect detection system according to an embodiment of the present invention. Detailed Implementation
[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.
[0025] Example 1, referring to Figure 1 As one embodiment of the present invention, a method for detecting defects in disconnect switches based on vibration signals is provided, comprising: S100: Vibration signals are collected by vibration sensors arranged on the housing of the disconnect switch, and the vibration signals are subjected to joint noise reduction processing to remove environmental noise and interference while retaining fault characteristics.
[0026] S200: Perform multimodal signal decomposition on the denoised vibration signal to obtain signal components. Multimodal signal decomposition includes obtaining multiple first signal components by adding adaptive noise and iteratively decomposing the signal, and obtaining second signal components by solving a constrained variational problem to decompose the first signal components.
[0027] S300: Extracts multi-dimensional features from the second signal component and detects defects in the disconnecting switch based on the multi-dimensional features.
[0028] It should be noted that by employing a joint denoising process combining frequency subband adaptive thresholding and matrix factorization, strong background noise was effectively filtered out. By combining adaptive noise iteration and constrained variational solution for multimodal signal decomposition, the modal aliasing defect of traditional methods was overcome, and clean signal components that can accurately characterize different fault sources were obtained. Furthermore, by extracting multi-dimensional features such as correlation, complexity, energy, spectrum, and power spectral density in parallel, a complementary feature set was constructed, which effectively distinguished composite faults of different types and stages. By inputting the feature set into a trained classifier for intelligent pattern recognition, automated and accurate detection and severity assessment of disconnector switch defects were achieved.
[0029] Example 2, refer to Figure 1 This is a second embodiment of the present invention, which provides a method for detecting defects in disconnect switches based on vibration signals, including: In step S100, the acquisition of vibration signals includes steps S101~S102: S101: Multiple high-sensitivity vibration sensors are arranged at different locations on the GIS housing to cover all key areas of the equipment.
[0030] S102: By deploying sensors at multiple points, the sensors capture vibration signals from different locations, effectively increasing the monitoring range of the system and ensuring that there are no blind spots.
[0031] Furthermore, in this embodiment of the application, in step S100, the joint denoising process includes first denoising, i.e., wavelet packet denoising, and second denoising, i.e., SVD denoising, specifically including steps S111~S112: S111: The vibration signal is decomposed into different frequency bands by wavelet packet decomposition, and the coefficients of each sub-band are processed by an improved adaptive threshold function. After reconstruction, a preliminary denoised signal is obtained. The wavelet packet decomposition formula is expressed as: in, The original vibration signal is the vibration data of the disconnector switch collected at time t, which serves as the time index. The number of wavelet packet decomposition levels. This is the index of the current decomposition level. For the node index at level a, Let be the wavelet packet coefficients at node k in layer a. These are wavelet packet basis functions; The improved adaptive threshold function formula is expressed as follows: in, The final selected coefficient index for signal reconstruction is used to find the metric. Index when taking the maximum value , This is an empirically corrected value.
[0032] S112: Construct the Hankel matrix to perform singular value decomposition, select the first n principal singular values to reconstruct the signal matrix, and recover the denoised vibration signal by the diagonal averaging method; The formula for restoring the denoised vibration signal is expressed as: in, The signal matrix after denoising. For the 0th singular value, It is a left singular vector. It is a right singular vector. This is the transpose of the right singular vector. The number of singular values selected. The index for the singular value.
[0033] In an optional implementation, in step S100, the joint denoising process further includes decomposing the vibration signal into multiple intrinsic mode functions (IMFs) using EMD, screening the main IMF components for reconstruction based on the correlation coefficient threshold, and performing PCA analysis on the reconstructed signal to construct a time-domain matrix, retaining the main eigenvectors to reconstruct the signal.
[0034] In another optional implementation, in step S100, the joint denoising process may further include: using an LMS adaptive filter to perform preliminary denoising on the vibration signal, automatically adjusting the filter coefficients according to the reference signal, and performing SSA decomposition on the filtered signal to decompose the time series into trend components, oscillation components and noise components, and selecting the first few main components to reconstruct the signal.
[0035] Furthermore, in this embodiment of the application, in step S111, the first denoising includes steps A1~A3: A1: Adaptive thresholding is achieved using wavelet packet decomposition and an improved soft thresholding function.
[0036] A2: Perform J-level wavelet packet decomposition on the vibration signal to obtain... For each sub-band coefficient, a scale-adaptive threshold function is designed.
[0037] A3: The threshold value is dynamically adjusted according to the coefficient distribution characteristics, and finally the denoised signal is obtained through wavelet packet reconstruction.
[0038] In an optional implementation, in step S111, the first denoising further includes decomposing the signal into multiple IMF components using EEMD, calculating the instantaneous frequency and amplitude for each IMF component, designing a time-varying threshold function based on the instantaneous frequency characteristics, and performing threshold processing in the time-frequency domain.
[0039] In another optional implementation, in step S111, the first denoising may further include analyzing signal features at different time scales using MSSA, designing multi-scale thresholds based on the correlation between each scale component and noise, and suppressing noise scale by scale.
[0040] In this embodiment of the application, step S200, the multimodal signal decomposition includes steps S201~S202: S201: The first decomposition includes using adaptive noise complete set empirical mode decomposition (CEEMDAN) to perform preliminary processing on the denoised vibration signal. By adding adaptive white noise to the original signal multiple times and performing decomposition, and integrating and averaging the results, the mode aliasing problem in traditional empirical mode decomposition is overcome. The complex non-stationary signal is decomposed into a series of intrinsic mode function (IMF) components arranged from high frequency to low frequency. After decomposition, the difference between each IMF component and the previous residual is calculated, and the current residual is used as the input for the next stage of decomposition. Through iterative extraction, until the final residual signal becomes a monotonic function or there are too few extreme points, a series of IMF components containing different fault information are obtained. The CEEMDAN decomposition formula is expressed as: in, The signal to be decomposed in the h-th experiment, The vibration signal is after joint processing of wavelet packet denoising and SVD denoising. The noise figure controls the intensity of noise added to the original signal. White noise with zero mean and unit variance. To extract the first eigenmode function from the EMD decomposition, Let q be the eigenmode function obtained from CEEMDAN decomposition. The total average number of times, Let H be the first intrinsic mode function obtained by EMD decomposition of the h-th test signal. For the q-th order residual signal, For the (q+1)th eigenmode function, Let be the noise figure for the q-th stage. To extract the q-th eigenmode function from the EMD decomposition, Let be the signal-to-noise ratio adjustment parameter for the q-th stage. The qth order residual signal standard deviation For the final residual, This represents the total number of intrinsic mode function components obtained from the final CEEMDAN decomposition. This is the index of the intrinsic mode function component.
[0041] S202: The second decomposition includes performing variational mode decomposition (VMD) on the first few high-frequency IMF components obtained after CEEMDAN decomposition. VMD involves constructing a constrained variational problem, decomposing each input IMF component into several variational mode functions with specific center frequencies and finite bandwidths, and introducing an augmented Lagrangian function to transform the constrained problem into an unconstrained problem. The problem is then solved iteratively in the frequency domain using the alternating direction multiplier method. In each iteration, the algorithm alternately updates each mode function and its corresponding center frequency until the convergence condition is met. The constrained variational problem is expressed as follows: in, Let g be the variational mode function. Let g be the center frequency of the g-th mode. The preset number of modal decompositions, The input signal to be decomposed is... For Dirac delta function, The imaginary unit, Let be the partial derivative with respect to time t. For convolution operations, It is the L2 norm. Pi As a limiting condition, For variational mode function index, This represents the total number of variational mode functions; The augmented Lagrange function is expressed as: in, For bandwidth parameters, To augment the Lagrange function, For Lagrange multipliers, This is an inner product operation; The iterative solution formula is expressed as: in, Let g be the frequency domain representation of the g-th mode function in the (S+1)-th iteration. The frequency domain representation of the original input signal to be decomposed. This is the frequency domain representation of all modal functions after the S-th iteration, excluding the currently updating g-th mode. For the frequency domain representation of Lagrange multipliers, For bandwidth parameters, Let g be the center frequency of the g-th mode in the current iteration (S-th iteration). Index for iteration count, Let g be the center frequency of the g-th mode after the (S+1)-th iteration. Let g be the frequency domain representation of the g-th mode function after the update in the (S+1)-th iteration. For modality The power spectrum represents the energy distribution of the current mode at different frequencies. This is the weighted integral of the current modal power spectrum, with the weights being the frequency. , The total energy of the current mode. For variational mode functions, For modal indexing, This represents the total number of modes.
[0042] In an optional implementation, in step S200, the multimodal signal decomposition further includes decomposing the signal into multiple internal scale components using an LCD, automatically determining the number of decomposition layers based on the local features of the signal, performing EWT decomposition on the main ISC components, adaptively dividing the frequency band according to the spectral characteristics, and extracting the modal components.
[0043] In another alternative implementation, in step S200, the multimodal signal decomposition may further include processing the non-stationary frequency-modulated signal by VNCMD, accurately separating the frequency modulation components, and using SCA to extract sparse fault components from the mixed signal.
[0044] In this embodiment of the application, step S300, the extraction of multi-dimensional features includes correlation feature extraction, complexity feature extraction, energy feature extraction, and spectral feature extraction, specifically including steps S301 to S304: S301: The correlation feature extraction includes calculating the mean of each signal component and the mean of the denoised vibration signal, calculating the covariance of each signal component and the denoised vibration signal, calculating the standard deviation of each signal component and the standard deviation of the denoised vibration signal respectively, obtaining the correlation coefficient value by the ratio of the product of the covariance and the two standard deviations, and selecting signal components with a correlation coefficient higher than the preset correlation coefficient threshold as sensitive signal components based on the preset correlation coefficient threshold. The formula for calculating the correlation coefficient is as follows: in, Let be the correlation coefficient between the i-th intrinsic mode function component and the denoised vibration signal. Let be the signal value at time t in the i-th eigenmode function. Let i be the average value of the i-th intrinsic mode function component over all time points. Let be the value of the vibration signal at time t after joint denoising processing. The noise-reduced vibration signal is the average value at all time points. The total length of the vibration signal. For time indexing.
[0045] S302: The complexity feature extraction includes coarsening each sensitive signal component under different scale factors, obtaining a coarsened sequence by segmenting the signal and calculating the average value of each segment, reconstructing each coarsened sequence into a multi-dimensional vector, obtaining a symbol sequence by arranging the vector elements, and counting the frequency of each symbol sequence, calculating the permutation entropy value based on the frequency, taking the permutation entropy values at multiple scales, and constructing a multi-scale permutation entropy feature vector to characterize the complexity change of the signal at different time scales; The formula for coarsening is expressed as follows: in, Let be the value of the j-th coarse-grained sequence under the scale factor τ. As a scale factor, For indexing coarse-grained sequences, This is the length of the coarsened sequence, rounded down. The permutation entropy value is calculated as follows: in, Let m be the reconstructed m-dimensional vector, representing m consecutive data points starting from the l-th point in the coarse-grained sequence. For the embedding dimension, The values for the l-th to l+m-1-th coarse-grained sequences, This is the starting index for reconstructing the vector. This represents the total number of reconstructed vectors; The formula for the permutation entropy value under multiple scales is expressed as follows: in, For the arrangement pattern, For the occurrence pattern Number of times, Arrangement mode The probability of occurrence Let be the permutation entropy value under the scale factor τ.
[0046] S303: The energy feature extraction includes summing the squares of the amplitudes of each signal component to obtain the energy value, calculating the total energy of all signal components, calculating the ratio of the energy of each signal component to the total energy, calculating the mean of the energy distribution based on the ratio, and obtaining the energy distribution skewness by calculating the ratio of the third moment of the difference between the energy ratio and the mean to the cube of the standard deviation. The formula for calculating the total energy of all signal components is expressed as: in, Let be the energy of the i-th eigenmode function component. The total energy of all intrinsic mode function components. The total number of intrinsic mode function components. The index of the intrinsic mode function component; The formula for calculating the mean of the energy distribution is expressed as: in, Energy distribution skewness characterizes the asymmetry of energy distribution. Let be the normalized energy of the i-th eigenmode function component. This represents the average value of the normalized energy.
[0047] S304: The spectral feature extraction includes performing a Hilbert transform on each signal component to obtain an analytical signal, calculating the amplitude of the analytical signal as an envelope signal, performing a Fourier transform on the envelope signal to obtain an envelope spectrum, and calculating the maximum value and root mean square value of the envelope spectrum, and obtaining the peak factor by the ratio of the maximum value to the root mean square value of the envelope spectrum. The formula for calculating the envelope signal is expressed as: in, Let i be the envelope signal of the i-th intrinsic mode function component at time t. The result of the Hilbert transform of the i-th intrinsic mode function component at time t; The formula for obtaining the peak factor is expressed as: in, Let be the peak factor of the envelope spectrum of the i-th intrinsic mode function component. Let i be the envelope spectrum of the i-th intrinsic mode function component. The number of frequency points in the envelope spectrum. is the root mean square value of the envelope spectrum, representing the overall energy level of the envelope spectrum.
[0048] In an optional implementation, in step S300, the extraction of multi-dimensional features further includes time-frequency moment features of short-time Fourier transform, Lyapunov exponent, nonlinear features such as correlation dimension, wavelet energy entropy, and bispectral features of the modulation signal.
[0049] In another optional implementation, in step S300, the extraction of multi-dimensional features may further include automatically learning the depth features of the vibration signal using a one-dimensional convolutional neural network, and performing feature-level fusion with traditional time-domain statistical features (peak value, kurtosis, waveform factor, etc.) and frequency-domain features (centroid frequency, frequency variance, etc.).
[0050] Furthermore, in step S300, the extraction of multi-dimensional features also includes calculating the power spectral density of the vibration signal, integrating the power spectral density in the fundamental frequency band to obtain the fundamental frequency band energy, integrating the power spectral density in the range from zero to the Nyquist frequency to obtain the total energy, the ratio of the fundamental frequency band energy to the total energy is the fundamental frequency band energy ratio, and integrating the power spectral density in the third harmonic band to obtain the third harmonic band energy, the ratio of the third harmonic band energy to the total energy is the third harmonic band energy ratio.
[0051] The formula for the fundamental band energy ratio is expressed as: in, The fundamental frequency band energy ratio, The power spectral density of the signal. For frequency, The sampling frequency of the signal. The Nyquist frequency; The formula for the third harmonic band energy ratio is expressed as follows: in, It has a three-fold energy ratio. It is the integral variable.
[0052] Furthermore, in step S300, the detection of defects in the disconnecting switch includes combining the multi-scale permutation entropy feature vector, energy distribution skewness, peak factor, fundamental band energy ratio, and third harmonic band energy ratio into a feature vector set, inputting the current feature vector set into a trained classifier for pattern recognition, and comparing it with a threshold set based on historical data. When the feature value exceeds the threshold range, it is determined to be a specific fault type, and automatic detection of defects in the disconnecting switch is performed.
[0053] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
[0054] Example 3, referring to Figure 2 The third embodiment of the present invention provides a fault detection system for disconnecting switches based on vibration signals, including a signal acquisition module, a signal preprocessing module, a signal decomposition module, and a feature extraction module.
[0055] The signal acquisition module is used to construct a monitoring network covering the key areas of the disconnect switch by arranging vibration sensors at different locations on the GIS shell, and to collect vibration signals generated during operation in real time.
[0056] The signal preprocessing module is used to process the original vibration signal using a joint denoising strategy, employing both wavelet packet denoising and singular value decomposition denoising methods, to suppress environmental noise and electromagnetic interference while retaining fault-related impact components.
[0057] The signal decomposition module is used to adopt a multimodal decomposition strategy to adaptively decompose the denoised signal into intrinsic mode functions through CEEMDAN decomposition, and to perform VMD secondary decomposition on the selected high-frequency IMF components.
[0058] The feature extraction module is used to extract multi-dimensional features such as correlation coefficient, multi-scale permutation entropy, energy entropy, envelope spectrum peak factor, and fault-sensitive frequency band energy ratio from the decomposed signal components.
[0059] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
[0060] Example 4, the fourth embodiment of the present invention, differs from the previous three embodiments in that: If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0061] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.
[0062] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0063] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
Claims
1. A method for detecting defects in disconnectors based on vibration signals, characterized in that: The application relates to a vibration signal acquisition method and device for disconnecting switches. The vibration signal is acquired through a vibration sensor arranged on the disconnecting switch shell, and joint denoising processing is performed on the vibration signal to remove environmental noise and interference and retain fault characteristics. The denoised vibration signal is subjected to multi-modal signal decomposition to obtain signal components, and the multi-modal signal decomposition comprises adding adaptive noise to the signal and iteratively decomposing the signal to obtain a plurality of first signal components, and decomposing the first signal components to obtain second signal components by solving a constrained variation problem. Multi-dimensional features are extracted from the second signal components, and defects of the disconnecting switch are detected based on the multi-dimensional features.
2. The vibration signal based detection method of a disconnector defect as claimed in claim 1, characterized in that: The joint denoising processing comprises first denoising, which decomposes the vibration signal into frequency subbands and performs adaptive threshold processing on each frequency subband to remove noise. Second denoising constructs the denoised signal into a signal matrix, performs matrix decomposition, extracts main components, and reconstructs the denoised signal based on the main components.
3. The vibration signal based detection method of a disconnector defect as claimed in claim 2, characterized in that: The multi-modal signal decomposition comprises first decomposition, which adds adaptive noise to the denoised signal and obtains first signal components through iterative calculation, and second decomposition, which constructs a constrained variation problem and obtains second signal components through frequency domain iterative solving.
4. The vibration signal based detection method of a disconnector defect according to claim 3, characterized in that: The multi-dimensional feature extraction comprises correlation feature extraction, which calculates the correlation between each signal component and the denoised vibration signal, evaluates the signal correlation degree, and selects sensitive signal components based on the correlation size. Complexity feature extraction, which analyzes the complexity of each signal component under coarse-grained processing and permutation pattern statistics to evaluate the signal randomness. Energy feature extraction, which calculates the energy distribution of each signal component to evaluate the energy symmetry. Spectrum feature extraction, which analyzes the envelope spectrum of each signal component to evaluate the impact component significance.
5. The vibration signal based detection method of a disconnector defect as claimed in claim 4, characterized in that: The correlation feature extraction comprises, The mean value of each signal component and the mean value of the denoised vibration signal are calculated, the covariance of each signal component and the denoised vibration signal is calculated, and the standard deviation of each signal component and the denoised vibration signal is calculated, respectively. The correlation coefficient value is obtained by multiplying the covariance by the two standard deviations, and the signal components with a correlation coefficient higher than a preset correlation coefficient threshold are selected as sensitive signal components based on the preset correlation coefficient threshold. wherein, is the correlation coefficient of the i-th EFM component and the denoised vibration signal, is the signal value at time t in the i-th EFM, is the average value of the i-th EFM component over all time points, is the value of the vibration signal at time t after joint denoising, is the average value of the denoised vibration signal over all time points, is the total length of the vibration signal, is the time index; The correlation coefficient value calculation formula is represented as: The complexity feature extraction comprises coarse-grained processing of each sensitive signal component under different scale factors, coarse-grained sequences are obtained by segmenting the signal and calculating the average value of each segment, each coarse-grained sequence is reconstructed into a multi-dimensional vector, a symbol sequence is obtained by arranging the vector elements, the frequency of each symbol sequence is counted, the permutation entropy value is calculated based on the frequency, the permutation entropy values under multiple scales are taken to form a multi-scale permutation entropy feature vector, and the complexity change of the signal under different time scales is represented. wherein, is the value of the jth coarsening sequence at scale factor τ, is the scale factor, is the index of the coarsening sequence, is the length of the coarsened sequence, rounded down. The coarse-grained processing formula is represented as: wherein, is a reconstructed m-dimensional vector, representing the consecutive m data points from the 1st point in the coarse-grained sequence, is the embedding dimension, is the value of the 1st to l+m-1th coarse-grained sequence, is the starting index of the reconstructed vector, is the total number of reconstructed vectors; The permutation entropy value calculation formula is represented as: The permutation entropy value under multiple scales is taken, and the formula is represented as: wherein, is the permutation pattern, is the occurrence pattern of the times, is the permutation pattern occurrence probability, is the permutation entropy value at scale factor τ; The energy feature extraction includes summing the square of the amplitude of each signal component to obtain an energy value, calculating the total energy of all signal components, calculating the ratio of the energy of each signal component to the total energy, and calculating the mean value of the energy distribution based on the ratio, and obtaining the energy distribution skewness by calculating the ratio of the third moment of the energy ratio to the cube of the standard deviation and the mean value; The total energy of all signal components is calculated according to the formula: wherein, is the energy of the i-th eigenmode function component, is the total energy of all eigenmode function components, is the total number of eigenmode function components, is an index of the eigenmode function component. The mean value of the energy distribution is calculated according to the formula: wherein, is the energy distribution skewness, characterizing the asymmetry of the energy distribution, is the normalized energy of the i-th eigenmode function component, is the average value of the normalized energy; The frequency spectrum feature extraction includes performing Hilbert transform on each signal component to obtain an analytic signal, calculating the amplitude of the analytic signal as an envelope signal, performing Fourier transform on the envelope signal to obtain an envelope spectrum, and calculating the maximum value and the root mean square value of the envelope spectrum, and obtaining the peak factor by the ratio of the maximum value to the root mean square value of the envelope spectrum; The envelope signal is calculated according to the formula: wherein is an envelope signal of the i-th eigenmode function component at time instant t, is a Hilbert transform result of the i-th eigenmode function component at time instant t. The peak factor is obtained according to the formula: wherein, is the envelope spectrum peak factor of the i-th EIM component, is the envelope spectrum of the i-th EIM component, is the number of frequency points of the envelope spectrum, is the root mean square value of the envelope spectrum, which represents the overall energy level of the envelope spectrum.
6. The vibration signal based detection method of a disconnector defect as claimed in claim 5, characterized in that: The multi-dimensional feature extraction further includes calculating the power spectral density of the vibration signal, integrating the power spectral density in the fundamental frequency band to obtain the fundamental frequency band energy, and integrating the power spectral density in the range of zero to the Nyquist frequency to obtain the total energy, and the ratio of the fundamental frequency band energy to the total energy is the fundamental frequency band energy ratio, and integrating the power spectral density in the three times frequency band to obtain the three times frequency band energy, and the ratio of the three times frequency band energy to the total energy is the three times frequency band energy ratio; The fundamental frequency band energy ratio is calculated according to the formula: wherein, is the baseband energy ratio, is the power spectral density of the signal, is the frequency, is the sampling frequency of the signal, is the Nyquist frequency; The three times frequency band energy ratio is calculated according to the formula: wherein is the triple-band energy ratio, is the integral variable.
7. The vibration signal based detection method of a disconnector defect as claimed in claim 6, characterized in that: The defect detection of the disconnecting switch includes combining the multi-scale permutation entropy feature vector, the energy distribution skewness, the peak factor, the fundamental frequency band energy ratio and the three times frequency band energy ratio into a feature vector set, inputting the current feature vector set into the trained classifier for pattern recognition, and comparing with the threshold set based on historical data, when the feature value exceeds the threshold range, it is determined as a specific fault type, and the automatic detection of the disconnecting switch defect is performed.
8. A vibration signal based disconnector defect detection system applying the vibration signal based disconnector defect detection method according to any one of claims 1 to 7, characterized in that It comprises a signal acquisition module, a signal preprocessing module, a signal decomposition module and a feature extraction module. The signal acquisition module is used to construct a monitoring network covering the key areas of the disconnecting switch by arranging vibration sensors at different positions of the GIS shell, and to collect vibration signals generated in the operation process in real time. The signal preprocessing module is used to process the original vibration signal by wavelet packet denoising and singular value decomposition denoising through a joint denoising strategy, to suppress environmental noise and electromagnetic interference, and to retain fault-related impact components. The signal decomposition module is used to adaptively decompose the denoised signal into intrinsic mode functions through CEEMDAN decomposition, and to perform VMD secondary decomposition on the screened high-frequency IMF components. The feature extraction module is used to extract multi-dimensional features such as correlation coefficient, multi-scale permutation entropy, energy entropy, envelope spectrum peak factor and fault-sensitive frequency band energy ratio from the decomposed signal components. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8. The processor executes the computer program to realize the steps of the disconnecting switch defect detection method based on vibration signals in any one of claims 1 to 7.
10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the disconnecting switch defect detection method based on vibration signals in any one of claims 1 to 7.