A mechanical pulse live warning system for insulators

Abstract

This application relates to the field of insulator fault early warning technology, specifically to an insulator mechanical pulse energized early warning system. The system includes: a pulse signal acquisition module, used to acquire vibration pulse signals from each sampling point at equal intervals at each height of the insulator; a vibration anomaly identification module, used to determine the lateral and longitudinal vibration anomalies of the vibration pulse signals at each sampling point; and a pulse signal denoising early warning module, used to obtain a standard comprehensive anomaly index of the vibration pulse signals at each sampling point; identify the non-low-frequency energy intensity of the vibration pulse signals at each sampling point, and adjust a fixed threshold in the wavelet denoising algorithm based on the standard comprehensive anomaly index, so as to use the denoised vibration pulse signals to provide early warning of insulator defects. This application improves the accuracy of insulator defect early warning.

Description

Technical Field

[0001] This application relates to the field of insulator fault early warning technology, specifically to an insulator mechanical pulse energization early warning system. Background Technology

[0002] In modern power systems, porcelain post insulators are key insulating components ensuring stable power transmission, and their performance directly affects the safe and reliable operation of the entire power grid. Affected by various complex factors, porcelain post insulators face severe challenges during long-term operation, necessitating regular inspection. Among these advancements, a new method based on vibration acoustics allows for live-line working, enabling non-destructive and rapid detection of internal defects in porcelain post insulators without disrupting power supply to local areas.

[0003] The vibration modes and dynamic responses generated by internal defects in porcelain post insulators differ significantly from those under defect-free conditions. Therefore, the vibroacoustic method identifies internal defects by applying excitation vibration signals to the porcelain post insulator and analyzing the feedback pulse signals. However, since porcelain post insulators typically operate in noisy environments, various noise interferences can easily be introduced during the vibration pulse data detection process, affecting the accuracy of internal defect identification.

[0004] Traditional wavelet transform denoising relies on a fixed denoising threshold derived from the characteristics of wavelet approximation coefficients to denoise wavelet decomposed signals at various scales. However, due to differences in the installation method and stress state of porcelain post insulators, the vibration pulse signals responding at different locations of the porcelain post insulator under mechanical pulse excitation vary significantly. Denoising the vibration pulse signals based on a fixed threshold would cause the effective signals that vary due to sampling at different locations of the porcelain post insulator to be filtered out as noise, affecting the accuracy of subsequent defect identification and early warning for porcelain post insulators. Summary of the Invention

[0005] To address the aforementioned technical problems, the purpose of this application is to provide an insulator mechanical pulse energization early warning system, the specific technical solution of which is as follows:

[0006] This application proposes an insulator mechanical pulse energization early warning system, the system comprising:

[0007] The pulse signal acquisition module is used to set sampling points at equal intervals at each height of the insulator and acquire the vibration pulse signal at each sampling point.

[0008] The vibration anomaly identification module is used to identify other sampling points at the same height as each sampling point as neighboring sampling points; analyze the difference in amplitude distribution of vibration pulse signals in the frequency domain between each sampling point and its neighboring sampling points, and determine the degree of lateral vibration anomaly of the vibration pulse signals of each sampling point.

[0009] The peak frequency difference of the vibration pulse signal in the frequency domain between each sampling point and its corresponding sampling point at the adjacent height is determined. The longitudinal vibration anomaly of the vibration pulse signal at each sampling point is determined by the degree of deviation of the peak frequency difference between each sampling point and its neighboring sampling points.

[0010] The pulse signal denoising and early warning module is used to correct the fusion result of the transverse vibration anomaly and the longitudinal vibration anomaly based on the energy decay rate of the vibration pulse signal at the sampling point, so as to obtain the standard comprehensive anomaly index of the vibration pulse signal at each sampling point.

[0011] The non-low frequency energy intensity of the vibration pulse signal at each sampling point is identified. Combined with the standard comprehensive anomaly index, the fixed threshold in the wavelet denoising algorithm is adjusted so as to use the denoised vibration pulse signal to provide early warning of insulator defects.

[0012] In one embodiment, determining the degree of lateral vibration anomaly includes:

[0013] The spectrum of vibration pulse signals at each sampling point is obtained, the maxima in the spectrum are identified, the distribution difference of the maxima of vibration pulse signals between each sampling point and its neighboring sampling points is evaluated, and the degree of transverse vibration anomaly is determined.

[0014] In one embodiment, a nonlinear fitting is performed on all the maxima of the vibration pulse signal at each sampling point to obtain a fitting curve, wherein the lateral vibration anomaly is a fused value of the difference distance between the fitting curves of the vibration pulse signals at each sampling point and all its neighboring sampling points.

[0015] In one embodiment, determining the peak frequency difference includes:

[0016] Calculate the mean of the frequencies corresponding to all maxima in the spectrum of the vibration pulse signal at each sampling point, whereby the peak frequency difference is the difference between the mean values ​​of each sampling point and its adjacent sampling points at the corresponding height.

[0017] In one embodiment, determining the longitudinal vibration anomaly degree includes:

[0018] The difference vector is formed by the peak frequency difference between each sampling point and all its adjacent height corresponding sampling points. The difference distance of the difference vector of the vibration pulse signal between each sampling point and its neighboring sampling points is determined. The longitudinal vibration anomaly degree is positively correlated with the difference distance.

[0019] In one embodiment, the determination of the standard composite anomaly index includes:

[0020] Calculate the peak half-width at most one peak in the vibration pulse signal at each sampling point. Based on the numerical distribution of the peak half-width at all sampling point vibration pulse signals, determine a threshold. If the peak half-width at any sampling point vibration pulse signal is less than or equal to the threshold, then the standard comprehensive anomaly index of the vibration pulse signal at any sampling point is 0; otherwise, the standard comprehensive anomaly index of the vibration pulse signal at any sampling point is the fusion result.

[0021] In one embodiment, determining the non-low-frequency energy intensity includes:

[0022] The vibration pulse signal at each sampling point is modally decomposed to obtain low-frequency, mid-frequency, and high-frequency components. The average energy ratio of the mid-frequency and high-frequency components in the low-frequency component is calculated, and the non-low-frequency energy intensity is the minimum value between the average value and the natural number 1.

[0023] In one embodiment, adjusting the fixed threshold in the wavelet denoising algorithm includes:

[0024] The sum of the standard comprehensive anomaly index and the natural number 1 is calculated. When denoising the vibration pulse signal at each sampling point, the sum and the non-low frequency energy intensity are used to correct the initial threshold of the wavelet denoising algorithm.

[0025] In one embodiment, the sum is calculated as the product of the sum and the non-low frequency energy intensity, and the threshold adjusted by the wavelet denoising algorithm is the result of multiplying the product with its initial threshold.

[0026] In one embodiment, the method of using the denoised vibration pulse signal to provide early warning of insulator defects includes:

[0027] A pre-trained classification model for insulator defect detection based on vibration pulse signals is obtained. The data features extracted from all vibration pulse signals of the insulator under test after denoising are used as the input of the classification model, and the defect detection results of the insulator under test are output.

[0028] This application has the following beneficial effects:

[0029] This application acquires vibration pulse signals by setting equally spaced sampling points at each height of the insulator. The system can accurately capture the real-time state of the insulator at different positions and effectively identify minute vibration changes, which helps to detect potential damage and faults in the insulator in advance. By determining the lateral and longitudinal vibration anomalies of the vibration pulse signals at each sampling point, the system can quickly respond to vibration changes, enhancing the ability to identify lateral and longitudinal vibration anomalies of the insulator. This helps to extract the distinctive features of vibration pulse signals at different sampling points of the insulator, providing a reliable basis for denoising vibration pulse signals. Furthermore, by combining non-low-frequency energy intensity and standard comprehensive anomaly index to adjust the fixed threshold in the wavelet denoising algorithm, the adaptability of the denoising algorithm to vibration pulse signals at multiple positions is significantly improved. This ensures that key vibration pulse signals can be effectively captured under various working conditions, effectively distinguishing the subtle differences between potential defects and normal vibrations, avoiding the loss of effective signals caused by fixed thresholds, and thus improving the accuracy of insulator defect early warning. Attached Figure Description

[0030] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 A block diagram of an insulator mechanical pulse energization early warning system provided in one embodiment of this application;

[0032] Figure 2 Block diagram for pulse signal denoising and early warning module implementation. Detailed Implementation

[0033] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an insulator mechanical pulse live-line warning system proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0035] The following description, in conjunction with the accompanying drawings, details a specific scheme for an insulator mechanical pulse live-line early warning system provided in this application.

[0036] Please see Figure 1 The diagram illustrates a block diagram of an insulator mechanical pulse energization early warning system according to an embodiment of this application. The system includes: a pulse signal acquisition module 101, a vibration anomaly identification module 102, and a pulse signal noise reduction and early warning module 103.

[0037] The pulse signal acquisition module 101 sets sampling points at equal intervals at each height of the insulator and acquires the vibration pulse signal at each sampling point.

[0038] This embodiment uses an insulated operating support rod, referred to as an insulated rod, to acquire the vibration pulse signal of the porcelain post insulator. The insulated operating support rod is a structural component consisting of a circular slot fastener, bolt fasteners, and a nut for adjusting and fixing the angle of the vibratory acoustic device, which includes an excitation generator and a vibration sensor. The porcelain post insulator is simply referred to as an insulator. The excitation generator can apply a step signal to the insulator for excitation, and the vibration sensor is used to acquire the insulator's response data to the excitation.

[0039] When acquiring the vibration pulse signal of the insulator, the vibratory acoustic device at the top of the insulating rod is kept parallel to the insulator. Vibration pulse signals are collected sequentially from top to bottom between the upper and lower sheds of the insulator. The upper shed refers to the midpoint between the upper flange and the first upper shed, and the lower shed refers to the midpoint between the first lower shed and the lower flange. Vibration pulse signals are collected evenly from top to bottom across the insulator. Vibration pulse signals at different horizontal heights are used, with the upper shed being the first sampling height and the lower shed being the last sampling height, thus dividing the area between the upper and lower sheds of the insulator into equally spaced intervals. In this embodiment, several points are set. The implementer can set it according to the actual situation, and this embodiment does not impose any restrictions on it.

[0040] In this embodiment, when collecting vibration pulse signals between the upper and lower skirts of the insulator, a sampling point is set at every 90 degrees on the same horizontal level. The vibration pulse signal of the insulator at each sampling point is collected. There are a total of 4 sampling points at different angles around the insulator at the same horizontal level, that is, N×4 sampling points are set from top to bottom on the insulator. In this embodiment, the sampling frequency of the vibration sensor on the insulating rod is 20kHz, and the single sampling duration is 1s. The implementer can set these values ​​as needed.

[0041] Vibration anomaly identification module 102, (1) takes other sampling points at the same height as each sampling point as the neighboring sampling points of each sampling point; analyzes the difference in amplitude distribution of vibration pulse signals in the frequency domain between each sampling point and its neighboring sampling points, and determines the transverse vibration anomaly degree of vibration pulse signals of each sampling point.

[0042] Different stress states on an insulator alter its equivalent stiffness, leading to significant differences in the response signal to applied mechanical pulse excitation at different locations. This causes the vibration pulse signal frequency to drift towards lower or higher frequencies. Traditional wavelet transform denoising algorithms use low-frequency wavelet approximation coefficients to set a fixed threshold for denoising the signal data. However, the energy distribution of the measured vibration pulse signal at different sampling points on the insulator varies considerably across different frequency bands. Using a fixed threshold for denoising can easily result in effective signals in localized areas being filtered out as noise, affecting the accuracy of subsequent insulator defect identification.

[0043] Based on the above analysis, firstly, the vibration pulse signal of a single insulator is obtained, and is represented as follows: , , , Let be the vibration pulse signal at the i-th sampling point at the n-th height of the insulator, where , representing the total number of sampling points around the insulator at the same horizontal height. For the i-th sampling point at the n-th height, the remaining I-1 sampling points at the same height are all taken as the neighboring sampling points of the i-th sampling point.

[0044] The shed structure of an insulator at the same horizontal height is a regular circle, but along the insulator axis, the size of the shed structure and the magnitude of stress may vary. Therefore, the spectrum of the insulator vibration pulse signal depends on the longitudinal vibration component and is almost unaffected by the transverse vibration component. That is, along the insulator axis, vibration pulse signals at different sampling points at the same horizontal height have relatively similar data characteristics. Therefore, the vibration pulse signal... The spectrum of the vibration pulse signal is obtained by performing a Fast Fourier Transform (FFT). Then, the peak detection algorithm AMPD (Automatic Multiscale-based Peak Detection) is used to acquire the vibration pulse signal. The maxima in the spectrum are further analyzed using the least squares method for the vibration pulse signal. All maxima in the frequency spectrum are fitted with a nonlinear polynomial function. Cross-validation is used to obtain the optimal order of the polynomial function fit, which is 3 in this embodiment, to finally obtain the frequency amplitude curve. Among them, Fast Fourier Transform, AMPD peak detection, and least squares method are all existing technologies, and the specific processes will not be described in detail. Implementers can choose other existing feasible nonlinear fitting algorithms at their own discretion, and this embodiment does not impose any restrictions on this.

[0045] Based on the above analysis, the vibration pulse signal is determined. The transverse vibration anomaly is specifically defined as follows: the fused value of the difference distance between the frequency amplitude curves of the vibration pulse signal of the i-th sampling point at the n-th height and all its neighboring sampling points is used as the vibration pulse signal. The degree of transverse vibration anomaly. It should be noted that fusion means combining multiple variables, which can be calculated by adding, multiplying, combining addition and multiplication, or taking the mean.

[0046] In this embodiment, the vibration pulse signal The expression for the transverse vibration anomaly degree is:

[0047] In the formula, Vibration pulse signal The degree of transverse vibration anomaly, This is the frequency amplitude curve of the vibration pulse signal at the i-th sampling point at the n-th height of the insulator. Let D be the frequency amplitude curve of the vibration pulse signal at the j-th sampling point at the n-th height of the insulator. D() is the difference distance calculation function, which can be calculated using Euclidean distance, DTW distance, Manhattan distance, etc. In this embodiment, DTW distance is used for calculation.

[0048] It should be understood that the presence of noise interference at sampling points is determined by assessing the degree of difference in the frequency amplitude curves of vibration pulse signals between sampling points at the same horizontal height. For those at the same insulator level... The vibration pulse signals at different sampling points at different horizontal heights have similar spectra, then the... a horizontal height The frequency amplitude curves at individual sampling points show high similarity; however, if the vibration pulse signal at a sampling point is subject to significant noise interference, it will cause amplitude anomalies in its local frequency band, resulting in low similarity between the frequency amplitude curves of that sampling point and other sampling points at the same height. This indicates the degree of lateral vibration anomaly. When it is smaller, it indicates a vibration pulse signal. A medium level indicates less noise interference, while a high level indicates stronger noise interference.

[0049] (2) Determine the peak frequency difference of the vibration pulse signal in the frequency domain between each sampling point and its adjacent sampling point. Determine the longitudinal vibration anomaly of the vibration pulse signal at each sampling point by the degree of deviation of the peak frequency difference between each sampling point and its neighboring sampling points.

[0050] After installation, insulators are mainly affected by the tensile stress of the conductor and their own weight. This causes the tensile stress on the insulator to gradually increase from the upper flange to the lower flange. The magnitude of the stress will change the equivalent stiffness of the structure. If the tensile stress is too strong, it will reduce the equivalent stiffness of the insulator, and the response signal to mechanical pulse excitation will be more easily attenuated in the high-frequency components. This causes the peak resonant frequency in the vibration pulse signal of the area near the lower flange to shift to the low-frequency region. On the other hand, the equivalent stiffness of the area near the upper flange is stronger, and the peak resonant frequency in its vibration pulse signal will shift to the high-frequency region.

[0051] Therefore, when obtaining vibration pulse signals When the frequency of each maximum value is obtained in the spectrum graph, the frequency value corresponding to each maximum value is obtained. To avoid the maximum values ​​containing a large amount of noise spikes, the vibration pulse signal is... Arrange the frequency values ​​corresponding to all maxima in the spectrum in descending order, obtain the frequency values ​​corresponding to the first K maxima, and calculate the vibration pulse signal. The mean of the frequency values ​​corresponding to the first K maxima in the spectrum. Then, obtain the vibration pulse signal at the (n+1)th height. The mean of the frequencies corresponding to the first K maxima of vibration pulse signals at the same angle, arranged in descending order. and the vibration pulse signal at the (n-1)th height. The mean of the frequencies corresponding to the first K maxima of vibration pulse signals at the same angle, arranged in descending order. The mean respectively with the mean mean The difference between them is taken as the peak frequency difference. In this embodiment, K=10 is set, but implementers can set it according to their actual situation.

[0052] Vibration pulse signal The differences in all corresponding peak frequencies form a difference vector. Specifically, in this embodiment, the vibration pulse signal... The difference vector is In another embodiment, the vibration pulse signal The difference vector can be Specifically, if the vibration pulse signal... When the height is the first height, then in this embodiment its difference vector is Vibration pulse signal When the height is the last height, its difference vector in this embodiment is .

[0053] Based on the above analysis, the vibration pulse signal is determined. The longitudinal vibration anomaly is expressed as follows:

[0054] In the formula, Vibration pulse signal longitudinal vibration anomaly Vibration pulse signal The difference vector, Let be the difference vector of the vibration pulse signal at the j-th sampling point at the n-th height of the insulator.

[0055] Through calculation It can quantify vibration pulse signals. The data characteristics are examined to determine if the variations along the longitudinal direction of the insulator are similar to those at other sampling points at the same horizontal level. If the vibration components remain similar across transverse measurement points at the same horizontal level, then the variations in the vibration components along the longitudinal direction of the insulator should also remain similar in the absence of noise interference. This relates to the longitudinal vibration anomaly. The longitudinal vibration anomaly should be relatively small; conversely, if there is significant noise interference at the sampling point, it will disrupt the similarity of this change. The value is relatively large.

[0056] The pulse signal denoising and early warning module 103 (1) corrects the fusion result of the transverse vibration anomaly degree and the longitudinal vibration anomaly degree based on the energy decay rate of the vibration pulse signal at the sampling point, and obtains the standard comprehensive anomaly index of the vibration pulse signal at each sampling point.

[0057] The transverse and longitudinal vibration anomalies reflect the degree of abnormality in the vibration mode of the acquired vibration pulse signals as the measurement position changes. Larger transverse and longitudinal vibration anomalies indicate significant noise interference at the sampling point, leading to abnormal changes in the vibration mode. Therefore, the initial comprehensive anomaly index of the vibration pulse signals at each sampling point is calculated to reflect the degree of noise interference. The initial comprehensive anomaly index is positively correlated with both the transverse and longitudinal vibration anomalies.

[0058] First, the transverse and longitudinal vibration anomalies of the vibration pulse signals at all sampling points of the insulator are normalized using the maximum-minimum normalization method, resulting in normalized transverse and longitudinal vibration anomalies for each sampling point. Based on these normalized transverse and longitudinal vibration anomalies, the initial comprehensive anomaly index of the vibration pulse signals at each sampling point is calculated. In this embodiment, the vibration pulse signal... The initial comprehensive anomaly index is calculated as follows:

[0059] In the formula, Vibration pulse signal The initial comprehensive anomaly index, Vibration pulse signal Normalized transverse vibration anomaly Vibration pulse signal Normalized longitudinal vibration anomaly As the first weight, As the second weight, satisfying Its size can be set by the implementer according to the relative importance of the lateral vibration anomaly degree and the longitudinal vibration anomaly degree. In this embodiment, .

[0060] It should be understood that when the vibration mode of the vibration pulse signal is subject to significant noise interference, the threshold needs to be increased to improve the denoising capability of the wavelet transform during subsequent wavelet denoising.

[0061] Furthermore, both defects and noise in the insulator can cause abnormal frequency changes in the acquired vibration pulse signal. However, when defects cause abnormal signal changes, they manifest as rapid attenuation of pulse response energy in the time domain. For example, the presence of cracks in the insulator surface differs from that of an intact insulator, causing the vibration pulse energy to decay rapidly. Noise also causes abnormal frequency changes in the vibration pulse signal, but it manifests as a scattered and irregular distribution in the time domain and does not significantly affect the energy attenuation rate of the vibration pulse. Based on the above analysis, the vibration pulse signal is calculated. Half-width at half-peak of the highest peak in the waveform The peak half-width at half-maximum (HWHM) reflects the energy decay rate of the strongest pulse response in the vibration pulse signal. If the HWHM is large, it indicates that the energy decay rate of the pulse response is slow; conversely, if the HWHM is small, it indicates that the energy decay rate of the pulse response is fast.

[0062] Therefore, the peak half-width at half-maximum (FWHM) of the highest peak in the vibration pulse signal waveform at all sampling points on the insulator is calculated, and the threshold values ​​for all peak FWHMs are calculated. In this embodiment ,in, This represents the average half-width at half-maximum (WHM) of the peak in the vibration pulse signal waveform at all sampling points on the insulator. Because defects cause rapid decay of the pulse response energy, the WHM decreases, therefore... The situation.

[0063] Based on the different effects of insulator defects and noise on the vibration pulse signal in the time domain, the initial comprehensive anomaly index is corrected to obtain the vibration pulse signal. Standard composite anomaly index Specifically:

[0064]

[0065] when When the abnormal change in the vibration pulse signal is due to a defect, the initial comprehensive anomaly index is [value missing]. It is impossible to accurately determine whether abnormal vibration pulse signals are caused by noise interference; therefore, a standard comprehensive anomaly index is set. If the value is 0, the threshold of the subsequent wavelet transform will not be changed; otherwise, if the energy in the vibration pulse signal does not decay rapidly, the abnormal change in the frequency of the vibration pulse signal may be caused by noise. In this case, the initial comprehensive anomaly index is kept as the standard comprehensive anomaly index to adjust the threshold of the wavelet transform denoising algorithm.

[0066] (2) Identify the non-low frequency energy intensity of the vibration pulse signal at each sampling point, and adjust the fixed threshold in the wavelet denoising algorithm in combination with the standard comprehensive anomaly index, so as to use the denoised vibration pulse signal to provide early warning of defects in the insulator.

[0067] In the frequency domain, the energy of each frequency band of a vibration pulse signal is guided by its resonant frequency peak. Even with noise interference, the noise effect will not significantly disrupt the distribution of resonant frequency peaks in each frequency band. The distribution of resonant frequency peaks of vibration pulse signals varies considerably at different locations on the upper and lower parts of the insulator. For example, the vibration frequencies at the lower part of the insulator are mainly distributed in the low-frequency region. Regarding vibration pulse signals... After wavelet decomposition, the wavelet transform denoising threshold is obtained by fixing the threshold based on the wavelet approximation coefficients, i.e., the signal characteristics in the low-frequency band. However, if the threshold is too high, useful signal features in the vibration pulse signal will be eliminated. Therefore, this embodiment also considers the distribution of the resonant frequency peak of the vibration pulse signal at different locations of the insulator and adaptively adjusts the wavelet transform threshold accordingly.

[0068] First, empirical mode decomposition (EMD) is used to analyze the vibration pulse signal. Decompose to obtain There are 1 IMF component, of which... Taking a value of 3, the IMF components in the low, medium, and high frequency bands are obtained respectively, denoted as . , Empirical mode decomposition is performed using existing well-known techniques; the specific process will not be elaborated here. The energy of each IMF component is calculated based on the IMF component signals. The calculation of signal energy is a current technique, and will not be elaborated further.

[0069] Evaluation of vibration pulse signals The non-low frequency energy intensity is specifically expressed as:

[0070] In the formula, Vibration pulse signal The non-low frequency energy intensity, min[] is the minimum value function. Vibration pulse signal The energy of the low-frequency IMF component, Vibration pulse signal The energy of the mid-frequency IMF component, Vibration pulse signal The energy of the high-frequency IMF component, To avoid a denominator of 0, a very small value greater than 0 is preset in this embodiment. The implementer can set it according to the actual situation, and this embodiment does not impose any restrictions on it. This indicates the energy proportion of the mid-frequency IMF component in the low-frequency IMF component. This indicates the energy proportion of the high-frequency IMF component in the low-frequency IMF component.

[0071] The vibration pulse signal is quantified by the ratio of mid- and high-frequency energy to low-frequency energy. The proportion of energy distribution in the low-frequency band. When the vibration pulse signal The energy is mainly concentrated in the low-frequency region, and is higher than that in the mid- and high-frequency regions. Therefore, the vibration pulse signal... The difference in sampling points on the insulator causes the low-frequency signal distribution to be dominated by the resonant frequency peak, resulting in higher low-frequency energy. At this time, the non-low-frequency energy intensity... The threshold is relatively small, and it needs to be lowered in subsequent steps to avoid setting the threshold too high for wavelet transform denoising based on the characteristics of low-frequency signals. This would cause the effective information in the vibration pulse signal to be filtered out, affecting the authenticity of the vibration pulse signal; when the low-frequency energy When the energy is less than or equal to the mid-to-high frequency range, the non-low frequency energy intensity is set to... An equal value of 1 indicates a vibration pulse signal. There was no issue of low-frequency energy being too high due to differences in sampling location; at this time, the vibration pulse signal... The vibration pulse signal is denoised by maintaining the original threshold of wavelet transform denoising when the energy distribution is uniform or mainly in the mid-to-high frequency band.

[0072] In summary, the vibration pulse signal was determined. The adaptive threshold for wavelet transform denoising is expressed as follows: ,in, This is a fixed threshold in wavelet transform denoising, and its expression will not be elaborated here.

[0073] When denoising the vibration pulse signals at each sampling point, the number of decomposition levels for wavelet transform denoising is... Its size can be set by the implementer according to the sampling frequency in the implementation scenario, without special restrictions. If the sampling frequency is large, then... The value can be appropriately increased to perform more refined signal decomposition and noise reduction. In this embodiment... The value is set to 3. The acquired insulator vibration pulse signal is denoised based on the adaptive threshold during wavelet transform denoising. Wavelet transform denoising is a well-known existing technology, and its specific process will not be described in detail.

[0074] The denoised vibration pulse signals of the insulator are transformed using the Mel-Cepstral Coefficient method to obtain the feature matrix of each vibration pulse signal, thus completing the feature extraction of the vibration pulse signals. The Mel-Cepstral Coefficient method is an existing technology, and its specific process will not be described in detail here.

[0075] Furthermore, a dataset is constructed by pre-collecting vibration pulse signals from both intact and defective insulators using a vibratory acoustic device. The vibration pulse signals in the dataset are then denoised using the aforementioned denoising method. Next, the Mel-frequency cepstral coefficient method is used to extract features from the denoised vibration pulse signals in the dataset, resulting in a training set. The classification model is trained using this training set. In this embodiment, the classification model is a Support Vector Machine (SVM) model, the model optimizer is the Adam optimizer, and the model loss function is cross-entropy loss. Implementers can choose other classification models, such as decision trees or random forests. The training of the Support Vector Machine (SVM) model is a well-known technique, and the specific process will not be elaborated upon. Each data point in the training set represents a feature matrix extracted from all vibration pulse signals of a single insulator.

[0076] The feature matrix of the extracted vibration pulse signal from the insulator under test is used to detect and identify whether there are defects in the insulator. If a defect is found, an early warning is issued for the insulator, and timely repairs are carried out. The block diagram of the pulse signal denoising and early warning module is shown below. Figure 2 As shown.

[0077] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0078] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0079] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A mechanical pulse live-line warning system for insulators, characterized in that, The system includes: The pulse signal acquisition module is used to set sampling points at equal intervals at each height of the insulator and acquire the vibration pulse signal at each sampling point. The vibration anomaly identification module is used to identify other sampling points at the same height as each sampling point as neighboring sampling points; analyze the difference in amplitude distribution of vibration pulse signals in the frequency domain between each sampling point and its neighboring sampling points, and determine the degree of lateral vibration anomaly of the vibration pulse signals of each sampling point. The peak frequency difference of the vibration pulse signal in the frequency domain between each sampling point and its corresponding sampling point at the adjacent height is determined. The longitudinal vibration anomaly of the vibration pulse signal at each sampling point is determined by the degree of deviation of the peak frequency difference between each sampling point and its neighboring sampling points. The pulse signal denoising and early warning module is used to correct the fusion result of the transverse vibration anomaly and the longitudinal vibration anomaly based on the energy decay rate of the vibration pulse signal at the sampling point, so as to obtain the standard comprehensive anomaly index of the vibration pulse signal at each sampling point. The non-low frequency energy intensity of the vibration pulse signal at each sampling point is identified. Combined with the standard comprehensive anomaly index, the fixed threshold in the wavelet denoising algorithm is adjusted so as to use the denoised vibration pulse signal to provide early warning of insulator defects. The determination of the transverse vibration anomaly degree includes: The spectrum of vibration pulse signal at each sampling point is obtained, the maximum value in the spectrum is identified, and nonlinear fitting is performed on all the maximum values ​​of vibration pulse signal at each sampling point to obtain the fitting curve. The transverse vibration anomaly is the fusion value of the difference distance between the fitting curve of vibration pulse signal at each sampling point and all its neighboring sampling points. The determination of the peak frequency difference includes: Calculate the mean value of the frequencies corresponding to all maxima in the spectrum of the vibration pulse signal at each sampling point, where the peak frequency difference is the difference between the mean values ​​of each sampling point and its adjacent sampling points at the corresponding height. The determination of the longitudinal vibration anomaly degree includes: The difference vector is formed by the peak frequency difference between each sampling point and all its adjacent height corresponding sampling points. The difference distance of the difference vector of the vibration pulse signal between each sampling point and its neighboring sampling points is determined. The longitudinal vibration anomaly is positively correlated with the difference distance. The determination of the standard comprehensive anomaly index includes: Calculate the peak half-width at most peak of the vibration pulse signal at each sampling point. Based on the numerical distribution of the peak half-width at all sampling points, determine a threshold. If the peak half-width at any sampling point is less than or equal to the threshold, the standard comprehensive anomaly index of the vibration pulse signal at any sampling point is 0. Otherwise, the standard comprehensive anomaly index of the vibration pulse signal at any sampling point is the fusion result. The determination of the non-low frequency energy intensity includes: The vibration pulse signal at each sampling point is modally decomposed to obtain low-frequency, mid-frequency, and high-frequency components; the average energy proportion of the mid-frequency and high-frequency components in the low-frequency component is calculated, and the non-low-frequency energy intensity is the minimum value between the average value and the natural number 1. The adjustment of the fixed threshold in the wavelet denoising algorithm includes: The sum of the standard comprehensive anomaly index and the natural number 1 is calculated, and the product of the sum and the non-low frequency energy intensity is calculated. The threshold adjusted by the wavelet denoising algorithm is the result of multiplying the product with its initial threshold.

2. The insulator mechanical pulse live-line early warning system according to claim 1, characterized in that, The method of using denoised vibration pulse signals to provide early warning of insulator defects includes: A pre-trained classification model for insulator defect detection based on vibration pulse signals is obtained. The data features extracted from all vibration pulse signals of the insulator under test after denoising are used as the input of the classification model, and the defect detection results of the insulator under test are output.

Images