Method for analyzing the erosion residual value of high-voltage direct-current contactor contacts based on acoustic signals

Abstract

The application is a high-voltage direct-current contactor contact erosion residual value analysis method based on an acoustic signal, including the following contents: collecting the acoustic signal sequence of the contact closure stage of the high-voltage direct-current contactor under different contact erosion degrees, and extracting the contact collision acoustic signal segment containing the topographic information; using the wavelet soft threshold method to complete low-frequency denoising, and then performing modal decomposition through the symplectic geometry modal decomposition to obtain a series of symplectic geometry modal components; selecting the mode based on the EW-TOPSIS evaluation mode of the adaptive distance calculation module to obtain the mode reconstruction acoustic signal containing the contact topographic sensitive component which can effectively represent the contact erosion degree; and then performing Hilbert transformation to obtain the skewness value of the Hilbert envelope spectrum for quantifying the contact erosion residual value. The application realizes the effective representation of the contact topographic parameters through the acoustic signal, and realizes the nondestructive detection of the contact erosion residual value.

Description

Technical Field

[0001] This invention belongs to the field of contact erosion residual value analysis technology, specifically a method for analyzing the erosion residual value of high-voltage DC contactors based on acoustic signals. Background Technology

[0002] With the rapid development of renewable energy, high-voltage DC contactors face severe challenges in miniaturization, high power, high reliability, and long lifespan. In scenarios such as electric vehicle collisions, high-voltage DC contactors need to frequently withstand high-voltage and high-current surges and quickly disconnect circuits in emergency situations to curb fault propagation, leading to a continuous increase in contact load. Among these challenges, contact electrical contact failure has become the main failure mode of contactors, and its surface morphology, as a core parameter affecting the electrical contact state, has become a key entry point for degradation analysis.

[0003] Existing research studies the degradation process of contact electrical contacts by analyzing the surface morphology of the contacts. For example, Li Kui et al. (Li Kui, Zhang Yue, Jiang Hui, et al. Reliability assessment of low-voltage DC circuit breakers based on contact morphology features [J]. Journal of Instrumentation, 2024, 45(12): 118-128.) used an adaptive multi-threshold segmentation algorithm based on image block weighting to segment the contact image, extract the area and centroid of the binary image of the region of interest, and thus establish a multi-stage Wiener degradation model with binary feature correlation. This model was verified by electrical life test. However, this study relies on a three-dimensional morphology instrument and requires the violent destruction of the sealed structure of the arc-extinguishing chamber, which affects the subsequent use of the contactor. Currently, the lack of a measurement method that can effectively and non-destructively characterize the erosion morphology of contacts has become a key technical bottleneck restricting the residual value analysis of electric vehicle high-voltage DC contactors (or high-voltage DC relays), which urgently needs to be overcome.

[0004] In recent years, in the condition monitoring of high-voltage DC contactors, acoustic signals have been used for feature extraction. For example, Sun Shuguang et al. (Sun Shuguang, Wang Zihang, Wang Jingqin, et al. Measurement of characteristic parameters of high-voltage DC contactors based on sub-band average kurtosis diagram and TMSST [J]. Journal of Instrumentation, 2025, 46(6):117-129.) proposed a measurement method based on sub-band average kurtosis diagram and time redistribution multi-synchronous compression transform (TMSST) to measure key dynamic characteristic parameters such as contactor engagement time and overtravel time. However, current research focuses more on monitoring the overall condition of the contactor. Compared with the impact effect of mechanical structure in acoustic signals, the effect of contact morphology degradation is relatively weak. Therefore, there is little research on the residual value analysis of contact erosion of high-voltage DC contactors based on acoustic signals. The main difficulties are as follows: First, because the degradation of the contact morphology has a relatively weak impact on the acoustic signal, and there is overlap of multiple events in the acoustic signal, it is difficult to extract the sensitive components of the contact erosion morphology; second, because it is difficult to capture the key information directly related to the contact erosion, the accuracy of the extracted parameters characterizing the degree of contact erosion is low. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the technical problem this invention aims to solve is to provide a method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals. This method effectively characterizes the contact morphology parameters through acoustic signals, enabling non-destructive testing of contact erosion residual values.

[0006] The technical solution adopted by the present invention to solve the aforementioned technical problem is as follows:

[0007] This invention provides a method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals. The method includes the following steps:

[0008] Step 1: Collect acoustic signal sequences of the contact closure phase of the high-voltage DC contactor under different contact erosion levels;

[0009] Step 2: Accurate identification of contact collision events based on the dual threshold method of short-time energy, and extraction of contact collision sound signal segments containing morphological information;

[0010] Step 3: Extract acoustic signal components sensitive to contact morphology:

[0011] Low-frequency denoising is achieved by using wavelet soft thresholding. Then, symplectic geometric mode decomposition (SGMD) is used to perform mode decomposition on the denoised contact collision sound signal segment containing morphological information to obtain a series of symplectic geometric mode components.

[0012] The EW-TOPSIS evaluation method based on an adaptive distance calculation module selects modes: Bach distance is used as a measure of the similarity between each modal component and the original signal, and kurtosis is used as a measure of sensitivity to contact collision events; the entropy weight method is used to calculate the information entropy and entropy weight of each measure; during the evaluation process, the correlation coefficient between the measures is used... ρ Standard deviation of relative closeness to each mode σ To determine the characteristics of the metric data and automatically trigger the selection of the most suitable distance metric algorithm; if there is a strong correlation between the metrics, Mahalanobis distance is selected as the distance metric algorithm; if the metrics are highly independent and the closeness between modes is too concentrated, Euclidean distance is switched to symmetric cross-entropy distance as the distance metric algorithm.

[0013] Calculate the distance between each mode and the positive and negative ideal solutions to obtain the final relative closeness of each mode;

[0014] Set a filtering threshold, select the corresponding modal components whose final relative proximity is greater than the filtering threshold for reconstruction, and obtain a modal reconstruction acoustic signal containing contact morphology sensitive components that can effectively characterize the degree of contact erosion.

[0015] Step 4: Perform Hilbert transform on the modal reconstruction acoustic signal containing the contact morphology-sensitive component to obtain the skewness value of the Hilbert envelope spectrum. Use the skewness value of the Hilbert envelope spectrum as the acoustic signal characteristic parameter characterizing the contact morphology state to quantify the contact erosion residual value.

[0016] Furthermore, the process of quantifying the contact erosion residual value is as follows: obtain the skewness values ​​of the Hilbert envelope spectrum of the new sample and the failed sample, and use them as the maximum and minimum values ​​of the skewness values ​​of the Hilbert envelope spectrum, respectively, to obtain the skewness value of the Hilbert envelope spectrum of the sample to be monitored. By uniformizing the calculation results, an intuitive and quantitative contact erosion residual value is obtained.

[0017] Furthermore, the process for determining the skewness value of the Hilbert envelope spectrum is as follows:

[0018] First, construct the envelope signal using the signal after Hilbert transform. Then, obtain the Hilbert envelope spectrum by performing a Fourier transform on the envelope signal, and finally calculate the skewness value of the Hilbert envelope spectrum.

[0019] Furthermore, the specific process of the second step is as follows:

[0020] First, the acoustic signal sequence acquired in the first step is processed by framing using a Hanning window of predetermined length. The short-time energy of each frame is calculated, an energy threshold is set, and the first point where the short-time energy exceeds the energy threshold is identified as the attraction moment. t1 ;

[0021] Then, from t 1 Begin by finding the energy peaks, setting the minimum distance between peaks, and the distance between peaks and... t 1 The minimum distance is used as a constraint condition; the overtravel time is... t 2 The threshold is set to 0.6 times the peak energy level to accurately pinpoint the overtravel moment. t 2 ;

[0022] Finally, extract the data from the moment of attraction. t 1 Overrun time t 2 The acoustic signal segment is a contact collision acoustic signal segment containing shape information.

[0023] Furthermore, the specific process of the EW-TOPSIS evaluation method based on the adaptive distance calculation module is as follows: Each modal component is considered an evaluation object, and the metric set for each modal component includes Bhattacharyya distance and kurtosis. A decision matrix is ​​constructed with the modal components as rows and the metric indicators as columns. W，

[0024] The Bhattacharyya distance is forwarded using a subtraction transformation method, unifying all metrics to a maximum size; then, the forwarded matrix is ​​vector normalized to obtain a standardized matrix. Z ;

[0025] The entropy weight vector is obtained by calculating the entropy weight method. w And use the obtained entropy weights to apply to the standardized matrix Z Weighting is performed to form a weighted normalized decision matrix. V ;

[0026] Based on the selection of the adaptive distance calculation module, the distance between each mode and the positive and negative ideal solutions is calculated, and finally the final relative closeness of each mode is obtained;

[0027] The adaptive distance calculation module is used to automatically trigger the selection of the most suitable distance metric algorithm based on the characteristics of the metric data. The specific implementation process is as follows:

[0028] Construct a library of alternative distance functions, including Mahalanobis distance, Euclidean distance, and symmetric interaction entropy distance. Based on the weighted normalized decision matrix, calculate the correlation coefficient between two metrics. ρ ;like Then, Mahalanobis distance is chosen to calculate the distance between each mode and the positive and negative ideal solutions;

[0029] like First, Euclidean distance is used for pre-calculation to obtain the standard deviation of the relative closeness of each pre-calculated mode. σ, like σ< If the value is 0.05, the system will automatically switch to symmetric cross-entropy distance to calculate the distance between each mode and the positive and negative ideal solutions; if... σ If the value is ≥0.05, then Euclidean distance will still be used to calculate the distance between each mode and the positive and negative ideal solutions.

[0030] The present invention also protects a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the steps of the method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals.

[0031] Compared with the prior art, the beneficial effects of the present invention are:

[0032] (1) This invention focuses on the degree of contact erosion. To extract key information about the degree of contact erosion from the acoustic signal, morphology-related components are extracted as the main research object, rather than interference factors are removed. The skewness value of the Hilbert envelope spectrum is used as a characteristic parameter to characterize the contact morphology, realizing non-invasive, high-accuracy, and non-destructive testing of the degree of contact erosion of high-voltage DC contactors. Analyzing the contact erosion state using the acoustic signal of high-voltage DC contactor contact movement, unlike traditional three-dimensional morphology analysis, not only avoids the need for irreversible destructive testing of the contactor, but also achieves non-destructive testing of the contactor erosion residue while ensuring accuracy. This effectively reduces the waste of resources caused by blindly replacing contactors, and solves the problem of traditional contactors being difficult to maintain based on condition, improving the accuracy and economy of equipment maintenance.

[0033] (2) Improved accuracy of status monitoring: The EW-TOPSIS evaluation method based on the adaptive distance calculation module proposed in this invention, with its adaptive distance selection logic, can improve the accuracy of status monitoring based on the correlation coefficient of the measurement index (when...). ρ When the value is greater than 0.7, Mahalanobis distance is used to eliminate the interference of correlation between metrics) and the standard deviation of modal closeness (when σ When the distance is less than 0.05, the method switches to symmetric cross-entropy distance to improve discriminative power. This dynamic selection of a suitable distance metric algorithm effectively solves the analysis error problem caused by the fixed distance metric mode in the traditional EW-TOPSIS method. Results show that the evaluation results of the method in this invention are consistent with the peak material quantity. V mpThe method achieves higher accuracy, while the traditional EW-TOPSIS method exhibits significant fluctuations and biases, further validating the superior accuracy of the proposed method in contact erosion residual value analysis. This method provides reliable data support for the condition monitoring of high-voltage DC contactors, significantly improving monitoring accuracy and stability, and reducing the risk of misjudgment during operation and maintenance. Furthermore, it eliminates the need for manual parameter adjustments, adapts to acoustic signal characteristics under different operating conditions, and possesses strong engineering practicality, providing crucial technical support for the full lifecycle management of contactors.

[0034] (3) Provides an effective metric for quantitative evaluation: Based on the morphology of the envelope spectrum under different erosion degrees, this invention uses the skewness value of the homogenized Hilbert envelope spectrum to quantify the residual value of contact erosion. The results show that, in terms of correlation with peak material quantity, the frequency domain waveform characteristic parameters FC and MSF are both negatively correlated, and the absolute values ​​of the correlation coefficients are not greater than 0.80. This indicates that the correlation between these two parameters and peak material quantity is weak and difficult to accurately reflect the degree of contact erosion. In contrast, the skewness of the Hilbert envelope spectrum is not only positively correlated with peak material quantity but also has a very strong correlation. The method of this invention achieves accurate and reliable quantification of contact life, which is beneficial for engineering applications and further improves the monitoring level and maintenance quality of high-voltage DC contactor operation status. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the overall process of the high-voltage DC contactor contact erosion residual value analysis method based on acoustic signals according to the present invention.

[0036] Figure 2 This is a flowchart illustrating the adaptive distance calculation module.

[0037] Figure 3 This is a segmentation diagram of acoustic events during the contact collision stage, which is a schematic diagram of the process of extracting contact collision acoustic signal segments containing morphological information.

[0038] Figure 4 The image shows a comparison before and after wavelet soft thresholding denoising.

[0039] Figure 5 This is a schematic diagram of symplectic geometric mode decomposition (SGMD).

[0040] Figure 6 This is a schematic diagram of the measurement results of EW-TOPSIS for a single measurement.

[0041] Figure 7 Hilbert envelope spectra of the sound produced by the collision of contact morphology under different degrees of erosion.

[0042] Figure 8 A comparison of the effects of different methods for extracting acoustic signal components that are sensitive to morphology. Detailed Implementation

[0043] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments, but this is not intended to limit the scope of protection of the present invention.

[0044] This invention provides a method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals (hereinafter referred to as the method), comprising the following steps:

[0045] Step 1: Build a test platform for residual value analysis of contact morphology of high voltage DC contactors and collect contactor operation sound signals with different degrees of contact morphology degradation.

[0046] Step 2: Based on the short-time energy dual threshold method (setting dual empirical thresholds: energy threshold and threshold threshold), the boundary of the contact collision sound signal segment is accurately identified, so as to realize the accurate identification of the contact collision event and extract the contact collision sound signal segment containing morphological information.

[0047] 2-1 The acoustic signal sequence acquired in the first step is framed using a window function of predetermined length. The formula for the short-time energy of a frame audio signal is as follows:

[0048] (1)

[0049] in, For the Hamming window function, L For the length of the Hanming window, This is a local index within the Hamming window function. ; Indicates the first The short-time energy of a frame, ; inc The frame shift length, This represents the total number of frames after framing. This is the sequence of acoustic signals collected in the first step.

[0050] 2-2 The absorption moment and overtravel moment are calculated using the short-time energy of the acoustic signal. A sharp increase in energy occurs at both the absorption moment and the overtravel moment. Setting the energy threshold to 10000J, the first point where the short-time energy exceeds 10000J is identified as the absorption moment. t 1 .

[0051] 2-3 Considering the time of overtravel t 2 The energy is not necessarily the maximum value, firstly from the moment of attraction. t 1 The `findpeaks` function is used to locate energy peaks, with a minimum distance of 10 frames between peaks, and the peak values ​​are set to...t 1 The minimum distance is 150 frames, which avoids the peak distance search. t 1 Too short and it will cause misidentification. t 2 The threshold value for time is set to 0.6 times the peak energy level to find the overtravel time. t 2 .

[0052] 2-4 Acquisition of acoustic signals at the moment of absorption t 1 Up to the time of overrun t 2 From the acoustic signal segments, we obtain contact collision acoustic signal segments containing morphological information.

[0053] Step 3: Extract acoustic signal components sensitive to contact morphology:

[0054] Low-frequency denoising is achieved using wavelet soft thresholding. Then, mode decomposition of the denoised contact collision acoustic signal segment containing morphology information is performed using symplectic geometric mode decomposition (SGMD). The acoustic signal mode reconstruction containing contact morphology-sensitive components is achieved using EW-TOPSIS (entropy weighted superior solution distance method) based on an adaptive distance calculation module.

[0055] In scenarios where metrics such as acoustic signal mode screening may exhibit complex correlations or require high discriminative power, traditional methods struggle to dynamically select the most suitable distance metric, thus impacting the accuracy and effectiveness of the final evaluation results. This invention introduces an adaptive distance calculation module, enabling it to intelligently select the optimal distance metric algorithm based on the inherent characteristics of the dataset, significantly improving the discriminative power and robustness of the evaluation results.

[0056] 3-1 The wavelet soft thresholding method is used to solve the problem of baseline drift caused by low-frequency noise interference.

[0057] First, the normalized acoustic signal sequence is obtained through the normalization preprocessing of the second step. Then, the wavelet basis was determined to be db4 and the decomposition scale to be 7. This was achieved through... Wavelet coefficients are obtained by wavelet decomposition. Then, the threshold function shown in formula (2) is used to obtain the result. :

[0058] (2)

[0059] In the formula, These are wavelet coefficients; The processed wavelet coefficients are denoted by sgn, which is the sign function. The threshold value is used.

[0060] Finally, Reconstruction is performed to obtain an acoustic signal with baseline drift removed, thereby achieving wavelet soft thresholding denoising.

[0061] 3-2 The symplectic geometric mode decomposition is performed on the denoised contact collision sound signal segment containing morphological information using wavelet soft thresholding, yielding the result... The set of symplectic geometric modal components. ;

[0062] in, Represents the set of symplectic geometric modal components; =1,2,…, Indexes for each modal component.

[0063] Assuming each modal component is an evaluation object, there are a total of There are 10 modal components to be evaluated. The metric for each modal component is... ,in Index of metrics (in this embodiment) =1, 2, representing Bhattacharyya distance and kurtosis, respectively. A decision matrix is ​​constructed with modal components as rows and metrics as columns. W :

[0064] (3)

[0065] in, Indicates the first The Bach distance values ​​of each modal component. Indicates the first The kurtosis value of each modal component.

[0066] Since the Bhattacharyya distance is a very small metric, a subtraction transformation method is used to perform a positive transformation on it. The kurtosis metric itself is extremely large, and after positive transformation... This unifies all metrics to an extremely large size for subsequent TOPSIS analysis. Then, vector normalization is performed on the normalized matrix to obtain the standardized matrix. Z .

[0067] (4)

[0068] in, Indicates the first The modal component in the ... A standardized value under a positively quantified metric; This is the metric after positive transformation.

[0069] Secondly, calculate the proportion of each category of metrics in the overall information entropy of metrics. The first metric under the The contribution (proportion) of each modal component. The calculation is as follows:

[0070] (5)

[0071] No. Information entropy of a metric The calculation is as follows:

[0072] (6)

[0073] No. Entropy weight of each metric The calculation is as follows:

[0074] (7)

[0075] Finally, the entropy weight vector is obtained. And use the obtained entropy weights to apply to the standardized matrix Z Weighting is performed to form a weighted normalized decision matrix. V The calculation is as follows:

[0076] (8)

[0077] in, Indicates the first The modal component in the ... The weighted values ​​under each metric.

[0078] The logical flow of the adaptive distance calculation module is as follows: Figure 2 As shown, its core is to automatically trigger the selection of the most suitable distance metric algorithm based on the characteristics of the metric data.

[0079] First, a candidate distance function library is constructed, integrating at least three distance metric algorithms: Euclidean distance, Mahalanobis distance, and symmetric cross-entropy distance. Euclidean distance is the distance metric algorithm used in the unmodified TOPSIS method. Mahalanobis distance considers the correlation between metrics, effectively eliminating interference caused by metric correlation. Symmetric cross-entropy distance is extremely sensitive to differences in data distribution, effectively amplifying small differences in the performance of evaluation objects on metrics, and significantly improving ranking discriminative power.

[0080] Then, the correlation coefficient between the two metrics (Bachton distance and kurtosis) is calculated. ρ The calculation is as follows:

[0081] (9)

[0082] in, and These are the average values ​​of the two metrics in the weighted normalized decision matrix.

[0083] like (Indicating a strong correlation between the metrics), Mahalanobis distance is chosen for calculation.

[0084] like (Indicating strong independence among the metrics), we first use traditional Euclidean distance for pre-calculation to obtain the standard deviation of the relative closeness of each mode. σ .like σ< A value of 0.05 (indicating that the modal proximity is too concentrated and the discrimination is insufficient) will automatically switch to symmetric cross-entropy distance for calculation. If σ If the value is ≥0.05, then Euclidean distance should still be used for calculation.

[0085] Based on the distance metric algorithm selected by the adaptive distance calculation module, the distance between each mode and the positive and negative ideal solutions is calculated. , This yields the final relative fit for each modality; the specific process is as follows:

[0086] First, the scheme consisting of the optimal values ​​of each metric is determined to be the positive ideal solution. This indicates that the ideal solution is... Simultaneously, the scheme consisting of the worst values ​​of each metric is determined to be the negative ideal solution. This indicates that the negative ideal solution is... .

[0087] Then calculate the distance, if you choose Mahalanobis distance:

[0088] (10)

[0089] (11)

[0090] in, S For weighted normalized decision matrix V The covariance matrix, For the first A weighted metric vector for each modality.

[0091] If we choose symmetric interaction entropy distance:

[0092] (12)

[0093] (13)

[0094] in, For the first k The optimal value of each metric, i.e. , For the first k The worst value of each metric, i.e. .

[0095] If Euclidean distance is chosen:

[0096] (14)

[0097] (15)

[0098] Finally, the final relative fit of each mode is calculated. And sort them, including relative proximity. The calculation formula is as follows:

[0099] (16)

[0100] The value range is [0,1]. The closer the value is to 1, the more sensitive the acoustic signal mode is to contact erosion. Based on... Values ​​from largest to smallest for all The modal components are sorted, and a screening threshold is set. The modal components with a relative similarity greater than the screening threshold are selected as the acoustic signal components sensitive to the contact morphology, thus obtaining the screening results of sensitive modal components. The selected sensitive modal components are then reconstructed by linearly superimposing all sensitive modal components to obtain a modal reconstruction acoustic signal containing contact morphology-sensitive components that can effectively characterize the degree of contact erosion. .

[0101] Step 4: Calculate the contact erosion residual value. A Hilbert envelope spectrum is constructed from the modal reconstruction acoustic signal containing components sensitive to contact morphology. The skewness value of the Hilbert envelope spectrum is calculated. This skewness value is used as a characteristic parameter of the acoustic signal representing the contact morphology state, quantifying the contact erosion residual value and accurately assessing the contact's health status. The contact erosion residual value is obtained by homogenizing the skewness value of the Hilbert envelope spectrum.

[0102] First, the modal reconstruction acoustic signal obtained in the third step, which includes components sensitive to the contact morphology, is processed. Perform the Hilbert transform, i.e.

[0103] (17)

[0104] In the formula: for Hilbert transform, For time, For a certain time.

[0105] Constructing envelope signals ,Right now

[0106] (18)

[0107] Performing a Fourier transform on the envelope signal yields the Hilbert envelope spectrum. ,Right now

[0108] (19)

[0109] in, f For frequency, j It is the imaginary unit.

[0110] Then, the skewness of the Hilbert envelope spectrum is calculated to obtain the acoustic signal characteristic parameters that characterize the contact morphology.

[0111] Finally, acoustic signal characteristic parameters representing the contact morphology of brand-new and failed test samples were calculated respectively. The contact erosion residual value was obtained by homogenizing the results. This value is an important parameter for evaluating the degree of contact erosion and is used to quantify the contact erosion residual value and accurately assess the health status of the contact.

[0112] The process of quantifying the residual value of contact erosion is as follows: obtain the skewness values ​​of the Hilbert envelope spectrum of the new sample and the failed sample, and use them as the maximum and minimum values ​​of the skewness values ​​of the Hilbert envelope spectrum, respectively, to obtain the skewness value of the Hilbert envelope spectrum of the sample to be monitored. By uniformizing the calculation results, an intuitive and quantitative residual value of contact erosion is obtained.

[0113] Example 1:

[0114] This embodiment presents a method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals, including the following steps:

[0115] Step 1: Collect acoustic signal sequences of the contact closing stage (closing stage) of the high-voltage DC contactor under different erosion conditions;

[0116] Twenty-four high-voltage DC contactor samples of the same batch were subjected to DC breaking tests under different operating conditions and with varying numbers of cycles to obtain samples with different degrees of contact erosion. These samples were then used as test objects for no-load testing. A synchronous acquisition system for the acoustic signals during operation and the electrical signals in the contact circuit was used to collect the acoustic signals of the contactors during the closing process under different contact erosion states. Under completely identical operating conditions, three samples were selected to perform the same number of operations, and this set of data was used as an analysis unit to eliminate potential interference from randomness between products on the analysis results. The test object information is shown in Table 1. The rated voltage of the contactor coil was set to 12V. The acoustic pressure signal was measured using a MA231 preamplifier and an MP251 1 / 2-inch pressure field microphone, with a frequency range of 3.15 Hz-20 kHz, placed directly above the contactor. A DH5922D dynamic test analyzer was used to acquire the acoustic signal data. Furthermore, a LabVIEW-controlled PLC was used to perform operation tests on the test objects at an operating frequency of 20 times / min, with the sampling frequency set to 256 kHz.

[0117] Table 1. Information on test subjects

[0118]

[0119] Step 2: Based on the short-time energy dual threshold method, accurately identify the boundary of the contact closure acoustic event (contact collision event) and extract the contact collision acoustic signal segment containing morphological information;

[0120] Signals during the contactor closing process, such as Figure 3 As shown. The sampling frequency of the audio signal is 256kHz, the selected window length is 50, and the frame shift is 10. t 1 At the moment when the moving and stationary contacts begin to collide and close, based on multiple measurement results, the energy threshold corresponding to that moment is set to 10 mJ. t 2 When the contacts are stably closed and the iron core begins to collide, its short-time energy value is equal to or slightly less than the maximum value of the total energy peak. In this embodiment, t 2 The threshold value for each moment is set to 0.6 times the maximum value of the total energy peak value. Simultaneously, to avoid peak interference during the contact collision phase, the peak value is required to be... t 1 The minimum distance is 150 frames. t 1 and t 2 The acoustic signal segments between them are extracted to obtain the contact collision acoustic signal segments containing shape information.

[0121] Step 3: Extract acoustic signal components that are sensitive to contact morphology to obtain modal reconstruction acoustic signals containing components sensitive to contact morphology;

[0122] The acoustic signal generated by the contact collision has a frequency distribution of 1200Hz and above. The portion below 1000Hz is mainly caused by baseline shift due to slight sensor jitter or environmental interference. A wavelet soft thresholding method is used to denoise the acoustic signal. The signal is decomposed into eight levels of wavelets, and the approximation coefficients of the eighth level (0-1000Hz) are set to zero to solve the baseline drift problem. The effects before and after denoising are shown below. Figure 4 As shown.

[0123] Symplectic geometric mode decomposition (SGMD) is used to perform mode decomposition on the contact collision sound signal segment containing morphological information after wavelet soft thresholding, and the embedding dimension is set. d=n / 3( n (Number of sampling points for the contact collision sound signal segment), delay time λ =1, resulting in 10 modal components ( d =10), the result is as follows Figure 5 As shown.

[0124] To further extract sensitive components containing contact morphology, the EW-TOPSIS evaluation method based on the adaptive distance calculation module was used to screen the modal components obtained from the symplectic geometric mode decomposition (SGMD). The specific process is as follows:

[0125] First, calculate the Bach distance and kurtosis value between each modal component and the original signal (the signal before decomposition). Construct a decision matrix with the modal components as rows and the metrics as columns. W By analyzing the decision matrix W By performing forward normalization and vector normalization, a standardized matrix is ​​obtained. Z .

[0126] Secondly, the entropy weight vector is calculated using the entropy weight method. w And use the obtained entropy weights to apply to the standardized matrix Z Weighting is performed to form a weighted normalized decision matrix. V .

[0127] Then, an adaptive distance calculation module is used to calculate the correlation coefficient between the metrics. ρ Standard deviation of relative closeness to each mode σ This is used to determine the characteristics of the metric data and automatically trigger the selection of the most suitable distance metric algorithm. Based on the selection of the adaptive distance calculation module, the distance between each mode and the positive and negative ideal solutions is calculated, and finally the final relative proximity of each mode is obtained. .in accordance with Values ​​from largest to smallest for all By sorting the modal components, the filtering results of sensitive modal components can be obtained.

[0128] Figure 6 The metrics for a given measurement and the final relative closeness The calculation results. A comprehensive analysis of multiple measurement results revealed that the modal components... Values ​​between 0.400 and 0.030 and below 0.015 should be considered when... A threshold below 0.020 contains too little valid information; therefore, 0.020 is set as the filtering threshold. Specifically, SGC7 and SGC9... The value is approximately 0, which is much smaller than the threshold of 0.020. Therefore, a set of modal components containing contact morphology-sensitive elements is selected to effectively characterize the degree of contact erosion. The signal is reconstructed to obtain a modal reconstruction acoustic signal containing contact morphology-sensitive components that can effectively characterize the degree of contact erosion.

[0129] Step 4: Calculate the residual value of contact erosion.

[0130] First, the Hilbert envelope spectrum is constructed from the modal reconstruction acoustic signal containing components sensitive to contact morphology. Figure 7 The Hilbert envelope spectra of the contact morphology under different degrees of erosion for impact sound generation are shown in the figure. As can be seen from the figure, the peak material content increases with... V mp As the frequency increases, the high-frequency proportion of the Hilbert envelope spectrum increases, and the center of gravity of the waveform shifts to the right. The acoustic signal characteristic parameters characterizing the contact morphology are derived by calculating the skewness of the Hilbert envelope spectrum in the frequency domain. Simultaneously, acoustic signal characteristic parameters characterizing the contact morphology are calculated for both brand-new and failed samples, serving as upper and lower limits. After homogenizing the sample parameters, the quantified contact erosion residual value is obtained.

[0131] To compare the consistency between changes in acoustic signal features and morphological features under different degradation states, the method proposed in this invention was used to extract features from acoustic signals at 10, 100, and 1000 cycles at 450V / 300A. Morphological scanning was performed using a DSX1000 digital microscope equipped with two high-performance objectives: the MPLFLN5XBDP (5x magnification) and the MPLFLN10XBDP (10x magnification). The peak material content of the contact was determined. V mp It refers to the material volume of the peak region in the surface morphology, which is the main contact area of ​​the contact during the collision process. During arc erosion, the contact surface undergoes melting, splashing, and resolidification, causing material to migrate from low-lying areas to higher areas, forming resolidified protrusions; peak material volume. V mpThese "malicious peaks" increase in volume and directly reflect the concentration of arc energy and the severity of surface reconstruction. V mp The larger the contact, the more severe the contact erosion.

[0132] Table 2 shows the calculation results of contact erosion residual value and peak material content. As can be seen from the table, with the increase of the number of operations, the peak material content of the contact increases. V mp As the contact erosion rate increases, the degree of erosion of the contact also gradually increases, while the residual value of the eroded contact gradually decreases. This indicates that the proposed method can effectively reflect the degree of contact erosion.

[0133] Table 2 Calculation results of contact erosion residual value and peak material quantity

[0134]

[0135] To verify the advantages of this invention and the contact morphology-sensitive acoustic signal component extraction method, it was compared with the unprocessed original signal and the traditional EW-TOPSIS modal reconstruction method. The results are as follows: Figure 8 As shown in the figure, the trend of the contact erosion residual value of the unprocessed original signal fluctuated multiple times, and was significantly different from that of the original signal. V mp The inconsistent trends fail to reflect the degree of contact erosion. While the traditional EW-TOPSIS modal reconstruction method... V mp While the trend remains consistent, its evaluation accuracy is significantly lower than the EW-TOPSIS evaluation method based on an adaptive distance calculation module proposed in this invention. In summary, this invention can effectively extract contact morphology-sensitive components, further improving the accuracy of contact erosion residual values.

[0136] To verify the advantage of Hilbert envelope spectrum in characterizing the degree of contact erosion, it was compared with the commonly used frequency domain waveform characteristic parameter, centroid frequency. FC and mean square frequency MSF For comparison, the calculation formula is as follows:

[0137] (20)

[0138] (twenty one)

[0139] in, For frequency, In frequency The power spectral density at that location.

[0140] To quantify contact morphology parameters ( V mpTo determine the correlation between the acoustic signal characteristic parameters and the extracted acoustic signal feature parameters, Spearman correlation coefficient and Kendall correlation coefficient were introduced, both of which are suitable for correlation analysis of ordered categorical variables. The correlation analysis results between the acoustic signal feature parameters and the contact morphology parameters are shown in Table 5. The centroid frequency and mean square frequency showed a negative correlation with the peak material content, while the skewness of the Hilbert envelope spectrum not only showed a positive correlation with the peak material content, but also a very strong correlation. Therefore, the skewness of the Hilbert envelope spectrum can effectively characterize the degree of contact erosion, and its normalized value can be used as the contact erosion residual value for evaluation of high-voltage DC contactor contacts.

[0141] Table 5. Correlation analysis between acoustic signal characteristic parameters and contact morphology parameters

[0142]

[0143] Any aspects not covered in this invention are applicable to existing technologies.

Claims

1. A method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals, characterized in that, The method includes the following steps: Step 1: Collect acoustic signal sequences of the contact closure phase of the high-voltage DC contactor under different contact erosion levels; Step 2: Accurate identification of contact collision events based on the dual threshold method of short-time energy, and extraction of contact collision sound signal segments containing morphological information; Step 3: Extract acoustic signal components sensitive to contact morphology: Low-frequency denoising is achieved by using wavelet soft thresholding. Then, symplectic geometric mode decomposition (SGMD) is used to perform mode decomposition on the denoised contact collision sound signal segment containing morphological information to obtain a series of symplectic geometric mode components. The EW-TOPSIS evaluation method based on an adaptive distance calculation module selects modes: Bach distance is used as a measure of the similarity between each modal component and the original signal, and kurtosis is used as a measure of sensitivity to contact collision events; the entropy weight method is used to calculate the information entropy and entropy weight of each measure; during the evaluation process, the correlation coefficient between the measures is used... ρ Standard deviation of relative closeness to each mode σ The system determines the characteristics of the measurement index data and automatically triggers the selection of the most suitable distance measurement algorithm. If there is a strong correlation between the measurement indexes, Mahalanobis distance is selected as the distance measurement algorithm. If the measurement indexes are highly independent and the proximity between modes is too concentrated, the Euclidean distance is switched to symmetric cross-entropy distance as the distance measurement algorithm. The distance between each mode and the positive and negative ideal solutions is calculated to obtain the final relative proximity of each mode. A screening threshold is set, and the corresponding modal components with a final relative proximity greater than the screening threshold are selected for reconstruction to obtain a modal reconstruction acoustic signal containing contact morphology sensitive components that can effectively characterize the degree of contact erosion. Among them, the standard deviation of the relative closeness of each mode. σ The determination process is as follows: First, Euclidean distance is used for pre-calculation to obtain the standard deviation of the relative closeness of each pre-calculated mode. σ ; Step 4: Perform Hilbert transform on the modal reconstruction acoustic signal containing the contact morphology-sensitive component to obtain the skewness value of the Hilbert envelope spectrum. Use the skewness value of the Hilbert envelope spectrum as the acoustic signal characteristic parameter characterizing the contact morphology state to quantify the contact erosion residual value.

2. The method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals according to claim 1, characterized in that, The process of quantifying the residual value of contact erosion is as follows: obtain the skewness values ​​of the Hilbert envelope spectrum of the new sample and the failed sample, and use them as the maximum and minimum values ​​of the skewness values ​​of the Hilbert envelope spectrum, respectively, to obtain the skewness value of the Hilbert envelope spectrum of the sample to be monitored. By uniformizing the calculation results, an intuitive and quantitative residual value of contact erosion is obtained.

3. The method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals according to claim 1, characterized in that, The process for determining the skewness value of the Hilbert envelope spectrum is as follows: First, construct the envelope signal using the signal after Hilbert transform. Then, obtain the Hilbert envelope spectrum by performing a Fourier transform on the envelope signal, and finally calculate the skewness value of the Hilbert envelope spectrum.

4. The method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals according to claim 1, characterized in that, The specific process of the second step is as follows: First, the acoustic signal sequence acquired in the first step is processed by framing using a Hanning window of predetermined length. The short-time energy of each frame is calculated, an energy threshold is set, and the first point where the short-time energy exceeds the energy threshold is identified as the attraction moment. t 1 ; Then, from t 1 Begin by finding the energy peaks, setting the minimum distance between peaks, and the distance between peaks and... t 1 The minimum distance is used as a constraint condition; the overtravel time is... t 2 The threshold is set to 0.6 times the peak energy level to accurately pinpoint the overtravel moment. t 2 ; Finally, extract the data from the moment of attraction. t 1 Overrun time t 2 The acoustic signal segment is a contact collision acoustic signal segment containing shape information.

5. The method for analyzing the residual value of high-voltage DC contactor contacts based on acoustic signals according to claim 1, characterized in that, The specific process of the EW-TOPSIS evaluation method based on the adaptive distance calculation module is as follows: Each modal component is considered an evaluation object, and the metric set for each modal component includes Bhattacharyya distance and kurtosis. A decision matrix W is constructed with the modal components as rows and the metric sets as columns. The Bhattacharyya distance is forwarded using a subtraction transformation method, unifying all metrics to a maximum size; then, the forwarded matrix is ​​normalized to obtain a standardized matrix. Z ; The entropy weight vector is obtained by calculating the entropy weight method. w And use the obtained entropy weights to apply to the standardized matrix Z Weighting is performed to form a weighted normalized decision matrix. V ; Based on the selection of the adaptive distance calculation module, the distance between each mode and the positive and negative ideal solutions is calculated, and finally the final relative closeness of each mode is obtained; The adaptive distance calculation module is used to automatically trigger the selection of the most suitable distance metric algorithm based on the characteristics of the metric data. The specific implementation process is as follows: Construct a library of alternative distance functions, including Mahalanobis distance, Euclidean distance, and symmetric interaction entropy distance. Based on the weighted normalized decision matrix, calculate the correlation coefficient between two metrics. ρ ;like Then, Mahalanobis distance is chosen to calculate the distance between each mode and the positive and negative ideal solutions; like First, Euclidean distance is used for pre-calculation to obtain the standard deviation of the relative closeness of each pre-calculated mode. σ, like σ< If the value is 0.05, the system will automatically switch to symmetric cross-entropy distance to calculate the distance between each mode and the positive and negative ideal solutions; if... σ If the value is ≥0.05, then Euclidean distance will still be used to calculate the distance between each mode and the positive and negative ideal solutions.

Images