Method for extracting mechanical defect vibration and sound characteristics of GIS (Gas Insulated Switchgear) equipment

Abstract

The invention discloses a mechanical defect vibration sound feature extraction method for GIS equipment, and belongs to the technical field of equipment detection. Comprising a data acquisition unit, a data transmission unit, a vibration sound processing unit, a data storage unit, a data processing unit and a display reminding unit, and the data acquisition unit is used for acquiring various data required in the use process of the mechanical defect vibration sound feature extraction system of the GIS equipment. The invention provides an optimal correction logarithmic spectrum estimation and Wiener filtering fusion algorithm, noise signals are estimated, the amplitude part in the noisy signals is processed and updated to obtain enhanced sound, the OM-LSA is used for estimating and filtering transient noise, the Wiener filtering is used for filtering steady-state noise, and the noise is filtered. According to the method, original vibration signals and acoustic signals of each operation state of the GIS are fused, a series of characteristic parameters are extracted from a multi-dimensional domain, and effective characterization of the state of the GIS equipment is realized.

Description

Technical Field

[0001] The invention relates to the technical field of equipment detection, and in particular to a method for extracting vibration and sound features of mechanical defects of GIS equipment. Background Art

[0002] Since the sound signal is easily interfered by external signals during the field test, thus affecting the quality of the test data, it is necessary to denoise the sound signal before GIS equipment fault diagnosis and feature extraction, filter the background signal in the collected signal and purify it to obtain pure GIS operation sound.

[0003] Substations are mainly composed of GIS, transformers, busbars, grounding switches, lightning arresters, transformers, capacitors, reactors, and live structures. Generally speaking, the higher the voltage level, the greater the substation noise value. The main noise sources of transformers in substations of different voltage levels are also different.

[0004] The on-site test results show that the main noise-generating equipment in 110kV voltage level substations are transformers and ventilation fans; the main noise-generating equipment in 220kV and 330kV voltage level substations are main transformers and live structures (metal fittings, wires and insulators, etc.); the main noise sources in 500kV and above voltage level substations are transformers, high-voltage shunt reactors and live structures.

[0005] The main transformer is generally located in the center of the substation. Its noise is mainly generated by the cooling system and the vibration of the transformer body (core, winding). The noise value is greatly affected by the transformer cooling device (fan, oil pump). The high-voltage shunt reactor is generally located near the substation wall. Its noise characteristics are similar to those of the transformer body and are mainly related to the line load. The live frame is mainly located on both sides of the transformer group. Its noise value is relatively small. The noise measurement result is greatly affected by the noise of the main transformer and is related to climatic conditions. With the increase of air humidity and surface contamination, the noise of the live frame tends to increase gradually.

[0006] When analyzing the acoustic signals of GIS equipment, the noise signal collected by the microphone is a mixed sound signal including transformer noise, noise from other power equipment in the substation, and background noise of the substation. Therefore, it is of great significance to explore a reliable method to suppress the GIS operation background noise and extract the pure original sound signal of GIS equipment from the mixed background sound signal of the substation. Summary of the invention

[0007] Purpose of the invention: The purpose of the present invention is to provide a method for extracting vibration and acoustic features of mechanical defects of GIS equipment; it can solve the problem that the noise signal collected by the microphone is a mixed sound signal containing transformer noise, noise from other power equipment in the substation and background noise of the substation.

[0008] Technical solution: To solve the above technical problems, according to one aspect of the present invention, more specifically, a method for extracting vibration and sound features of mechanical defects of GIS equipment includes: a data acquisition unit, a data transmission unit, a vibration and sound processing unit, a data storage unit, a data processing unit, and a display and reminder unit; Data acquisition unit: used to collect various data required during the use of the mechanical defect vibration and sound feature extraction system of GIS equipment; Data transmission unit: used to transmit the corresponding data collected by the data collection unit to the vibration and sound processing unit, the data storage module and the data processing module, and transmit the corresponding data stored in the data storage module to the data processing unit, and transmit the report on the mechanical defect vibration and sound feature extraction system of the GIS equipment generated by the data processing unit to the display reminder unit; Vibration and sound processing unit: used to receive and purify the mechanical defect vibration and sound data collected by the data collection unit, and extract and analyze the characteristic parameters of the purified vibration and sound data; Data storage unit: used to store various data on quantity, duration and data volume collected by the data collection unit; Data processing unit: used to receive various data on quantity, duration and data volume collected by the data collection unit and various data stored by the data storage unit, and analyze and process the above received data, and generate a report on the mechanical defect vibration and sound feature extraction system of GIS equipment according to the analysis and processing results; Display reminder unit: used to set the threshold of relevant data of the report of the vibration and sound feature extraction system of mechanical defects of GIS equipment, receive and display the report of the vibration and sound feature extraction system of mechanical defects of GIS equipment, and issue warning reminders according to the set threshold of relevant data and reports.

[0009] Furthermore, the data acquisition unit includes: a vibration and sound acquisition module, a quantity acquisition module, a duration acquisition module and a data volume acquisition module; Vibration and sound acquisition module: used to collect the original vibration and sound data that need to be extracted and analyzed for characteristic parameters; Quantity collection module: used to collect various quantity data required by this system; Duration collection module: used to collect various duration data required by this system; Data volume collection module: used to collect various data on data volume required by this system.

[0010] Furthermore, the vibration-acoustic processing unit includes: a vibration-acoustic receiving module, a vibration-acoustic purification module and a characteristic parameter extraction and analysis module; Vibration and sound receiving module: used to receive the original vibration and sound data collected by the data collection unit; Vibration sound purification module: used to filter and purify the original vibration sound data received by the vibration sound receiving module to obtain pure GIS operation sound; Feature parameter extraction and analysis module: used to extract feature parameters that can characterize the GIS operation status from the filtered and purified GIS operation sound and perform feature sensitive parameter analysis.

[0011] Furthermore, the data processing unit includes: a data receiving module, an analysis and processing module, and a report generating module; Data receiving module: used for receiving various data on quantity, duration and data volume collected by the data collection unit and various data stored by the data storage unit; Analysis and processing module: used to analyze and process the data received by the data receiving module; Report generation module: used to generate a report on the mechanical defect vibration and acoustic feature extraction system of GIS equipment based on the analysis and processing results of the analysis and processing module.

[0012] Furthermore, the display reminder unit includes: a threshold setting module, a report receiving module, a report display module and a warning reminder module; Threshold setting module: used to set the threshold of relevant data reported by the mechanical defect vibration and sound feature extraction system of GIS equipment; Report receiving module: used to receive the report generated by the data processing unit on the mechanical defect vibration and acoustic feature extraction system of the GIS equipment; Report display module: used to display the report on the mechanical defect vibration and acoustic feature extraction system of GIS equipment received by the report receiving module; Warning and reminder module: It is used to warn relevant personnel and discover problems in time when the relevant data of the report on the mechanical defect vibration and acoustic feature extraction system of GIS equipment received by the report receiving module reaches or exceeds the set threshold.

[0013] Furthermore, during the process of analyzing and processing, the analysis and processing module can obtain the number of mechanical defects identified by the mechanical defect vibration and sound feature extraction system of the GIS equipment every day, the number of mechanical defects correctly identified every day, the number of mechanical defects not identified every day, the number of mechanical defects actually existing every day, the last mechanical defect vibration and sound data processing time, the current mechanical defect vibration and sound data processing time, the amount of mechanical defect vibration and sound data processed last time, and the amount of mechanical defect vibration and sound data processed currently, and comprehensively analyze the above data to obtain the performance evaluation index of the mechanical defect vibration and sound feature extraction system of the GIS equipment: ; Therefore, the performance of the system can be obtained according to the performance evaluation index of the mechanical defect vibration and sound feature extraction system of GIS equipment. The larger the performance evaluation index, the better the system performance, and vice versa. Among them, PZ is the performance evaluation index of the mechanical defect vibration and sound feature extraction system of GIS equipment, is the number of mechanical defects identified per day, is the number of mechanical defects correctly identified per day, is the number of unidentified mechanical defects per day, is the actual number of mechanical defects per day, is the duration of the last mechanical defect vibration and sound data processing, is the processing time of the current mechanical defect vibration and sound data, is the amount of mechanical defect vibration and sound data processed last time, The amount of mechanical defect vibration and acoustic data currently being processed.

[0014] According to another aspect of the present invention, a method for extracting vibration and acoustic features of mechanical defects of GIS equipment is provided. The method is implemented based on the above-mentioned system for extracting vibration and acoustic features of mechanical defects of GIS equipment, and specifically comprises the following steps: S1, acquisition of raw vibration and sound data; S2, filtering and purification of raw vibration and sound data; S3. Extraction and analysis of vibration and sound characteristic parameters; S4. Collect and store various data required for analysis and processing; S5. Threshold setting for relevant data reported; S6, data reception, analysis and report generation required for analysis and processing; S7. Report display and warning reminders.

[0015] Furthermore, in step S2, the original data of the vibration and sound are filtered and purified by using an optimal modified log spectral estimation (OM-LSA) algorithm and a Wiener filter denoising method derived from psychoacoustic theory.

[0016] Furthermore, in step S3, the extracted vibro-acoustic characteristic parameters include nine characteristic values: odd-even subharmonic ratio, frequency complexity, root mean square, spectral kurtosis, spectral entropy, spectral centroid, spectral diffusion, energy similarity and weighted harmonic distortion.

[0017] Beneficial effect: The present invention proposes an algorithm for fusion of optimal modified logarithmic spectrum estimation and Wiener filtering, estimates the noise signal, and processes and updates the amplitude part of the noisy signal to obtain an enhanced sound, and uses the optimal modified logarithmic spectrum (OM-LSA) to estimate and filter transient noise, and uses Wiener filtering to filter steady-state noise. The original vibration signal and acoustic signal of each operating state of GIS are integrated, and a series of characteristic parameters are extracted from multi-dimensional domains to achieve effective characterization of the GIS equipment state. The result spectrum under the defects of insufficient conductor insertion depth, loose shielding cover and loose external fasteners is analyzed, and nine characteristic values ​​of odd-even harmonic ratio, frequency complexity, root mean square, spectrum peak, spectrum entropy, spectrum center of gravity, spectrum diffusion, energy similarity and weighted harmonic distortion are selected for correlation analysis under typical conditions, and the characteristic values ​​of different GIS states are displayed in box plots, and the amplitude distribution is visualized. The results show that the selected characteristic parameters can significantly distinguish the normal operating state of GIS from the other three typical mechanical defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the system principle; Figure 2 is a schematic diagram of the method flow chart; Figure 3 It is the flow chart of OM-LSA algorithm; Figure 4 yes Calculation flow chart of Figure 5 It is the sound spectrum of the original signal using the optimal modified logarithmic spectrum estimation method; Figure 6 It is the sound spectrum of GIS operation sound after being purified by the optimal modified logarithmic spectrum estimation method; Figure 7 This is a schematic diagram of the input and output of the Wiener filter; Figure 8 are (a) clean speech spectra (solid line) and the corresponding masking threshold (dashed line) and (b) noisy speech spectra (solid line) and the corresponding masking threshold; Fig. 9 is (a) the difference spectrum and the corresponding masking threshold and (b) audible noise spectrum ; Fig.10It is a schematic diagram of the implementation flow of the audible noise suppression algorithm based on the masking threshold method; Fig.11 It is a schematic diagram of the parameter extraction process of the audible noise suppression algorithm based on the masking threshold method; Fig.12 It is the original mixed signal and the denoised pure signal. DETAILED DESCRIPTION

[0019] In order to make the technical solution of the present invention clearer, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] Example 1

[0021] In the first step, the vibration and sound raw data that need to be extracted and analyzed for characteristic parameters are collected through the vibration and sound collection module of the data collection unit.

[0022] In the second step, the original data of vibration and sound are filtered and purified through the vibration and sound purification module of the vibration and sound processing unit using the optimal modified log spectral estimation (OM-LSA) algorithm and the Wiener filter denoising method derived from psychoacoustic theory.

[0023] In order to achieve noise suppression in low signal-to-noise ratio and non-stationary background noise conditions, Cohen et al. proposed an algorithm based on optimal modified logarithmic spectral amplitude estimation (OM-LSA). This algorithm has good robustness in low signal-to-noise ratio and non-stationary noise environments. The idea is to obtain the optimal spectral gain function of the observed signal through the prior signal-to-noise ratio estimation, noise estimation, and equipment operation sound existence probability estimation of the observed signal, and finally multiply the frequency domain observed signal by the gain function to obtain the denoised equipment operation sound signal. The flowchart of the OM-LSA algorithm is shown in the figure. Figure 3 shown.

[0024] Assume that the observed signal is , the pure signal is , the additive noise is . Then we have: ; For the observed signal Perform frame division and windowing operations, and perform short-time Fourier transform. ; in, is the frequency index number, is the index number of the time domain frame, is the window function, and M is the frame shift. Here, LSA estimation is selected to calculate the spectrum gain. The core idea of ​​LSA is to minimize ,in represents the amplitude spectrum of pure speech, and yes Assuming that each spectral component is statistically independent, the estimate of the short-time log spectrum is defined as: ; Here is the following binary hypothesis model: ; in, represents the STFT of the noise, and They represent two states: no sound and sound.

[0025] The conditional probability density functions of the observed signals can be expressed as: ; in, Represent the variance of pure equipment sound and noise respectively. Apply the Bayesian criterion.

[0026] ; in, Respectively represent the prior signal ratio and the posterior signal-to-noise ratio of the observed signal. According to the binary hypothesis model, we have: ; Then the short-time log spectrum amplitude estimation based on OM-LSA can be expressed as: ; Setting the Gain Threshold , is used to characterize the lower limit of the gain of the non-speech segment, and its value depends on the subjective evaluation of the noise. As a state gain function with segments.

[0027] ; The optimal gain function finally obtained by OM-LSA is: ; The prior knowledge for calculating the optimal gain includes the prior signal-to-noise ratio, the posterior signal-to-noise ratio, the prior probability of silence, and the conditional probability of sound.

[0028] From the above part, we can see that we need a priori signal-to-noise ratio estimate and The OM-LSA method adopts a decision-oriented approach to estimate the prior SNR.

[0029] ; in, represents the weight factor, which is used to control the trade-off between noise elimination and transient signal distortion. When it is uncertain whether speech exists, the above formula gives the unconditional priori signal-to-noise ratio, and the final priori signal-to-noise ratio , so and The noise is easy to interact with each other, seriously affecting the result of gain estimation and introducing residual noise. Therefore, the uncertainty in the sound needs to be taken into account when calculating the prior SNR.

[0030]

[0031] When sound exists, becomes , and when there is no sound, Decay to , effectively avoiding the problems in the above formula.

[0032] In the OM-LSA algorithm, a soft decision method is used to calculate the three The parameters related to the time-frequency distribution of First, the smoothing factor , recursively smooth the prior SNR estimate.

[0033]

[0034] Then, the local and global average values ​​of the prior SNR are obtained by locally and globally smoothing the prior SNR using a smoothing window in the frequency domain.

[0035] ; in, You can choose between local and global options. is the selected smoothing window. Definition and Two parameters are used to represent the relationship between the probability of sound existence and the smoothed signal-to-noise ratio.

[0036] ; in, and is an empirical value that can minimize noise while maintaining weaker device sound components. To further eliminate the impact of noise, define parameters based on sound frame energy. First, in the frequency domain The average value of To prevent the sound of the device or weaker components from being cut, When it increases, there are audio frames. The calculation process is as follows Figure 4 shown.

[0037] exist Figure 4 In the , it indicates a soft transition from the device sound frame to the noise frame, and its definition is as follows: ; in, Indicates restricted Finally, the estimated value of the prior probability of silence can be obtained from the above three parameters.

[0038] According to the introduction of the above method, in order to solve the transient noise problem such as voice and footsteps mixed in the GIS operation sound during the field test, the optimal modified logarithmic spectrum estimation (OM-LSA) method is used to filter such noise and extract the pure GIS operation sound. The time-frequency diagram of the original signal and the time-frequency diagram after calculation and purification are shown in Figure 5 , Figure 6 As shown: Figure 5 , Figure 6 In the figure, the horizontal axis represents time, the vertical axis represents frequency, and the color represents the magnitude of the sound. Figure 5 It can be seen that non-periodic noise appears at different times, and the frequency of the signal is relatively wide, including the range of 400~3500Hz. After using the filtering algorithm, non-equipment operation sounds are identified and eliminated at different times. Figure 6 It can be seen that the sound spectrum is mainly composed of 100Hz harmonics, so the optimal modified logarithmic spectrum estimation (OM-LSA) algorithm is used to remove intermittent transient noise.

[0039] The Wiener filter is an optimized linear filter under the minimum mean square error criterion. It has strict requirements when processing signals. The speech signal needs to be stable, but in reality, neither speech nor noise meets the stability requirement. Speech needs to be processed in segments to become short-term stable speech. The Wiener filter algorithm first needs to estimate the power spectrum of the noisy speech signal and the noise signal in the frequency domain through Fourier transform, and then construct a gain function through the above two power values. The power of the filtered pure speech signal is calculated from the noisy speech power and the gain function, and the obtained power is restored in the time domain after Fourier inverse transform to enhance the speech signal.

[0040] Introduction to Wiener filtering: Assume a linear system with an impulse response of h(n). Input a random signal with an observation value of x(n). The random signal contains a signal s(n) and a noise signal n(n). The formula is as follows: ; The output signal y(n) through the Wiener filter is: ; In the filtering process, we hope that the output signal y(n) is as close as possible to the expected target signal s(n). y(n) is the estimated value of s(n). To express, that is ; The input and output of the Wiener filter are as follows: Figure 7 As shown; Assuming that the Wiener filter processing system is a causal system, the output It can be regarded as the estimated value of the observed value at the past time and the observed value at the present time x(n), x(n-1), x(n-2), x(n-3)... Figure 7 It can be seen that the estimated value of the signal is different from the expected signal, and the difference between the two is represented by e(n), that is, ; because It is estimated based on the filter processing, so e(n) can be positive or negative, and it is a random variable, so it is more reasonable to use the minimum mean square value to represent e(n): ; The current output value of the Wiener filter is calculated from the past input value and the current input value, so the expected output value is Mainly based on the system's impulse response h(n) analysis, we can get: ; In order to minimize the mean square error, it is necessary to find the partial derivative of the above formula for h(m), m=0,1,2…N-1 to be equal to 0: ; Further calculation of the above formula can be obtained: ; By expressing equation (2-9) with a related function, we get the Wiener-Hoff equation: ; Subdividing the above formula into smaller parts, we can get a linear equation system, which can be expressed in matrix form as follows: ; The simplified form is: ; in: is the cross-correlation sequence; are the filter coefficients; is the autocorrelation matrix.

[0041] From the above, it can be seen that the Wiener filtering process is mainly to find the expression of the unit impulse response h(n) under the minimum mean square error. It can be seen that the accuracy and precision of the Wiener filter are directly related to the order N of the filter. In order to make the error between the output estimated signal and the expected pure signal as small as possible, a feasible way is to increase the order of the Wiener filter. The increase in the order will increase the amount of calculation of the entire filtering process, which will reduce the efficiency of the filtering.

[0042] The above is the principle of Wiener filtering. When Wiener filtering is applied to speech enhancement algorithms, it needs to be transformed into the frequency domain for filtering. The Fourier transform of the above formula can be obtained: ; , and are the amplitude spectrum representations of the kth spectral component of the time domain signal in the jth frame. The frequency response function of the Wiener filter is: ; In the above formula is the pure speech spectrum power, is the power spectrum of the noise frequency domain. H(k,j) is the gain function and must be a non-negative real number, similar to the impulse response h(n).

[0043] By expanding the gain function of the Wiener filter, we can obtain a generalized parameterized Wiener filter gain function: ; When the parameters α and β change, the effect of the filter will be different. Because in real processing, if the estimated noise power is too small, the speech after Wiener enhancement will contain more residual noise, and if the estimated noise power is large, the enhanced speech will be seriously distorted. When the estimated noise power is too small, α >1 is taken. When the estimated noise power is too large, the value range of α is 0<α<1. The β value is equivalent to amplifying the enhanced signal. According to experience, the value range is β ≥ 1. When α = 1 and β = 1, it is the basic Wiener filtering algorithm.

[0044] The spectrum of the enhanced signal can be obtained as: ;

[0045] Where S(k, j) is the amplitude spectrum of the enhanced signal in the corresponding frequency band.

[0046] This project uses a Wiener filter denoising method derived from psychoacoustic theory. This method is based on the size of the masking threshold and is mathematically derived to the audible spectrum components. Generally speaking, the audible spectrum components are those spectrum components that are greater than the masking threshold, so they can be obtained by taking the maximum value between each frequency component of the signal power spectrum and its corresponding masking threshold. Its definition is as follows: ; in represents the masking threshold of the kth frequency component at the jth time. The spectral components perceived as noise are the audible spectrum of additive noise, which can be expressed as the difference between the audible spectrum of the noise-bearing noise and the audible spectrum of the pure sound. That is, the audible spectrum of the noise can be expressed as: ; like Figure 3-5 The figure shows the power spectra of the clean signal and the noisy signal, as well as the masking threshold obtained by the clean signal. Figure 3-6 The figure shows the difference spectrum between the noisy power spectrum and the pure power spectrum, that is, , and the masking threshold. In theory, the difference spectrum corresponds to the noise spectrum (if the cross power spectrum term). The audible noise spectrum calculated according to formula (2-20) will also be Figure 3-6 , most of its spectral components fall above the masking threshold. Since these noise spectral components are larger than the masking threshold, they can be perceived by hearing and therefore need to be cancelled or suppressed.

[0047] It can be deduced A more specific expression: ;

[0048] In the above formula, for scenarios (III) and (IV), the audible spectrum can only be negative or 0. There will be no audible noise in these two cases (due to ), so there is no need to adjust the power spectrum of the noisy sound However, for scenario (II), the audible spectrum is always positive or 0 (if ), it is obvious that the audible noise needs to be eliminated or suppressed. For scenario (I), according to different and , the audible spectrum may be positive, zero or negative. Therefore, the main goal of the audible noise suppression algorithm is to adjust the noisy audio spectrum in some way , so that the audible noise spectrum is a negative value or 0. In this way, the residual noise (after spectral adjustment) becomes inaudible.

[0049] Now suppose that the noisy power spectrum has been adjusted , and the enhanced speech spectrum is obtained , denoted as. Then the adjusted audible noise spectrum , denoted as, the following conditions must be met: ; Where N is the FFT length. Since scenarios (III) and (IV) do not cause audible noise, they will not be affected by spectrum adjustment. For the spectrum adjustment method of noisy sound spectrum, a parameterized Wiener-type function is selected, which is as follows: ; in and are time-frequency variable parameters, which are assumed to be positive. The above Wiener-type function is chosen because it can be obtained by appropriately choosing the parameters and The main difference between the above suppression algorithm and the Wiener method is that it can provide relatively constant suppression in the low SNR area. In contrast, both the spectral subtraction and Wiener filtering algorithms provide more suppression in the lower SNR position.

[0050] By adjusting the parameters and , can provide the optimal spectral adjustment for the noisy speech spectrum to satisfy constraints (I) and (II) (constraints (III) and (IV) are not affected by the spectral adjustment). The parameters and of constraints (I) and (II) are obtained as follows: ; In order to simplify the problem, we choose to use a fixed while changing The optimal value is derived The solution has the disadvantage that it relies on a good estimate of the entire spectrum of the clean speech. Therefore, we can choose to estimate the spectrum of the key frequency band of the clean speech signal. This reduces the scope of the problem. Previously, N frequency components needed to be estimated, but now only B key frequency band components need to be estimated (for 16kHz sampling frequency, B=22; for telephone speech, B=18).

[0051] There are two ways to estimate the parameters There are two types of algorithms, one relies on the estimation of the spectral minimum in each critical frequency band and the other is based on the estimation of the auditory masking threshold.

[0052] The scheme for spectral minima has the following form: ; The masked thresholding scheme has the following form: ; Where B is the number of critical bands, is the noise spectrum estimate for the ith critical band, is the minimum value of the pure speech signal spectrum of the ith key band, is the estimated masking threshold for the ith frequency band. It should be noted that now the parameter and is a function of the key frequency band, not a function of discrete spectral points, and is assumed to be constant within each frequency band. Since a pure speech signal spectrum is generally not available, if the minimum spectral component is assumed to follow the Rayleigh distribution, we can derive A statistical estimator of .

[0053] A good estimate of the auditory masking threshold can be obtained through an iterative process. Specifically, the process requires passing the noisy signal through the suppression function multiple times, where Each iteration can get a better estimate of the pure speech signal, and thus a more accurate estimate of the auditory masking threshold. The pth iteration process is expressed as follows: ; in , expressed by the following formula: ; in represents the noise power estimate of the i-th frequency band at the p-th iteration. The above iterative algorithm uses the power spectrum of noisy speech, that is, To initialize. The initialization approximate value of the masking threshold in the formula is , comes from the estimated value of the pure spectrum obtained after subtracting the power spectrum of the noisy speech signal.

[0054] The audible noise suppression algorithm based on the masking threshold method is summarized as follows. For each speech frame: Step 1: Calculate the power spectrum of the noisy speech signal through FFT and save the phase of the noisy speech spectrum.

[0055] Step 2: Get an estimate of the pure signal power spectrum by power spectrum subtraction .

[0056] Step 3: Use the spectrum obtained in step 2 to estimate the masking threshold , and use the obtained masking threshold as the initial value. Use the iterative algorithm to obtain the masking threshold A better estimate of .

[0057] Step 4: Use the masking threshold obtained in step 3 to adjust the noisy power spectrum .make .

[0058] Step 5: Using the noisy speech phase spectrum, calculate the inverse FFT of the amplitude spectrum estimated in step 4.

[0059] The implementation process of the algorithm is as follows Fig.10 The parameter extraction process is shown in Fig.11 shown.

[0060] In order to solve the problem that the fan sound of peripheral equipment is mixed into the GIS operation sound during the on-site test, the Wiener filter method is used to filter the fan sound and extract the pure GIS operation sound. Fig.12 It can be seen that the fan operation noise is a type of broadband noise, covering a wide frequency range from several hundred to several thousand hertz. After filtering using the Wiener filtering method, the harmonic components in the sound are more obvious than the original signal, and the broadband noise between 300Hz-400Hz and 1200Hz-1400Hz is eliminated more significantly. Therefore, the Wiener filtering method has a very obvious filtering effect on the broadband noise in the station.

[0061] The third step is to extract and analyze the characteristic parameters of the filtered and purified vibration-acoustic data through the characteristic parameter extraction and analysis module of the vibration-acoustic processing unit. The extracted vibration-acoustic characteristic parameters include nine characteristic values: odd-even harmonic ratio, frequency complexity, root mean square, spectral peaks, spectral entropy, spectral center of gravity, spectral diffusion, energy similarity and weighted harmonic distortion.

[0062] The fourth step is to collect various data required for analysis and processing through the quantity collection module, duration collection module and data volume collection module of the data collection unit, and transmit them to the data storage unit for storage.

[0063] The fifth step is to set the threshold of relevant data reported by the mechanical defect vibration and sound feature extraction system of the GIS equipment through the threshold setting module of the display reminder unit.

[0064] The sixth step is to receive and analyze the data to be analyzed and processed through the data receiving module, analysis and processing module and report generating module of the data processing unit, and generate a report on the mechanical defect vibration and acoustic feature extraction system of the GIS equipment according to the analysis and processing results.

[0065] During the analysis and processing, the analysis and processing module can obtain the number of mechanical defects identified by the mechanical defect vibration and sound feature extraction system of the GIS equipment every day, the number of mechanical defects correctly identified every day, the number of mechanical defects not identified every day, the number of mechanical defects actually existing every day, the last mechanical defect vibration and sound data processing time, the current mechanical defect vibration and sound data processing time, the amount of mechanical defect vibration and sound data processed last time, and the amount of mechanical defect vibration and sound data processed currently, and comprehensively analyze the above data to obtain the performance evaluation index of the mechanical defect vibration and sound feature extraction system of the GIS equipment: ; Therefore, the performance of the system can be obtained according to the performance evaluation index of the mechanical defect vibration and sound feature extraction system of GIS equipment. The larger the performance evaluation index, the better the system performance, and vice versa. Among them, PZ is the performance evaluation index of the mechanical defect vibration and sound feature extraction system of GIS equipment, is the number of mechanical defects identified per day, is the number of mechanical defects correctly identified per day, is the number of unidentified mechanical defects per day, is the actual number of mechanical defects per day, is the duration of the last mechanical defect vibration and sound data processing, is the processing time of the current mechanical defect vibration and sound data, is the amount of mechanical defect vibration and sound data processed last time, The amount of mechanical defect vibration and acoustic data currently being processed.

[0066] The seventh step is to display the report through the report display module of the display reminder unit, and when the relevant data of the report on the mechanical defect vibration and acoustic feature extraction system of GIS equipment received by the report receiving module reaches or exceeds the set threshold, the relevant personnel are warned to discover the problem in time.

[0067] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A mechanical defect vibration and acoustic feature extraction system for GIS equipment, characterized in that: include: Data collection unit, data transmission unit, vibration and sound processing unit, data storage unit, data processing unit and display reminder unit; Data acquisition unit: used to collect various data required during the use of the mechanical defect vibration and sound feature extraction system of GIS equipment; Data transmission unit: used to transmit the corresponding data collected by the data collection unit to the vibration and sound processing unit, the data storage module and the data processing module, and transmit the corresponding data stored in the data storage module to the data processing unit, and transmit the report on the mechanical defect vibration and sound feature extraction system of the GIS equipment generated by the data processing unit to the display reminder unit; Vibration and sound processing unit: used to receive and purify the mechanical defect vibration and sound data collected by the data collection unit, and extract and analyze the characteristic parameters of the purified vibration and sound data; Data storage unit: used to store various data on quantity, duration and data volume collected by the data collection unit; Data processing unit: used to receive various data on quantity, duration and data volume collected by the data collection unit and various data stored by the data storage unit, and analyze and process the above received data, and generate a report on the mechanical defect vibration and sound feature extraction system of GIS equipment according to the analysis and processing results; Display reminder unit: used to set the threshold of relevant data of the report of the vibration and sound feature extraction system of mechanical defects of GIS equipment, receive and display the report of the vibration and sound feature extraction system of mechanical defects of GIS equipment, and issue warning reminders according to the set threshold of relevant data and reports.

2. The mechanical defect vibration and acoustic feature extraction system of GIS equipment according to claim 1, characterized in that: The data collection unit includes: a vibration and sound collection module, a quantity collection module, a duration collection module and a data volume collection module; Vibration and sound acquisition module: used to collect the original vibration and sound data that need to be extracted and analyzed for characteristic parameters; Quantity collection module: used to collect various quantity data required by this system; Duration collection module: used to collect various duration data required by this system; Data volume collection module: used to collect various data on data volume required by this system.

3. The mechanical defect vibration and acoustic feature extraction system of GIS equipment according to claim 2 is characterized in that: The vibration-acoustic processing unit includes: a vibration-acoustic receiving module, a vibration-acoustic purification module and a characteristic parameter extraction and analysis module; Vibration and sound receiving module: used to receive the original vibration and sound data collected by the data collection unit; Vibration sound purification module: used to filter and purify the original vibration sound data received by the vibration sound receiving module to obtain pure GIS operation sound; Feature parameter extraction and analysis module: used to extract feature parameters that can characterize the GIS operation status from the filtered and purified GIS operation sound and perform feature sensitive parameter analysis.

4. The mechanical defect vibration and acoustic feature extraction system of GIS equipment according to claim 3 is characterized in that: The data processing unit includes: a data receiving module, an analysis and processing module and a report generating module; Data receiving module: used for receiving various data on quantity, duration and data volume collected by the data collection unit and various data stored by the data storage unit; Analysis and processing module: used to analyze and process the data received by the data receiving module; Report generation module: used to generate a report on the mechanical defect vibration and acoustic feature extraction system of GIS equipment based on the analysis and processing results of the analysis and processing module.

5. The mechanical defect vibration and acoustic feature extraction system of GIS equipment according to claim 4 is characterized in that: The display reminder unit includes: a threshold setting module, a report receiving module, a report display module and a warning reminder module; Threshold setting module: used to set the threshold of relevant data reported by the mechanical defect vibration and sound feature extraction system of GIS equipment; Report receiving module: used to receive the report generated by the data processing unit on the mechanical defect vibration and acoustic feature extraction system of the GIS equipment; Report display module: used to display the report on the mechanical defect vibration and acoustic feature extraction system of GIS equipment received by the report receiving module; Warning and reminder module: It is used to warn relevant personnel and discover problems in time when the relevant data of the report on the mechanical defect vibration and acoustic feature extraction system of GIS equipment received by the report receiving module reaches or exceeds the set threshold.

6. The mechanical defect vibration and acoustic feature extraction system of GIS equipment according to claim 5, characterized in that: In the process of analyzing and processing, the analysis and processing module can obtain the number of mechanical defects identified by the mechanical defect vibration and sound feature extraction system of the GIS equipment every day, the number of mechanical defects correctly identified every day, the number of mechanical defects not identified every day, the number of mechanical defects actually existing every day, the duration of the last mechanical defect vibration and sound data processing, the duration of the current mechanical defect vibration and sound data processing, the amount of mechanical defect vibration and sound data processed last time, and the amount of mechanical defect vibration and sound data processed currently, and comprehensively analyze the above data to obtain the performance evaluation index of the mechanical defect vibration and sound feature extraction system of the GIS equipment: ； Therefore, the performance of the system can be obtained according to the performance evaluation index of the mechanical defect vibration and sound feature extraction system of GIS equipment. The larger the performance evaluation index, the better the system performance, and vice versa. Among them, PZ is the performance evaluation index of the mechanical defect vibration and sound feature extraction system of GIS equipment, is the number of mechanical defects identified per day, is the number of mechanical defects correctly identified per day, is the number of unidentified mechanical defects per day, is the actual number of mechanical defects per day, is the duration of the last mechanical defect vibration and sound data processing, is the processing time of the current mechanical defect vibration and sound data, is the amount of mechanical defect vibration and sound data processed last time, The amount of mechanical defect vibration and acoustic data currently being processed.

7. A method for extracting vibration and acoustic features of mechanical defects of GIS equipment, characterized by: The method is implemented based on a mechanical defect vibration and acoustic feature extraction system of a GIS device according to any one of claims 1 to 6, and specifically comprises the following steps: S1, acquisition of raw vibration and sound data; S2, filtering and purification of raw vibration and sound data; S3. Extraction and analysis of vibration and sound characteristic parameters; S4. Collect and store various data required for analysis and processing; S5. Threshold setting for relevant data reported; S6, data reception, analysis and report generation required for analysis and processing; S7. Report display and warning reminders.

8. The method for extracting vibration and acoustic features of mechanical defects of GIS equipment according to claim 7, characterized in that: In the step S2, the original data of the vibration and sound are filtered and purified by using the optimal modified log spectral estimation (OM-LSA) algorithm and the Wiener filter denoising method derived from psychoacoustic theory.

9. The method for extracting vibration and acoustic features of mechanical defects of GIS equipment according to claim 1, characterized in that: In step S3, the extracted vibro-acoustic characteristic parameters include nine characteristic values: odd-even subharmonic ratio, frequency complexity, root mean square, spectrum kurtosis, spectrum entropy, spectrum center of gravity, spectrum diffusion, energy similarity and weighted harmonic distortion.

Images