Electric vibration sound signal feature extraction method for operating mechanism in railway high-voltage circuit breaker

Abstract

The invention relates to an electric vibration sound signal feature extraction method for an operating mechanism in a railway high-voltage circuit breaker. The method comprises the steps that feature points of coil current signals are three extreme points and zero points of waveforms of the coil current signals; the feature point extraction of the sound signal adopts wavelet packet decomposition and equal-time frequency band division to obtain energy values of different time frequency bands, and a frequency band with a large energy value difference and obvious features is selected as a feature vector of the sound signal; the feature point extraction of the vibration signal adopts EEMD decomposition to obtain an IMF component, and a box dimension is calculated based on the IMF component, so that the limitation that feature extraction and fault identification are carried out by only depending on a single signal is overcome by respectively carrying out feature extraction on the electric vibration sound signal, and the accuracy of fault diagnosis is improved; the defects of component mode aliasing and end effect existing in EMD are effectively overcome through EEMD; complexity and texture features of a vibration signal image are quantitatively described through fractal box dimensions, and classification and quantification of various mechanical states of the circuit breaker are realized.

Description

Technical Field

[0001] The invention relates to the field of intelligent diagnosis of high-voltage circuit breakers, and in particular to a method for extracting electric vibration and sound signal characteristics of an operating mechanism in a railway high-voltage circuit breaker. Background Art

[0002] The complex operating environment and sophisticated internal structure of railway high-voltage circuit breakers make them prone to failure, especially mechanical failure of the operating mechanism, which poses a potential threat to the stability of the railway power supply system. The traditional manual inspection and regular maintenance mode has many limitations. Human factors can easily lead to missed inspections or misjudgments. With the continuous expansion of the railway network and the continuous increase in train speed, traditional methods can no longer meet the needs of efficient and accurate fault diagnosis.

[0003] Existing technologies can collect circuit breaker operating status signals in real time through advanced sensors, such as the current signal of the opening and closing coils, vibration signals, and sound signals, which can reflect the operating status of the circuit breaker operating mechanism and transmission mechanism. Subsequently, these signals are preprocessed and feature extracted using modern signal processing technology to extract key information reflecting the fault characteristics. However, how to extract effective features from complex and changeable signals and integrate multiple signals for comprehensive analysis is a technical problem currently faced in the field of railway high-voltage circuit breaker fault diagnosis. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and to provide a method for extracting electric vibration and acoustic signal features of an operating mechanism in a railway high-voltage circuit breaker. The method comprises the following steps:

[0005] Collecting electric vibration and sound signals of the operating mechanism in the high-voltage circuit breaker, wherein the electric vibration and sound signals include coil current signals, vibration signals and sound signals;

[0006] Extract characteristic points corresponding to different mechanical states of the high-voltage circuit breaker from the coil current signal, vibration signal and sound signal, analyze and select characteristic points representing the coil current signal, vibration signal and sound signal to form a characteristic vector;

[0007] The characteristic points of the coil current signal are the three extreme points and the zero point of its waveform;

[0008] The feature point extraction of the sound signal uses wavelet packet decomposition to obtain the energy value of each frequency band, and further divides the energy value of each frequency band by equal time to obtain the energy value of different time-frequency bands, and selects the frequency band with a large energy value difference and obvious characteristics from the energy values ​​of different time-frequency bands as part of the feature vector of the sound signal;

[0009] The feature point extraction of the vibration signal adopts EEMD decomposition to obtain IMF components, and the box dimension is calculated based on the IMF components, and the box dimension is used as the feature vector of the vibration signal.

[0010] Preferably, the different mechanical states include a normal state, a screw loose state, an electromagnet stuck state, and a spring aging state.

[0011] Preferably, multiple groups of coil current signals are selected from each mechanical state for comparison, and it is verified whether the coil current signals have the same current characteristics in different mechanical states;

[0012] When the coil current signal has the same current characteristics in different mechanical states, characteristic points for distinguishing the coil current signal in different mechanical states are obtained by comparing the coil current signals in different mechanical states.

[0013] Preferably, the characteristic points of the sound signal include the following steps:

[0014] S41, through the conjugate orthogonal filter h k Filter the sound signal and set g k =(-1) k-1 h 1-k ;

[0015] S42, by setting 2 according to the sound signal N sampling points, the sampling frequency is set to f s , the number of decomposition layers is set to M, and M < N; the wavelet packet decomposition technology is used to process the sound signal and the bandwidth of each layer of the decomposed signal is f s / 2 M+1 , the starting frequency of the nth frequency band is f = (n-1)f s / 2 M+1 ;

[0016] S43, the decomposed signals of each layer are obtained by recursively decomposing the high-pass filtering result Wm 2n+1 and the low-pass filtering signal result Wm 2n according to formula (1), and the length of the high-pass filtering result Wm 2n+1 and the low-pass filtering signal result Wm 2n are both N / 2 M ;;

[0017]

[0018] Where: n = 0, 1, 2, ..., 2 m-1 -1; W0 = f(t); i = 0, 1, 2, 3, ..., N / 2m-1; m = 0, 1, 2, 3, ..., M. When the decomposition reaches the mth level, the number of decomposed frequency bands is 2. m-1 ;

[0019] S44, reconstructing the signal of the designated node according to formula (2) to obtain a signal after wavelet packet transformation;

[0020]

[0021] S45, extracting the energy value of each frequency band according to the wavelet packet space energy operator E(j,n);

[0022]

[0023] Among them, E(j,n) represents the energy value of the i-th node at the decomposition level j, p s (n,j,k) are the wavelet packet transform coefficients.

[0024] Preferably, the percentage of the total energy accounted for by the energy value of each frequency band is calculated, and the mechanical state of the operating mechanism in the high-voltage circuit breaker is determined according to the different percentages under different mechanical states.

[0025] Preferably, the energy values ​​of each frequency band divided equally in time are obtained by decomposing the wavelet packet. m This is achieved by dividing the frequency band signal into N equal parts in time and obtaining the time-frequency energy spectrum of the sound signal.

[0026] Preferably, the feature extraction of the vibration signal comprises the following steps:

[0027] S71, setting the number of overall averages M;

[0028] S72, through the white noise n with standard normal distribution i (t) is added to the original signal x(t) to get the new signal x i (t):

[0029] x i (t) = x(t) + n i (t);

[0030] Where: n i (t) represents the i-th white noise sequence added, x i (t) represents the i-th additional noise signal, i = 1, 2, 3, ..., M;

[0031] S73, using EMD to i (t) decomposes to obtain the IMF components:

[0032]

[0033] Where: c i,j (t) is the jth IMF component when the i-th white noise is added, r i,j(t) represents the residual component, indicating the signal movement trend, and J is the number of IMF components;

[0034] S74, repeat steps S62 and S63 M times, and obtain the set of IMFs obtained by adding white noise with different amplitudes to each decomposition:

[0035] c 1,j (t)c 2,j (t)…c M,j (t),j=1,2,…,J;

[0036] S75, performing a set average operation on the IMF set obtained in step S64 by adopting the principle that the statistical average value of unrelated sequences is zero, to obtain the final IMF after EEMD decomposition:

[0037]

[0038] In the formula, c j (t) is the j-th IMF component of EEMD decomposition, i = 1, 2, 3, ..., M, j = 1, 2, 3, ..., J;

[0039] S76. The threshold TH is obtained by the correlation coefficient r(j) between the IMF and the original signal, and the effective IMF is screened out according to the comparison result between the IMF and the threshold TH:

[0040]

[0041] Where: r j is the correlation coefficient between the jth IMF component and the original signal, t is the sampling point of the signal; when r j When it is greater than the threshold TH, the jth IMF component is retained, otherwise it is removed.

[0042] 8. The method for extracting electric vibration and acoustic signal features of an operating mechanism in a railway high-voltage circuit breaker according to claim 2, characterized in that the calculation of the box dimension based on the IMF component comprises the following steps:

[0043] S81, dividing the IMF component equally on the time axis and determining the grayscale value in the box according to the amplitude of the IMF component in the time period;

[0044] S82. Calculate the box dimension of the IMF component according to the maximum grayscale value and the minimum grayscale value in the box.

[0045] The beneficial effects of the present invention are as follows: 1. By extracting the features of the three signals of electric vibration and sound respectively, a joint feature of electric vibration and sound is constructed. This method overcomes the limitation of relying on a single signal for feature extraction and fault identification, and can more comprehensively reflect the working status of the circuit breaker operating mechanism, thereby significantly improving the accuracy of fault diagnosis;

[0046] 2. The wavelet packet decomposition technology is used to decompose the high-frequency signal of the sound signal more carefully, solving the problems of nonlinearity, non-stationary characteristics and a large number of harmonics;

[0047] 3. The defects of component mode aliasing and tail effect existing in traditional empirical mode decomposition (EMD) are effectively overcome through ensemble empirical mode decomposition (EEMD).

[0048] 4. The fractal box dimension is used to quantitatively describe the complexity and texture characteristics of the vibration signal image, and the classification and quantification of various mechanical states of the circuit breaker are realized, which not only improves the accuracy of fault diagnosis, but also provides strong support for the early detection and precise positioning of faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying creative labor, other drawings obtained based on these drawings still belong to the scope of the present invention.

[0050] Figure 1 It is a principle block diagram of the present invention;

[0051] Figure 2 The coil current waveform and characteristic point comparison in four states, where Figure (a) is the current waveform comparison in four states, and Figure (b) is the characteristic point distribution in four states;

[0052] Figure 3 The three-layer wavelet packet decomposition of each component signal diagram of the present invention, wherein Figure (a) is the sound signal component of the F1 band, Figure (b) is the sound signal component of the F2 band, Figure (c) is the sound signal component of the F3 band, Figure (d) is the sound signal component of the F4 band, Figure (e) is the sound signal component of the F5 band, Figure (f) is the sound signal component of the F6 band, Figure (g) is the sound signal component of the F7 band, and Figure (h) is the sound signal component of the F8 band;

[0053] Figure 4 The energy percentages of each frequency band in the four states of the present invention are shown in Figure (a), wherein Figure (b) shows the energy percentages of each frequency band in the normal state, Figure (c) shows the energy percentages of each frequency band in the state of the electromagnet being stuck, and Figure (d) shows the energy percentages of each frequency band in the state of the spring being aged;

[0054] Figure 5The time-frequency characteristic distribution diagrams of the sound signal under four states of the present invention, wherein Figure (a) is the time-frequency energy spectrum of the normal state, Figure (b) is the time-frequency energy spectrum of the screw loose state, Figure (c) is the time-frequency energy spectrum of the electromagnet stuck state, and Figure (d) is the time-frequency energy spectrum of the spring aging state;

[0055] Figure 6 The diagram is an IMF component diagram of a vibration signal in a normal state of the present invention, wherein diagram (a) is an IMF1 component, diagram (b) is an IMF2 component, and diagram (c) is an IMF component diagram of a vibration signal in a normal state. DETAILED DESCRIPTION

[0056] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings.

[0057] It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are for distinguishing two non-identical entities with the same name or non-identical parameters. It can be seen that "first" and "second" are only for the convenience of expression and should not be understood as limitations on the embodiments of the present invention. The subsequent embodiments will not explain this one by one.

[0058] The directions and positions mentioned in the present invention, such as "upper", "lower", "front", "back", "left", "right", "inside", "outside", "top", "bottom", "side", etc., are only for reference to the directions or positions of the drawings. Therefore, the directions and positions used are for explaining and understanding the present invention, but not for limiting the scope of protection of the present invention.

[0059] like Figures 1 to 6 As shown in the embodiment of the present invention, the present invention includes the following steps:

[0060] Collecting electric vibration and sound signals of the operating mechanism in the high-voltage circuit breaker, wherein the electric vibration and sound signals include coil current signals, vibration signals and sound signals;

[0061] Extract characteristic points corresponding to different mechanical states of the high-voltage circuit breaker from the coil current signal, vibration signal and sound signal, analyze and select characteristic points representing the coil current signal, vibration signal and sound signal to form a characteristic vector;

[0062] The characteristic points of the coil current signal are the three extreme points and the zero point of its waveform;

[0063] The feature point extraction of the sound signal uses wavelet packet decomposition to obtain the energy value of each frequency band, and further divides the energy value of each frequency band by equal time to obtain the energy value of different time-frequency bands, and selects the frequency band with a large energy value difference and obvious characteristics from the energy values ​​of different time-frequency bands as part of the feature vector of the sound signal;

[0064] The feature point extraction of the vibration signal adopts EEMD decomposition to obtain IMF components, and the box dimension is calculated based on the IMF components, and the box dimension is used as the feature vector of the vibration signal.

[0065] It is understandable that the present invention constructs a joint feature of electric, vibration and sound by extracting features from the three signals of electric, vibration and sound respectively, and comparing the features under normal and fault conditions. This method overcomes the limitation of relying on a single signal for feature extraction and fault identification, and can more comprehensively reflect the working state of the circuit breaker operating mechanism, thereby significantly improving the accuracy of fault diagnosis.

[0066] The different mechanical states include normal state, screw loose state, electromagnet stuck state, and spring aging state. It is understandable that by specifying the fault problem, when the collected electro-vibration and acoustic signals are consistent with the signals in these four states, it not only improves the accuracy of fault diagnosis, but also provides strong support for early detection and accurate positioning of faults.

[0067] By selecting multiple groups of coil current signals from each mechanical state for comparison, and verifying whether the coil current signals have the same current characteristics under different mechanical states;

[0068] When the coil current signal has the same current characteristics in different mechanical states, characteristic points for distinguishing the coil current signal in different mechanical states are obtained by comparing the coil current signals in different mechanical states.

[0069] It is understandable that when the coil current signal does not have the same current characteristics under different mechanical states, it means that there are other factors affecting the current characteristics of the coil current signal. Therefore, when it is verified that the current characteristics of the coil current signal are consistent, other factors are excluded, and the mechanical state of the coil current signal can be judged based on the characteristic points of the coil current signal.

[0070] The characteristic points of the sound signal include the following steps:

[0071] S41, through the conjugate orthogonal filter h k Filter the sound signal and set g k =(-1) k-1 h 1-k ;

[0072] S42, by setting 2 according to the sound signal N sampling points, the sampling frequency is set to f s , the number of decomposition layers is set to M, and M < N; the wavelet packet decomposition technology is used to process the sound signal and the bandwidth of each layer of the decomposed signal is f s / 2 M+1, the starting frequency of the nth frequency band is f = (n-1)f s / 2 M+1 ;

[0073] S43, the decomposed signals of each layer are obtained by recursively decomposing the high-pass filtering result Wm 2n+1 and the low-pass filtering signal result Wm 2n according to formula (1), and the length of the high-pass filtering result Wm 2n+1 and the low-pass filtering signal result Wm 2n are both N / 2 M ;;

[0074]

[0075] Where: n = 0, 1, 2, ..., 2 m-1 -1; W0 = f(t); i = 0, 1, 2, 3, ..., N / 2m-1; m = 0, 1, 2, 3, ..., M. When the decomposition reaches the mth level, the number of decomposed frequency bands is 2. m-1 ;

[0076] S44, reconstructing the signal of the designated node according to formula (2) to obtain a signal after wavelet packet transformation;

[0077]

[0078] S45, extracting the energy value of each frequency band according to the wavelet packet space energy operator E(j,n);

[0079]

[0080] Among them, E(j,n) represents the energy value of the i-th node at the decomposition level j, p s (n,j,k) are the wavelet packet transform coefficients.

[0081] It can be understood that in the present invention, after repeated tests and comparisons, the wavelet basis function with the most accurate decomposition result is selected to perform three-layer wavelet packet decomposition on the sound signal. The sampling frequency of the sound signal is 50kHz. According to the Nyquist theorem, the bandwidth of the sound signal is 25kHZ. Taking the sound signal in a normal state as an example, it is equally divided into 8 frequency bands by three layers of wavelet packets. Adaptively selecting the best wavelet basis function avoids the process of multiple tests to select the wavelet basis function, and improves the efficiency and accuracy of signal processing;

[0082] The percentage of the total energy accounted for by the energy value of each frequency band is calculated, and the mechanical state of the operating mechanism in the high-voltage circuit breaker is determined according to the different percentages under different mechanical states.

[0083] The energy value of each frequency band divided equally in time is obtained by wavelet packet decomposition. mThis is achieved by dividing the frequency band signal into N equal parts in time and obtaining the time-frequency energy spectrum of the sound signal.

[0084] It is understandable that since the frequency domain energy value cannot distinguish the four states well, an equal time division is performed on this basis, and the energy values ​​of the three time-frequency domains are extracted as the characteristic vectors of the sound signal. The signal of each frequency band is divided into N parts in time, and the energy of each part is calculated. The audio signal in the normal state is decomposed into 8 sub-frequency bands by three layers of wavelet packets, and then divided into 10 time-frequency bands according to time to obtain the time-frequency energy spectrum of the sound signal. Compare the energy value differences of different states in different frequency bands, and select the frequency band with a larger energy value difference, that is, the frequency band with obvious energy value characteristics, as the characteristic vector of the sound signal.

[0085] The feature extraction of vibration signals includes the following steps:

[0086] S71, setting the number of overall averages M;

[0087] S72, through the white noise n with standard normal distribution i (t) is added to the original signal x(t) to get the new signal x i (t):

[0088] x i (t) = x(t) + n i (t);

[0089] Where: n i (t) represents the i-th white noise sequence added, x i (t) represents the i-th additional noise signal, i = 1, 2, 3, ..., M;

[0090] S73, using EMD to i (t) decomposes to obtain the IMF components:

[0091]

[0092] Where: c i,j (t) is the jth IMF component when the i-th white noise is added, r i,j (t) represents the residual component, indicating the signal movement trend, and J is the number of IMF components;

[0093] S74, repeat step S62 and step S63 M times, and obtain the set of IMFs obtained by adding white noise with different amplitudes to each decomposition:

[0094] c 1,j (t)c 2,j (t)…c M,j (t),j=1,2,…,J;

[0095] S75, performing a set average operation on the IMF set obtained in step S64 by adopting the principle that the statistical average value of unrelated sequences is zero, to obtain the final IMF after EEMD decomposition:

[0096]

[0097] In the formula, c j (t) is the j-th IMF component of EEMD decomposition, i = 1, 2, 3, ..., M, j = 1, 2, 3, ..., J;

[0098] S76. The threshold TH is obtained by the correlation coefficient r(j) between the IMF and the original signal, and the effective IMF is screened out according to the comparison result between the IMF and the threshold TH:

[0099]

[0100] Where: r j is the correlation coefficient between the jth IMF component and the original signal, t is the sampling point of the signal; when r j When it is greater than the threshold TH, the jth IMF component is retained, otherwise it is removed.

[0101] It can be understood that in the processing of vibration signals, the present invention introduces the ensemble empirical mode decomposition (EEMD) method, which effectively overcomes the defects of component mode aliasing and tail effect existing in the traditional empirical mode decomposition (EMD). By superimposing Gaussian white noise and performing multiple empirical mode decompositions, and then averaging the corresponding IMFs, the EEMD method can significantly suppress the generation of modal aliasing and improve the accuracy and stability of signal decomposition.

[0102] The calculation of the box dimension based on the IMF component comprises the following steps:

[0103] S81, dividing the IMF component equally on the time axis and determining the grayscale value in the box according to the amplitude of the IMF component in the time period;

[0104] S82, calculate the box dimension of the IMF component according to the maximum grayscale value and the minimum grayscale value in the box. The principle of the classification box dimension is analyzed as follows:

[0105] Divide the image with M×M pixels into L×L parts (1<L≤M / 2), and let r=L / M, so that each part is a set of L×L×L1 boxes. The expression of L1 is shown in formula (3), where G is the grayscale.

[0106] L1=L×r=L×G / M (3);

[0107] Assuming that the qth box in the (a, b)th grid contains the maximum grayscale value of the image, and the pth box contains the minimum grayscale value of the image, then the box dimension n of the image in the grid is δ (a,b) is;

[0108] n δ (a,b)=p-q+1

[0109] Finally, N is needed δ Number of boxes to cover the entire image:

[0110] N δ =∑n δ (a,b)

[0111] The final formula for obtaining the box dimension is as follows (4):

[0112]

[0113] It can be understood that the vibration signals under four conditions, namely normal state, loose screws, electromagnet jamming, and spring aging, are decomposed by EEMD and the box dimension is obtained. By comparing the box dimension horizontally and vertically, it can be found that as the IMF component increases, the box dimension will decrease successively, and the box dimension of the same IMF component under different conditions is different. It can be seen that by comparing the box dimension horizontally and vertically, the mechanical characteristics and operating status of the circuit breaker can be reflected.

[0114] Here are some experimental data:

[0115] like Figure 2 The closing current comparison under the four states is described; a group of closing current signal data groups in each state is selected and compared together, such as Figure 2 As shown in (a), it can be clearly seen that the current waveforms in different states are also significantly different, especially at the selected characteristic points; Figure 2 (b) It can be seen that the distribution of feature quantities in different states is also different.

[0116] The four feature points (t1, i1), (t2, i2), (t3, i3), and (t4, i4) are distributed differently with the different states of the circuit breaker. In the two-dimensional coordinate system, they can also be regarded as different vectors with the origin as the starting point and the feature point as the focus. At the same time, at a certain feature position, the position of the feature point of each state is also different. Therefore, the feature vector W = [t1, t2, t3, t4, i1, i2, i3] is selected as the feature vector for study. The distance between its various feature quantities can be calculated to classify the state of the circuit breaker.

[0117] Taking the sound signal in normal state as an example, it is equally divided into 8 frequency bands by three layers of wavelet packets. The frequency band distribution is shown in Table 1.

[0118] Table 1 Frequency band distribution of three-layer wavelet packet decomposition of sound signal

[0119] Frequency band number Frequency distribution / kHz Frequency band number Frequency distribution / kHz <![CDATA[F1]]> 0-3.125 <![CDATA[F5]]> 12.500-15.625 <![CDATA[F2]]> 3.125-6.250 <![CDATA[F6]]> 15.625-18.750 <![CDATA[F3]]> 6.250-9.375 <![CDATA[F7]]> 18.750-21.875 <![CDATA[F4]]> 9.375-12.500 <![CDATA[F8]]> 21.875-25.000

[0120] The frequency band components obtained after decomposition are as follows Figure 3 (a) to 3(h).

[0121] The percentage of each frequency band energy in the total energy is calculated as follows: Figure 4 shown.

[0122] Depend on Figure 4 It can be seen that the energy percentage of each frequency band in each state is different. The most obvious feature is the F1 frequency band, which has a large difference in the four states; the remaining energy percentages always have similar situations in two states. For example, the energy percentages of the F2 frequency band in the loose screw state and the electromagnet stuck state are not much different, and the energy percentages of the F3 frequency band in the electromagnet stuck state and the spring aging state are almost the same, so the characteristics are not obvious. If this state is used for fault classification, the accuracy will be greatly reduced, affecting the correctness of the final fault diagnosis result. Therefore, in order to better extract the feature vector of the sound signal and improve the accuracy of fault diagnosis, the sound signal decomposed by the three-layer wavelet packet is further divided in the time domain.

[0123] The normal audio signal is decomposed into 8 sub-bands by three-layer wavelet packet decomposition, which are then divided into 10 time-frequency bands according to time, and the time-frequency energy spectrum of the sound signal is obtained, as shown in Table 2.

[0124] Table 2 Time-frequency energy characteristic spectrum of normal state sound signal

[0125] Serial number <![CDATA[F1]]> … <![CDATA[F7]]> <![CDATA[F8]]> <![CDATA[T1]]> 30.57 … 8.69 15.40 <![CDATA[T2]]> 8.73 … 2.48 3.83 … … … … … <![CDATA[T 10 ]]> 10.37 … 5.34 8.21

[0126] The time-frequency characteristics of the four different states are distributed as follows: Figure 5 As shown by Figure 5 It can be seen that the energy values ​​of the four states in the F1, F7, and F8 frequency bands, i.e., 0-3.125Hz, 18.750-21.875Hz, and 21.875-25.000Hz, are quite different. 11 、E 71 、E 83 The energy value characteristics are obvious.

[0127] From the above analysis, it can be seen that the mechanical state of the high-voltage circuit breaker operating mechanism can be expressed by the time-frequency energy spectrum of the sound signal. Therefore, by analyzing the time-frequency characteristic energy, the mechanical state of the circuit breaker can be classified. In this paper, S=[E 11 , E 71 ,E 83 ] is studied as the characteristic vector of the sound signal.

[0128] After EEMD decomposes the vibration signal, the correlation coefficients of each IMF component are obtained. There are three IMF components that are greater than the threshold. The component graph is as follows: Figure 6 shown.

[0129] The vibration signals under four states, namely, normal state, screw loosening, electromagnet jamming, and spring aging, are decomposed by EEMD and the box dimension is obtained. As shown in Table 3, it can be seen that the obtained box dimensions are all between 1 and 2. By comparing the box dimensions horizontally and vertically, it can be found that as the IMF components increase, the box dimensions will decrease successively. The box dimensions of the same IMF components in different states are different. It can be seen that by comparing the box dimensions horizontally and vertically, the mechanical characteristics and operating state of the circuit breaker can be reflected. Therefore, the vector Z = [D1, D2, D3] is defined as the characteristic vector of the vibration signal state.

[0130] Table 3 Dimensions of vibration signal box in four states

[0131] State / Box Dimension <![CDATA[D1]]> <![CDATA[D2]]> <![CDATA[D3]]> Normal state 1.5236 1.5122 1.5248 Loose screws 1.8643 1.8545 1.8610 Electromagnet stuck 1.1355 1.1764 1.1467 Spring aging 1.3677 1.3425 1.2774

[0132] The present invention proposes an innovative method for extracting the characteristics of the electric vibration and acoustic signals of the operating mechanism of a railway high-voltage circuit breaker by comprehensively using advanced technologies such as wavelet packet decomposition, EEMD, and fractal box dimension. This method not only improves the accuracy and efficiency of fault diagnosis, but also has strong practicality and applicability. It can be widely used in the maintenance and fault diagnosis of railway high-voltage circuit breakers, and has important technical value and practical significance.

[0133] A person skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as ROM / RAM, disk, CD-ROM, etc.

[0134] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

[0135] It should be noted that the embodiments of the present invention can be implemented by hardware, software or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those skilled in the art will appreciate that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in processor control codes, such as such codes provided on a programmable memory or a data carrier such as an optical or electronic signal carrier.

[0136] In addition, although the operation of the method of the present invention is described in a particular order in the accompanying drawings, this does not require or imply that these operations must be performed in this particular order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps described in the flow chart can change the order of execution. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step for execution, and / or one step can be decomposed into multiple steps for execution. It should also be noted that the features and functions of two or more devices according to the present invention can be embodied in one device. On the contrary, the features and functions of a device described above can be further divided into being embodied by multiple devices.

[0137] Although the invention has been described with reference to several specific embodiments, it should be understood that the invention is not limited to the specific embodiments disclosed. The invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims

1. A method for extracting electric vibration and acoustic signal features of an operating mechanism in a railway high-voltage circuit breaker, characterized in that: The method comprises the following steps: Collecting electric vibration and sound signals of the operating mechanism in the high-voltage circuit breaker, wherein the electric vibration and sound signals include coil current signals, vibration signals and sound signals; Extract characteristic points corresponding to different mechanical states of the high-voltage circuit breaker from the coil current signal, vibration signal and sound signal, analyze and select characteristic points representing the coil current signal, vibration signal and sound signal to form a characteristic vector; The characteristic points of the coil current signal are the three extreme points and the zero point of its waveform; The feature point extraction of the sound signal uses wavelet packet decomposition to obtain the energy value of each frequency band, and further divides the energy value of each frequency band by equal time to obtain the energy value of different time-frequency bands, and selects the frequency band with a large energy value difference and obvious characteristics from the energy values ​​of different time-frequency bands as part of the feature vector of the sound signal; The feature point extraction of the vibration signal adopts EEMD decomposition to obtain IMF components, and the box dimension is calculated based on the IMF components, and the box dimension is used as the feature vector of the vibration signal.

2. The method for extracting electric vibration and acoustic signal characteristics of an operating mechanism in a railway high-voltage circuit breaker according to claim 1 is characterized in that: The different mechanical states include a normal state, a screw loose state, an electromagnet stuck state, and a spring aging state.

3. The method for extracting electric vibration and acoustic signal characteristics of an operating mechanism in a railway high-voltage circuit breaker according to claim 2 is characterized in that: By selecting multiple groups of coil current signals from each mechanical state for comparison, and verifying whether the coil current signals have the same current characteristics under different mechanical states; When the coil current signal has the same current characteristics in different mechanical states, characteristic points for distinguishing the coil current signal in different mechanical states are obtained by comparing the coil current signals in different mechanical states.

4. The method for extracting electric vibration and acoustic signal characteristics of an operating mechanism in a railway high-voltage circuit breaker according to claim 1, characterized in that: The characteristic points of the sound signal include the following steps: S41, through the conjugate orthogonal filter h k Filter the sound signal and set g k =(-1) k-1 h 1-k ; S42, by setting 2 according to the sound signal N sampling points, the sampling frequency is set to f s , the number of decomposition layers is set to M, and M < N; the wavelet packet decomposition technology is used to process the sound signal and the bandwidth of each layer of the decomposed signal is f s / 2 M+1 , the starting frequency of the nth frequency band is f = (n-1)f s / 2 M+1 ; S43, the decomposed signals of each layer are obtained by recursively decomposing the high-pass filtering result Wm 2n+1 and the low-pass filtering signal result Wm 2n according to formula (1), and the length of the high-pass filtering result Wm 2n+1 and the low-pass filtering signal result Wm 2n are both N / 2 M ; Where: n = 0, 1, 2, ..., 2 m-1 -1; W0 = f(t); i = 0, 1, 2, 3, ..., N / 2m-1; m = 0, 1, 2, 3, ..., M. When the decomposition reaches the mth level, the number of decomposed frequency bands is 2. m-1 ; S44, reconstructing the signal of the designated node according to formula (2) to obtain a signal after wavelet packet transformation; S45, extracting the energy value of each frequency band according to the wavelet packet space energy operator E(j,n); Among them, E(j,n) represents the energy value of the i-th node at the decomposition level j, p s (n,j,k) are the wavelet packet transform coefficients.

5. The method for extracting electric vibration and acoustic signal characteristics of an operating mechanism in a railway high-voltage circuit breaker according to claim 4, characterized in that: The percentage of the total energy accounted for by the energy value of each frequency band is calculated, and the mechanical state of the operating mechanism in the high-voltage circuit breaker is determined according to the different percentages under different mechanical states.

6. The method for extracting electric vibration and acoustic signal characteristics of an operating mechanism in a railway high-voltage circuit breaker according to claim 4, characterized in that: The energy value of each frequency band divided equally in time is obtained by wavelet packet decomposition. m This is achieved by dividing the frequency band signal into N equal parts in time and obtaining the time-frequency energy spectrum of the sound signal.

7. The method for extracting electric vibration and acoustic signal characteristics of an operating mechanism in a railway high-voltage circuit breaker according to claim 2, characterized in that: The feature extraction of vibration signals includes the following steps: S71, setting the number of overall averages M; S72, through the white noise n with standard normal distribution i (t) is added to the original signal x(t) to get the new signal x i (t): x i (t)=x(t)+n i (t); Where: n i (t) represents the i-th white noise sequence added, x i (t) represents the i-th additional noise signal, i = 1, 2, 3, ..., M; S73, using EMD to i (t) decomposes to obtain the IMF components: Where: c i,j (t) is the jth IMF component when the i-th white noise is added, r i,j (t) represents the residual component, indicating the signal movement trend, and J is the number of IMF components; S74, repeat steps S62 and S63 M times, and obtain the set of IMFs obtained by adding white noise with different amplitudes to each decomposition: c 1,j (t)c 2,j (t)…c M,j (t),j=1,2,…,J; S75, performing a set average operation on the IMF set obtained in step S64 by adopting the principle that the statistical average value of unrelated sequences is zero, to obtain the final IMF after EEMD decomposition: In the formula, c j (t) is the j-th IMF component of EEMD decomposition, i = 1, 2, 3, ..., M, j = 1, 2, 3, ..., J; S76. The threshold TH is obtained by the correlation coefficient r(j) between the IMF and the original signal, and the effective IMF is screened out according to the comparison result between the IMF and the threshold TH: Where: r j is the correlation coefficient between the jth IMF component and the original signal, t is the sampling point of the signal; when r j When it is greater than the threshold TH, the jth IMF component is retained, otherwise it is removed.

8. The method for extracting electric vibration and acoustic signal characteristics of an operating mechanism in a railway high-voltage circuit breaker according to claim 2, characterized in that: The calculation of the box dimension based on the IMF component comprises the following steps: S81, dividing the IMF component equally on the time axis and determining the grayscale value in the box according to the amplitude of the IMF component in the time period; S82. Calculate the box dimension of the IMF component according to the maximum grayscale value and the minimum grayscale value in the box.

Images