Method for synchronously extracting multiple channels of composite fault features of electric locomotive running gear bearing

Abstract

This invention provides a method for synchronously extracting multi-channel composite fault features of bearings in the running gear of electric locomotives. The method mainly includes: Step S1, preprocessing the multi-channel vibration signals of composite faults in the bearings of electric locomotives using a tensor synchronous denoising method based on higher-order singular value decomposition; Step S2, adaptively filtering and decomposing the preprocessed signals using a multi-layer K-value MVMD algorithm; and Step S3, calculating the fault peak factor of the envelope spectrum of each channel and plotting a tower-shaped EC diagram. Based on the tower-shaped EC diagram, the optimal analysis result of the multi-channel vibration signal is selected and output, synchronously extracting the composite fault features of the bearings. This method overcomes the problems of severe noise interference and ineffective detection of composite faults encountered by traditional methods when processing multi-channel signals of composite faults in electric locomotive bearings. It achieves synchronous and visualized output of composite faults, providing a strong basis for the extraction and identification of weak and composite fault features in complex dynamic signals of the running gear.

Description

Technical Field

[0001] This invention belongs to the field of electric locomotive fault analysis technology, specifically relating to a multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives. Background Technology

[0002] For electric locomotives, the continuous increase in railway speed and heavy-haul transportation places higher demands on their running gear. The running gear is one of the most critical parts of a locomotive; it is the part that propels the entire locomotive body along the rails under traction power. It is required to maintain a stable running condition on the rails and to maintain good performance even on curves. Therefore, the design, production, and maintenance quality of the running gear are subject to strict requirements. Rolling bearings are one of the important components of the electric locomotive's running gear. During operation, the running gear bearings are subjected to high-temperature and heavy-load environments for extended periods, making them more prone to failure and affecting the safe operation of the electric locomotive. Therefore, a method for accurately identifying running gear bearing failures is needed, which is of great significance for improving operational safety.

[0003] Existing methods for diagnosing bearings in the running gear of electric locomotives rely on temperature and vibration signals. For example: (1) Patent application CN202110856057.1 discloses a noise reduction algorithm based on complementary set empirical mode decomposition, which highlights the high-frequency resonance components in the signal and improves the diagnostic effect of resonance demodulation technology in the fault diagnosis of bearings in the running gear of trains. (2) Patent application CN201510482513.5 discloses a method for detecting locomotive bearing cage faults by combining temperature detection and vibration detection methods. (3) Patent application CN202121629769.1 discloses an online monitoring system for bearings in the running gear of locomotives and rolling stock, which mainly optimizes the detection system to address the situation of missed and false alarms caused by electromagnetic interference in the current locomotive and rolling stock operating environment.

[0004] Most of the aforementioned methods for diagnosing bearing faults in electric locomotive running gear still focus on single-channel signals or single faults. However, the vibration modes of the locomotive's running gear body and tracks couple with the bearing's vibration signals, causing the bearing's fault characteristic signals to be submerged in noise, making it difficult to accurately identify the fault components. Furthermore, the harsh working environment of electric locomotive running gear bearings can easily lead to multiple faults, with various fault characteristics superimposed and interfering with each other, increasing the difficulty of extracting complex fault features. Using single-fault diagnostic methods may result in misdiagnosis, necessitating more effective signal detection methods.

[0005] With the development of bearing fault diagnosis technology, signal acquisition has evolved from single-channel to multi-channel, multi-dimensional synchronous acquisition. Multi-channel signals can reflect rich status information of equipment, but how to effectively process them has become a new problem. Traditional methods treat each multi-dimensional signal as a single-channel signal. Multi-channel signals contain potential structural information between each channel, which has high-dimensional characteristics. Tensors are expansions of vectors and matrices in high-dimensional space and are a natural expression of high-dimensional signals. Therefore, addressing the problems in the synchronous extraction of composite fault features from bearings in the running gear of electric locomotives, this invention combines a tensor synchronous denoising method based on high-order singular value decomposition and multilevel K-valued multivariate variational mode decomposition (MVMD) to achieve synchronous adaptive filtering and decomposition of multi-channel signals. Finally, for typical bearing damage, the optimal analysis results of the multi-channel signals are visualized based on the peak factor map of the tower-shaped envelope spectrum, achieving synchronous extraction of composite bearing fault features. Patents and literature similar to the basic theory of this invention include: (1) Invention patent CN202110826154.6 discloses a method for suppressing random noise in three-dimensional seismic data based on MVMD and multi-channel singular spectrum analysis. This invention first uses MVMD to decompose three-dimensional seismic data into multiple narrowband data volumes. These data volumes help to select multi-channel singular spectrum analysis parameters, thereby helping to suppress random noise and protect effective waves. (2) In the journal article "Research on Early Fault Feature Extraction of Rolling Bearings Based on MVMD and Fractional Fourier Transform", Hong Da et al. proposed a feature extraction method based on MVMD and fractional Fourier transform. This method is applied to the fault diagnosis of rolling bearings to effectively avoid mode aliasing, make full use of fault feature information, weaken the interference of low-frequency signals and noise, and effectively extract the fault feature information of rolling bearings. (3) Invention patent CN202210121549.0 provides a method for diagnosing composite faults of rolling bearings. This method decomposes the vibration signal of the acquired rolling bearing, then performs noise reduction and evaluation processing to obtain a filtered signal, and then performs envelope analysis processing on the filtered signal to diagnose composite faults of rolling bearings. While the aforementioned patents and literature have improved upon traditional fault diagnosis and signal processing methods, they still have shortcomings. For example, invention patent CN202210121549.0 shares a similar approach with this invention, but it only uses a single-channel signal, making it difficult to accurately extract complex fault features from the complex multi-channel signals of electric locomotive running gear bearings. Similarly, the method proposed by Hongda, based on MVMD, only studies single faults and has weak identification capabilities for complex faults in electric locomotive running gear bearings. These problems lead to poor performance of these methods when applied to the simultaneous multi-channel extraction of complex fault features from electric locomotive running gear bearings, resulting in the inability to identify complex faults. Summary of the Invention

[0006] This invention addresses the aforementioned problems and aims to provide a multi-channel synchronous extraction method for bearing composite fault features that can overcome the influence of complex noise in the working environment of electric locomotives and can be effectively applied to the analysis of composite faults in electric locomotives. The invention employs the following technical solution:

[0007] This invention provides a multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives, characterized by comprising the following steps:

[0008] Step S1: Use the tensor synchronization denoising method based on high-order singular value decomposition to preprocess the multi-channel vibration signal of the composite fault of the running gear bearing of the electric locomotive.

[0009] Step S2: The preprocessed multi-channel vibration signal is adaptively filtered and decomposed using a multi-layer K-value MVMD algorithm to obtain the decomposition result;

[0010] Step S3: Based on the decomposition results, calculate the fault peak factor of the envelope spectrum of each channel and draw the tower-type EC diagram. Based on the tower-type EC diagram, select and output the optimal analysis result of the multi-channel vibration signal, and simultaneously extract the bearing composite fault features.

[0011] The multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives provided by the present invention may also have the following technical features, wherein step S1 includes the following sub-steps:

[0012] Step S1-1: Reconstruct the multi-channel vibration signal into a trajectory matrix using phase space;

[0013] Step S1-2: Arrange the trajectory matrix along the third dimension into multiple tensor forward slices, and superimpose the multiple tensor forward slices to form a tensor;

[0014] Step S1-3: Perform higher-order singular value decomposition on the tensor to obtain a singular value diagonal matrix, which contains a left singular value matrix.

[0015] Step S1-4: Calculate the singular entropy increment of the singular value diagonal matrix, and take the first point where the singular entropy increment tends to stabilize after the first rapid decrease as the denoising singular order, where rapid decrease means that the decrease rate of the singular entropy increment exceeds a preset threshold.

[0016] Steps S1-5: The singular value diagonal matrix is ​​truncated into a truncated singular value diagonal matrix by the noise reduction singular order, and the truncated left singular value matrix is ​​also obtained.

[0017] Steps S1-6: Based on the truncated left singular value matrix, solve for the core tensor and the denoised tensor;

[0018] Steps S1-7: Inverse transform the denoised tensor into the denoised multichannel vibration signal.

[0019] The multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives provided by this invention also has the following technical feature, wherein the acquired multi-channel vibration signals are represented as follows:

[0020] y = [y1(t); y2(t); ...; y q [(t)], t=1,…,N

[0021] In the formula, N represents the number of sampling points, and q represents the number of channels.

[0022] In step S1-1, the phase space reconstruction is represented as:

[0023]

[0024] In the formula, m represents the embedding dimension, τ represents the delay time, and n represents the window length.

[0025] In step S1-2, the tensor forward slice is represented as:

[0026]

[0027] In the formula, It is a third-order tensor.

[0028] In steps S1-3, the third-order tensor is... Perform higher-order singular value decomposition to obtain the singular value diagonal matrix Σ. (n) , is represented as:

[0029]

[0030] In the formula, U (n) For a third-order tensor The left singular value matrix of the pattern-n expansion matrix, V (n) For a third-order tensor The right singular value matrix of the pattern-n expansion matrix, Σ (n) =[diag{σ1 n ,σ2 n ,...,σ i n},σ i n ≥σ i-1 n [A third-order tensor] The singular value diagonal matrix of the pattern-n expansion matrix,

[0031] In steps S1-4, the singular entropy increment ΔSEs n Expressed as:

[0032]

[0033] In the formula, s is the order of the singular entropy increment.

[0034] In steps S1-5, the truncated singular value diagonal matrix is ​​expressed as:

[0035]

[0036] The truncated left singular value matrix is ​​expressed as:

[0037]

[0038] In steps S1-6, the core tensor Represented as:

[0039]

[0040] The denoising tensor Represented as:

[0041]

[0042] In steps S1-7, the noise-reduced multi-channel vibration signal is represented as follows:

[0043]

[0044] The multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives provided by this invention also has the following technical feature: In step S1-1, during the phase space reconstruction, to accelerate the calculation speed and convert it into a square matrix, even if n = τ, In the formula, ceil represents rounding up, and Y i If there are any empty spaces, fill them with 0.

[0045] The multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives provided by this invention may also have the following technical features, wherein, in step S2, the decomposition result includes multiple IMF component signals, and step S3 includes the following sub-steps:

[0046] Step S3-1: Calculate the envelope spectrum peak factor of various typical faults of the bearing from each of the IMF component signals;

[0047] Step S3-2: Find the maximum value of each envelope spectrum peak factor, and locate the corresponding multi-layer K-value MVMD algorithm decomposition layer K, IMF number k, channel number i and fault type f based on the maximum value;

[0048] Step S3-3: Use the number of decomposition layers K, the IMF number k, the number of channels i, and the fault type f to draw a tower-type EC diagram. In the diagram, the vertical axis is divided into L levels, each level representing a different number of decomposition layers K. The horizontal axis corresponds to each IMF of k. Different colors of the blocks represent different channels, and different block filling methods represent different fault types.

[0049] Step S3-4: Output the time-domain waveform and envelope spectrum of the optimal decomposition result based on the synchronous selection and extraction of the tower-type EC diagram. Combine the characteristic frequencies of various typical faults to determine whether the frequency in the envelope spectrum matches the characteristic frequency of the typical fault, thereby realizing the composite fault diagnosis of the running gear bearing of the electric locomotive.

[0050] The multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives provided by this invention may also have the following technical feature: In step S3-1, the calculated envelope spectrum peak factor includes... These are the bearing inner ring failure factor, bearing outer ring failure factor, rolling element failure factor, and cage failure factor, respectively, and their expressions are:

[0051]

[0052] In the formula, f is the fault characteristic frequency of the running gear bearing of the electric locomotive, P is the harmonic number, and E(hf) represents the amplitude of the h-th harmonic of the result component obtained by the MVMD algorithm for each K value. It is the average value of the envelope spectrum amplitude. The four typical fault characteristic frequencies of the running gear bearings of electric locomotives are represented by the inner ring fault frequency f. i Outer ring fault frequency f o Rolling element failure frequency f b cage failure frequency f c .

[0053] The multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives provided by this invention may also have the following technical feature, wherein the inner ring fault frequency f i The calculation formula is:

[0054]

[0055] The outer ring fault frequency f o The calculation formula is:

[0056]

[0057] The rolling element failure frequency f b The calculation formula is:

[0058]

[0059] The cage failure frequency f c The calculation formula is:

[0060]

[0061] In the formula, f r The value represents the rotational frequency, B represents the number of rotating elements, d represents the diameter of the rolling elements, D represents the pitch diameter, and β represents the contact angle of the bearing.

[0062] Invention Function and Effect

[0063] The multi-channel synchronous extraction method for composite fault features of electric locomotive running gear bearings according to the present invention employs a tensor synchronous denoising method based on high-order singular value decomposition to preprocess the acquired multi-channel vibration signals of composite bearing faults. This effectively reduces noise components in the signals, thus solving the problem of complex fault feature extraction caused by the complex working environment and high noise levels of electric locomotive bearings. Furthermore, by using a multi-layer K-value MVMD algorithm for synchronous adaptive filtering and decomposition of multi-channel vibration signals, the drawbacks of manual parameter setting are avoided, enabling adaptive selection of the critical K value of MVMD. Moreover, by using multi-channel envelope spectrum fault peak factor and corresponding tower-type EC diagram for selection analysis, the composite bearing fault features can be accurately, intuitively, and efficiently extracted synchronously, thereby facilitating the diagnosis of composite bearing faults.

[0064] In summary, the method of the present invention can overcome the problems of severe noise interference and ineffective detection of composite faults encountered by traditional methods when processing multi-channel signals of composite faults in electric locomotive bearings. It can also achieve visualized synchronous output of composite faults, providing a strong basis for the extraction and identification of weak and composite fault features in the complex dynamic signals of the running gear of electric locomotives. Attached Figure Description

[0065] Figure 1 This is a flowchart of the multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives in this embodiment of the invention;

[0066] Figure 2 These are the time-domain waveform and frequency-domain waveform of the two-channel vibration signal of the combined fault of the running gear bearing of the electric locomotive collected in this embodiment of the invention;

[0067] Figure 3 This is a tower-type EC diagram, which is a visual output of a composite fault in the running gear bearing of an electric locomotive, in an embodiment of the present invention.

[0068] Figure 4These are the optimal IMF components and envelope spectrum obtained from solving the vibration signal of the combined fault of the running gear bearing in the embodiment of the present invention.

[0069] Figure 5 This is the decomposition result and envelope spectrum of the vibration signal of the composite fault of the running gear bearing of an electric locomotive, which is decomposed using the traditional fixed-value MVMD method in this embodiment of the invention.

[0070] Figure 6 This is the decomposition result and envelope spectrum of the vibration signal of the composite fault of the running gear bearing of an electric locomotive, which is decomposed using the Integrated Empirical Mode Decomposition (EEMD) method in this embodiment of the invention.

[0071] Figure 7 This is a schematic diagram of the two-channel calculation results for calculating the composite fault vibration signal of the running gear bearing of an electric locomotive using the fast spectral kurtosis method in an embodiment of the present invention. Detailed Implementation

[0072] To make the technical means, creative features, objectives and effects of this invention easy to understand, the following describes in detail the method for multi-channel synchronous extraction of composite fault features of bearings in the running gear of electric locomotives.

[0073] <Example>

[0074] Figure 1 This is a flowchart of the multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives in this embodiment.

[0075] like Figure 1 As shown, the multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives specifically includes the following steps:

[0076] Step S0: Collect multi-channel vibration data (signals) when the running gear bearing of the electric locomotive is in a fault state.

[0077] In this embodiment, the bearing of the running gear of the electric locomotive under test is model 552732QT, and the sensor used to collect vibration signals is an acceleration sensor. In order to simulate the actual working conditions, a load of 5000N is applied to the bearing through hydraulic system drive and loading.

[0078] First, data was collected from the running gear bearing that was in a faulty state. The collected multi-channel vibration signals are represented as follows:

[0079] y = [y1(t); y2(t); ...; y q [(t)], t=1,…,N

[0080] In the formula, N represents the number of sampling points and q represents the number of channels.

[0081] In this embodiment, the tested bearing has faults in both the outer and inner rings, i.e., a combined fault. Two-channel vibration signals of the bearing are collected during the operation of the traveling section, with 7000 sampling points. The collected two-channel vibration signals can be represented as follows:

[0082] y=[y1(t);y2(t)],t=1,…,7000

[0083] Figure 2 These are the time-domain waveforms and frequency-domain waveforms of the two-channel vibration signals of the combined bearing fault in the running gear of the electric locomotive collected in this embodiment.

[0084] like Figure 2 As shown, the frequencies marked by the ellipses in the time-domain waveform and frequency-domain graphs reveal the outer ring fault characteristics of the bearing, but other fault characteristics cannot be identified.

[0085] Step S1: The multi-channel vibration signal acquired in step S0 is preprocessed using a tensor synchronization denoising method based on higher-order singular value decomposition.

[0086] Step S1 specifically includes the following sub-steps:

[0087] Step S1-1: Reconstruct the multi-channel vibration signal y into a trajectory matrix Y using phase space. i The expression for phase space reconstruction is:

[0088]

[0089] In the formula, m represents the embedding dimension, τ represents the delay time, and n represents the window length. To speed up the calculation, it is converted into a square matrix such that n = τ. Where ceil represents rounding up, and Y i If there are any empty spaces, fill them with 0.

[0090] In this embodiment, based on the two-channel vibration signals obtained in step S0, the expression for phase space reconstruction is:

[0091]

[0092] In the formula, Y, similarly i Fill in the empty spaces with 0.

[0093] Step S1-2: Arrange the trajectory matrix obtained in step S1-1 along the third dimension into multiple tensor forward slices, i.e. Then, these forward slices of tensor are stacked to form a third-order tensor. .

[0094] Steps S1-3, for the 3rd order tensor Perform higher-order singular value decomposition to obtain the singular value diagonal matrix Σ. (n) Its expression is as follows:

[0095]

[0096] In the formula, U (n) For tensor The left singular value matrix of the pattern-n expansion matrix, V (n) For tensor The right singular value matrix of the pattern-n expansion matrix, Σ (n) =[diag{σ1 n ,σ2 n ,...,σ i n},σ i n ≥σ i-1 n ] is a tensor The singular value diagonal matrix of the pattern-n expansion matrix.

[0097] Steps S1-4: Calculate the singular value diagonal matrix Σ (n) The singular entropy increment ΔSE s n And take the singular entropy increment ΔSE s n The first point at which the price stabilizes after the first rapid descent (i.e., the descent rate exceeds a preset threshold) is the singular order p of the noise reduction process. n The expression is as follows:

[0098]

[0099] In the formula, s is the order of the singular entropy increment.

[0100] Steps S1-5 involve denoising the singular order p. n The original singular value diagonal matrix Σ (n) =[diag{σ1 n ,σ2 n ,...,σ i n},σ i n ≥σ i-1 n Extracted as a truncated singular value diagonal matrix Similarly, the truncated left singular value matrix can be obtained.

[0101] Steps S1-6: Based on the truncated left singular value matrix obtained in step S1-5, solve for the core tensor. and denoising tensor Core Tensor Represented as:

[0102]

[0103] Denoising Tensor Represented as:

[0104]

[0105] Steps S1-7, denoise the tensor Inverse transformation into a denoised multi-channel vibration signal:

[0106]

[0107] In this embodiment, the denoising tensor Inverse transformation to a denoised two-channel vibration signal:

[0108]

[0109] The core tensor obtained from higher-order singular value decomposition is beneficial for exploring the tensor structure of multidimensional data, and its local reconstruction is an important method for multi-channel signal denoising. The denoised signal will greatly improve the accuracy of the subsequent multi-layer K-valued MVMD.

[0110] Step S2: The multi-channel signal preprocessed in step S1 is subjected to adaptive filtering and decomposition using the multi-layer K-value MVMD algorithm to obtain the decomposition result, i.e., multiple IMF component signals.

[0111] In this embodiment, the decomposition results of each layer of MVMD are calculated when K values ​​are between 4 and 10. The MVMD algorithm is existing technology and will not be described in detail here.

[0112] By designing a multi-level K-value MVMD decomposition, the drawbacks of manual parameter setting can be avoided, and the adaptive selection of the critical K value of MVMD can be achieved.

[0113] Step S3: Based on the multiple IMF component signals obtained from step S2, calculate the fault peak factor of the envelope spectrum of each channel, output the optimal analysis results of the multi-channel vibration signal, and simultaneously extract the bearing composite fault characteristics.

[0114] Step S3 specifically includes the following sub-steps:

[0115] Step S3-1: Calculate the envelope spectrum peak factor (EC) for the four typical bearing faults based on each IMF result from the multi-layer K-value MVMD. Its expression is:

[0116]

[0117] In the formula, f is the fault characteristic frequency of the running gear bearing of the electric locomotive, P is the harmonic number, and E(hf) represents the amplitude of the h-th harmonic of the result component obtained by the MVMD algorithm for each K value. It is the average value of the envelope spectrum amplitude. The four typical fault characteristic frequencies of the running gear bearings of electric locomotives are represented by the inner ring fault frequency f. i =43.3Hz, outer ring fault frequency f o =58.7Hz, rolling element failure frequency f b =19.4Hz, cage failure frequency f c =2.6Hz, and the calculation formulas are as follows:

[0118]

[0119]

[0120]

[0121]

[0122] In the formula, f r The value represents the rotational frequency, B represents the number of rotating elements, d represents the diameter of the rolling elements, D represents the pitch diameter, and β represents the contact angle of the bearing.

[0123] Step S3-2, all the results calculated in step S3-1 Find the maximum value of the peak factor of each envelope spectrum, and locate the corresponding multi-layer K value MVMD algorithm decomposition layer number K, IMF number k, channel number i and fault type f according to the maximum value, that is, optimize these parameters.

[0124] In this embodiment, the largest The IMF2 at channel 2 with decomposition layer K=8 is the largest. The largest IMF4 is located when the number of decomposition layers K=4 in channel 1. The IMF5 at channel 1 with decomposition layer K=5 is the largest. The IMF2 is located when the number of decomposition layers K = 6 in channel 2.

[0125] Step S3-3: Use the decomposition layer number K, IMF number k, channel number i, and fault type f obtained in step S3-2 to draw the tower type EC diagram.

[0126] Figure 3 This is a tower-type EC diagram, which is a visual output of a combined fault in the running gear bearings of an electric locomotive in this embodiment.

[0127] like Figure 3As shown, the vertical axis of the tower EC diagram is divided into L levels, with L ranging from 4 to 10. Each level represents the decomposition parameter K in different MVMDs. The horizontal axis of the tower EC diagram corresponds to each IMF of k. Different colors of the blocks in the tower EC diagram represent different channels, and different block filling methods represent different fault types.

[0128] Step S3-4: Output the time-domain waveform and envelope spectrum of the optimal decomposition results based on the synchronous selection and extraction of the tower-type EC diagram. Combine the characteristic frequencies of various typical faults of the running gear bearings of electric locomotives to check whether the fault characteristics in the envelope spectrum are obvious (i.e., whether the frequencies in the envelope spectrum match the characteristic frequencies of typical faults), thereby realizing the diagnosis of composite faults of the running gear bearings of electric locomotives.

[0129] Figure 4 These are the optimal IMF components and envelope spectrum obtained from solving the vibration signal of the combined fault in the running gear bearing of the electric locomotive in this embodiment. Figure 4 (a)-(d) are the maximum values ​​respectively. maximum maximum maximum The corresponding result.

[0130] like Figure 4 As shown, according to Figure 4 (a) Inner ring faults can be easily found. i and its harmonics, according to Figure 4 (b) Outer ring fault f o Its harmonics are also quite evident. However, Figure 4 (c) Figure 4 (d) No rolling element or cage faults could be detected. In summary, the fault of the tested electric locomotive running gear bearing can be diagnosed as a composite fault including inner ring fault and outer ring fault, which is consistent with the actual bearing fault situation in this embodiment.

[0131] Figure 5-7 The results of signal decomposition of bearing faults in this embodiment using existing decomposition algorithms are shown, and compared with the effect of the multi-layer K-value MVMD algorithm in this embodiment.

[0132] Figure 5 This embodiment presents the decomposition results and envelope spectrum of the vibration signal from a composite fault in the running gear bearing of an electric locomotive, obtained using the traditional fixed-value MVMD algorithm. Figure 5 (a) shows the decomposition results. Figure 5 (b) is the envelope spectrum.

[0133] from Figure 5 (b) The envelope spectrum shows two channels with bearing outer ring faults. o(Indicated by dashed lines), but no characteristic frequency of inner ring fault is found in any component of either channel.

[0134] Figure 6 This embodiment presents the decomposition results and envelope spectrum of the vibration signal from a composite fault in the running gear bearing of an electric locomotive, obtained using the EEMD algorithm. Figure 6 (a) shows the decomposition results. Figure 6 (b) is the envelope spectrum.

[0135] from Figure 6 (b) The envelope spectrum shows that each component of both channels contains bearing outer race fault f. o (Indicated by dashed lines), but no characteristic frequency of inner ring fault is found in any component of either channel.

[0136] Figure 7 This is a schematic diagram of the two-channel calculation results for calculating the composite fault vibration signal of the running gear bearing of an electric locomotive using the fast spectral kurtosis algorithm in this embodiment. Figure 7 (a) shows the calculation results for channel 1. Figure 7 (b) shows the calculation results for channel 2. Figure 7 (a) Figure 7 (b) shows the center frequency and bandwidth parameters of the filter selected by the fast spectral kurtosis on the left, and the time-domain waveform and frequency-domain graph after filtering on the right.

[0137] from Figure 7 As can be seen, the fast spectrum kurtosis only extracts the bearing outer ring fault f. o (Indicated by dashed lines), but the characteristic frequency of the inner ring fault cannot be extracted.

[0138] In summary, for the bearing composite faults in this embodiment, none of the three existing algorithms mentioned above can extract the characteristic frequency of the inner ring fault. However, the method of this embodiment can extract the bearing composite faults more effectively, providing a reliable basis for fault diagnosis of bearings in the running gear of electric locomotives.

[0139] Functions and effects of the embodiments

[0140] The multi-channel synchronous extraction method for composite fault features of bearings in the running gear of electric locomotives provided in this embodiment effectively reduces noise components in the signals by employing a tensor synchronous denoising method based on high-order singular value decomposition to preprocess the acquired multi-channel vibration signals of composite bearing faults. This solves the problem of complex fault feature extraction caused by the complex working environment and high noise levels of electric locomotive bearings. Furthermore, the use of a multi-layer K-value MVMD algorithm for synchronous adaptive filtering and decomposition of multi-channel vibration signals avoids the drawbacks of manual parameter setting and enables adaptive selection of the critical K value of MVMD. Moreover, by using the fault peak factor of the multi-channel envelope spectrum and the corresponding tower-type EC diagram for selection analysis, the composite bearing fault features can be extracted accurately, intuitively, and efficiently, which is beneficial for the diagnosis of composite bearing faults.

[0141] In summary, the method of this embodiment can overcome the problems of severe noise interference and ineffective detection of composite faults encountered by traditional methods when processing multi-channel signals of composite faults in electric locomotive bearings. It can also achieve the visualized synchronous output of composite faults, providing a strong basis for the extraction and identification of weak and composite fault features in the complex dynamic signals of the running gear of electric locomotives.

[0142] The above embodiments are only used to illustrate specific implementations of the present invention, and the present invention is not limited to the scope of the description of the above embodiments.

Claims

1. A method for simultaneous multi-channel extraction of composite fault features of bearings in the running gear of an electric locomotive, characterized in that, Includes the following steps: Step S1: Use the tensor synchronization denoising method based on high-order singular value decomposition to preprocess the multi-channel vibration signal of the composite fault of the running gear bearing of the electric locomotive. Step S2: The preprocessed multi-channel vibration signal is adaptively filtered and decomposed using a multi-layer K-value MVMD algorithm to obtain the decomposition result; Step S3: Based on the decomposition results, calculate the fault peak factor of the envelope spectrum of each channel and plot the tower-shaped EC diagram. Based on the tower-shaped EC diagram, select and output the optimal analysis result of the multi-channel vibration signal, and simultaneously extract the bearing composite fault characteristics. Step S1 includes the following sub-steps: Step S1-1: Reconstruct the multi-channel vibration signal into a trajectory matrix using phase space; Step S1-2: Arrange the trajectory matrix along the third dimension into multiple tensor forward slices, and superimpose the multiple tensor forward slices to form a tensor; Step S1-3: Perform higher-order singular value decomposition on the tensor to obtain a singular value diagonal matrix, which contains a left singular value matrix. Step S1-4: Calculate the singular entropy increment of the singular value diagonal matrix, and take the first point where the singular entropy increment tends to stabilize after the first rapid decrease as the denoising singular order, where rapid decrease means that the decrease rate of the singular entropy increment exceeds a preset threshold. Steps S1-5: The singular value diagonal matrix is ​​truncated into a truncated singular value diagonal matrix by the noise reduction singular order, and the truncated left singular value matrix is ​​also obtained. Steps S1-6: Based on the truncated left singular value matrix, solve for the core tensor and the denoised tensor; Steps S1-7: Inverse transform the denoised tensor into the denoised multichannel vibration signal.

2. The method for multi-channel synchronous extraction of composite fault features of bearings in the running gear of electric locomotives according to claim 1, characterized in that: in, The acquired multi-channel vibration signals are represented as follows: In the formula, N represents the number of sampling points, and q represents the number of channels. In step S1-1, the phase space reconstruction is represented as: In the formula, m represents the embedding dimension, τ represents the delay time, and n represents the window length. To speed up the computation, it is converted into a square matrix as much as possible. , ceil represents rounding up, where If there are any empty spaces, fill them with 0. In step S1-2, the tensor forward slice is represented as: In the formula, It is a third-order tensor. In steps S1-3, the third-order tensor is... Perform higher-order singular value decomposition to obtain the singular value diagonal matrix. , is represented as: In the formula, U (n) For a third-order tensor The left singular value matrix of the pattern-n expansion matrix, V (n) For a third-order tensor The right singular value matrix of the pattern-n expansion matrix, For a third-order tensor The singular value diagonal matrix of the pattern-n expansion matrix, In steps S1-4, the singular entropy increment Expressed as: In the formula, s is the order of the singular entropy increment. In steps S1-5, the truncated singular value diagonal matrix is ​​expressed as: The truncated left singular value matrix is ​​expressed as: In steps S1-6, the core tensor Represented as: The denoising tensor Represented as: In steps S1-7, the noise-reduced multi-channel vibration signal is represented as follows: 。 3. The method for multi-channel synchronous extraction of composite fault features of bearings in the running gear of electric locomotives according to claim 1, characterized in that: in, In step S2, the decomposition result includes multiple IMF component signals. Step S3 includes the following sub-steps: Step S3-1: Calculate the envelope spectrum peak factor of various typical faults of the bearing from each of the IMF component signals; Step S3-2: Find the maximum value of each envelope spectrum peak factor, and locate the corresponding multi-layer K-value MVMD algorithm decomposition layer K, IMF number k, channel number i and fault type f based on the maximum value; Step S3-3: Use the number of decomposition layers K, the IMF number k, the number of channels i, and the fault type f to draw the tower-type EC diagram. In the diagram, the vertical axis is divided into L levels, each level representing a different number of decomposition layers K. The horizontal axis corresponds to each IMF of k. Different colors of the blocks represent different channels, and different block filling methods represent different fault types. Step S3-4: Output the time-domain waveform and envelope spectrum of the optimal decomposition result based on the synchronous selection and extraction of the tower-type EC diagram. Combine the characteristic frequencies of various typical faults to determine whether the frequency in the envelope spectrum matches the characteristic frequency of the typical fault, thereby realizing the composite fault diagnosis of the running gear bearing of the electric locomotive.

4. The method for multi-channel synchronous extraction of composite fault features of bearings in the running gear of electric locomotives according to claim 3, characterized in that: in, In step S3-1, the calculated envelope spectrum peak factor includes These are the bearing inner ring failure factor, bearing outer ring failure factor, rolling element failure factor, and cage failure factor, respectively, and their expressions are: In the formula, f is the fault characteristic frequency of the running gear bearing of the electric locomotive, and P is the harmonic number. This represents the amplitude of the h-th harmonic of the result component obtained by the MVMD algorithm for each K value. It is the average value of the envelope spectrum amplitude. The four typical fault characteristic frequencies of the running gear bearings of electric locomotives are respectively represented as the inner ring fault frequencies. outer ring failure frequency rolling element failure frequency cage failure frequency .

Images