A bearing fault diagnosis method and system based on time-frequency envelope spectrum peak analysis

Abstract

The present invention discloses a bearing fault diagnosis method and system based on time-frequency envelope spectrum peak analysis, which relates to the technical field of time-frequency analysis of non-stationary rotating machinery fault signals. The method includes obtaining the time-frequency spectrum of the vibration signal to be diagnosed, preprocessing the time-frequency spectrum to obtain the preprocessed envelope; constructing a fast time-frequency envelope spectrum peak diagram, and using the fast time-frequency envelope spectrum peak diagram to decompose the preprocessed envelope; extracting the frequency point with the most prominent pulse characteristics in the fast time-frequency envelope spectrum peak diagram to obtain the time domain envelope spectrum peak; calculating the short-time Fourier transform result of the frequency point corresponding to the time domain envelope spectrum peak, which is the fault characteristic frequency, and diagnosing the bearing fault type through the fault characteristic frequency. The present invention can detect periodic pulse information in complex signals, while suppressing the influence of interference such as random pulses and harmonics, and can more accurately extract the fault characteristic frequency of the bearing and effectively diagnose the health status of the machinery.

Description

Technical Field

[0001] The present invention relates to the technical field of time-frequency analysis of non-stationary rotating machinery fault signals, and in particular to a bearing fault diagnosis method and system based on time-frequency envelope spectrum peak analysis. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] As a basic signal, periodic pulse signals play a very important role in many fields. Taking the field of signal processing as an example, the type of bearing fault is determined by analyzing the periodic pulse signals generated by damaged bearings of mechanical equipment. However, the vibration signals of faulty bearings of rotating machinery usually contain multiple components, and it is difficult to predict which types of signals should be included. During the operation of the equipment, periodic pulses, sporadic pulses, harmonics, and non-Gaussian noise containing fault information may be generated. How to accurately distinguish various components in complex vibration signals, extract effective sensitive signal features of faults, and then accurately diagnose bearing faults is the key to current bearing fault factor analysis. A key criterion for an effective bearing fault diagnosis procedure is the ability to detect information about defects at an early stage, even in the presence of machine operation noise.

[0004] Narrowband demodulation of vibration signals makes it possible to extract the components carrying fault information in rotating machinery. However, the quality of the demodulated signal depends on the frequency band selected for demodulation. Spectral kurtosis is currently a very effective method and is generally considered a powerful tool for pulse detection. Thanks to the significant efforts of Antoni, spectral kurtosis has become recognized as a milestone in characterizing nonstationary signals, particularly bearing fault signals. Some fault diagnosis methods determine the presence of periodic pulse components by comparing the kurtosis index of the original signal. However, spectral kurtosis cannot accurately detect periodic pulses in the presence of random noise and single pulses. One of the most serious limitations of spectral kurtosis is its inability to distinguish between repetitive pulses within a series of pulses. Furthermore, due to the uncontrollable signal-to-noise ratio, the periodic pulse components carrying fault information are often buried in the noise, making them difficult to effectively separate. In fact, kurtosis decreases with increasing pulse repetition rate. Furthermore, spectral kurtosis is sensitive to noise. Furthermore, due to the limitations of its theoretical basis, spectral kurtosis is not applicable to signals obtained from variable speed experiments on machinery.

[0005] Therefore, how to quickly detect periodic pulse information in complex signals during bearing fault diagnosis while suppressing the influence of interference such as random pulses and harmonics has become an urgent problem to be solved. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a bearing fault diagnosis method and system based on time-frequency envelope spectrum peak analysis. By using the fast time-frequency envelope spectrum peak diagram to obtain more periodic pulse information, it can detect periodic pulse information in complex signals while suppressing the influence of interference such as random pulses and harmonics. In the noisy working environment of rotating machinery, it can more accurately extract the fault characteristic frequency of the bearing and effectively diagnose the health status of the machinery.

[0007] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0008] A first aspect of the present invention provides a bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis, comprising the following steps:

[0009] Obtaining the time-frequency spectrum of the vibration signal to be diagnosed, preprocessing the time-frequency spectrum, and obtaining a preprocessed envelope;

[0010] Construct a fast time-frequency envelope spectrum peak map, and use the fast time-frequency envelope spectrum peak map to decompose the preprocessed envelope;

[0011] Extract the fault characteristic frequency from the fast time-frequency envelope spectrum peak diagram and diagnose the bearing fault type through the fault characteristic frequency;

[0012] Among them, the specific steps of extracting the fault characteristic frequency in the fast time-frequency envelope spectrum peak diagram include: extracting the frequency point with the most prominent pulse characteristics in the fast time-frequency envelope spectrum peak diagram to obtain the time domain envelope spectrum peak; calculating the short-time Fourier transform result of the frequency point corresponding to the time domain envelope spectrum peak, which is the fault characteristic frequency.

[0013] Furthermore, the time-frequency spectrum is preprocessed by performing a mean removal process on each frequency point in the signal time-frequency spectrum.

[0014] Furthermore, the specific process of extracting the frequency point with the most prominent pulse characteristics in the envelope and obtaining the peak value of the time domain envelope spectrum is as follows:

[0015] Calculate the envelope spectrum value of each frequency point of short-time Fourier transform;

[0016] The maximum value is used to represent the frequency point where the pulse feature in the fault signal is most prominent, which is the peak value of the time domain envelope spectrum.

[0017] Furthermore, the time domain envelope spectrum peak value calculation formula is:

[0018]

[0019] Among them, TFES(ω) is the envelope spectrum value of the frequency point, ω is the frequency point, N w is the window length of the window function, r is the current window length, G STFT (i,ωk ) is the short-time Fourier transform function, φ(ω) represents the average value of the short-time Fourier transform result at the frequency point ω, ω k is the discrete frequency, F s is the sampling frequency, and i is the time point corresponding to the current window length.

[0020] Furthermore, the short-time Fourier transform function G STFT (i,ω k ) is calculated as:

[0021]

[0022] Among them, ω k is the discrete frequency, g[r] is the window function, r is the current window length, N w is the window length of the window function, F s is the sampling frequency, i is the time point corresponding to the current window length, ω k is the discrete frequency, x is the vibration signal, and L is the time shift between consecutive windows.

[0023] Furthermore, the first level 0 of the fast time-frequency envelope spectrum peak map is the original signal to be analyzed. The second level 1 decomposes the signal into a binary tree structure, the third level 1.6 decomposes the signal into a 1 / 3 tree structure, the fourth level 2 decomposes the signal into a binary tree structure, the fifth level 2.6 decomposes the signal into a ternary tree structure, and so on for the remaining levels.

[0024] Furthermore, the accuracy of periodic pulse signal recognition is verified by using the error rate between the center frequency of the periodic pulse signal located by the time-frequency envelope spectrum peak and the inherent center frequency of the periodic pulse signal.

[0025] A second aspect of the present invention provides a bearing fault diagnosis system based on time-frequency envelope spectrum peak analysis, comprising:

[0026] a signal acquisition module configured to acquire a time-frequency spectrum of a vibration signal to be diagnosed, preprocess the time-frequency spectrum, and obtain a preprocessed envelope;

[0027] A fast time-frequency envelope spectrum peak diagram module is configured to construct a fast time-frequency envelope spectrum peak diagram and decompose the preprocessed envelope using the fast time-frequency envelope spectrum peak diagram;

[0028] A fault diagnosis module is configured to extract a fault characteristic frequency in the envelope and diagnose the bearing fault type according to the fault characteristic frequency;

[0029] Among them, the fault diagnosis module includes an envelope analysis module, which is configured to extract the fault characteristic frequency in the fast time-frequency envelope spectrum peak diagram. The specific steps include: extracting the frequency point with the most prominent pulse characteristics in the fast time-frequency envelope spectrum peak diagram to obtain the time domain envelope spectrum peak; calculating the short-time Fourier transform result of the frequency point corresponding to the time domain envelope spectrum peak, which is the fault characteristic frequency.

[0030] A third aspect of the present invention provides a medium having a program stored thereon, which, when executed by a processor, implements the steps of the bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis as described in the first aspect of the present invention.

[0031] The fourth aspect of the present invention provides a device comprising a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps of the bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis as described in the first aspect of the present invention are implemented.

[0032] One or more of the above technical solutions have the following beneficial effects:

[0033] This invention discloses a bearing fault diagnosis method and system based on time-frequency envelope spectrum peak analysis. To address the inability of spectral kurtosis to accurately detect periodic pulses, the invention defines the time-frequency envelope spectrum peak and constructs a fast time-frequency envelope spectrum peak diagram to rapidly identify the center frequency of periodic pulse signals. This method can detect periodic pulse information in complex signals while suppressing the influence of interference such as random pulses and harmonics, enabling efficient extraction of fault characteristic frequencies and ultimately obtaining more accurate mechanical bearing fault diagnosis results.

[0034] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0036] Figure 1 This is a flow chart of a bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis in Example 1 of the present invention;

[0037] Figure 2 Schematic diagram of the time domain waveform, short-time Fourier transform result, and spectral kurtosis result of the composite signal in Example 1 of the present invention when the signal-to-noise ratio is 5dB;

[0038] Figure 3Schematic diagram of the time domain waveform, short-time Fourier transform result, and spectral kurtosis result of the composite signal when the signal-to-noise ratio is 0 dB in Example 1 of the present invention;

[0039] Figure 4 Schematic diagram of the time domain waveform, short-time Fourier transform result, and spectral kurtosis result of the composite signal when the signal-to-noise ratio is -5dB in Example 1 of the present invention;

[0040] Figure 5 : is the time-frequency envelope diagram of the short-time Fourier transform result at the frequency f=200 Hz in the first embodiment of the present invention;

[0041] Figure 6 Schematic diagram of the time domain waveform, short-time Fourier transform result, and time-frequency envelope spectrum peak result of the composite signal when the signal-to-noise ratio is 5dB in Example 1 of the present invention;

[0042] Figure 7 Schematic diagram of the time domain waveform, short-time Fourier transform result, and time-frequency envelope spectrum peak result of the composite signal when the signal-to-noise ratio is 0 dB in Example 1 of the present invention;

[0043] Figure 8 Schematic diagram of the time domain waveform, short-time Fourier transform result, and time-frequency envelope spectrum peak result of the composite signal when the signal-to-noise ratio is -5dB in Example 1 of the present invention;

[0044] Figure 9 Schematic diagram of the error rate of locating the center frequency of the periodic pulse component in the composite signal using the time-frequency envelope spectrum peak at different signal-to-noise ratios in Example 1 of the present invention;

[0045] Figure 10 Schematic diagram of the waveforms of a sinusoidal signal, sporadic pulses, a periodic pulse signal, Gaussian white noise, a mixed signal, and the short-time Fourier transform results of the mixed signal in Example 1 of the present invention;

[0046] Figure 11 This is a decomposition structure diagram of the fast time-frequency envelope spectrum peak diagram in Example 1 of the present invention;

[0047] Figure 12 The result of fast time-frequency envelope spectrum peak diagram processing signal S(t), the time domain waveform of the filtered signal, and the square envelope spectrum of the filtered signal in embodiment 1 of the present invention;

[0048] Figure 13 The result of fast spectral kurtosis processing of the signal S(t), the time domain waveform of the filtered signal, and the square envelope spectrum of the filtered signal in the first embodiment of the present invention;

[0049] Figure 14 The time domain waveform of the signal after filtering out the second possible frequency band of the fast spectral kurtosis diagram in the first embodiment of the present invention and the square envelope spectrum of the filtered signal;

[0050] Figure 15 The waveform of the bearing outer ring vibration signal collected under the variable speed working condition and its short-time Fourier transform result in Example 1 of the present invention;

[0051] Figure 16 Schematic diagram of the result of processing the vibration signal using the fast time-frequency envelope spectrum peak diagram, the time domain waveform of the filtered signal, and the short-time Fourier transform result of the square envelope spectrum of the filtered signal in Example 1 of the present invention;

[0052] Figure 17 Schematic diagram of the result of fast spectral kurtosis diagram processing vibration signal, the time domain waveform of the filtered signal and the short-time Fourier transform result of the square envelope spectrum of the filtered signal in embodiment 1 of the present invention. DETAILED DESCRIPTION

[0053] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0054] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations;

[0055] Example 1:

[0056] The first embodiment of the present invention provides a bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis, such as Figure 1 As shown, the following steps are included:

[0057] Step 1: Obtain the time-frequency spectrum of the vibration signal to be diagnosed, preprocess the time-frequency spectrum, and obtain the preprocessed envelope.

[0058] Step 1.1: preprocessing the time-frequency spectrum is to remove the mean value of each frequency point in the signal time-frequency spectrum, and then take the processed envelope.

[0059] In step 1.2, the vibration signal collected from the bearings of rotating machinery often contains a variety of information, including fault information, equipment operating sounds, and environmental noise. The vibration characteristics of a rolling bearing with a local defect can be represented by an amplitude modulation process. Therefore, the vibration signal x(t) is modeled as follows:

[0060]

[0061] Among them, A k is the amplitude of the kth fault pulse, 2K is the number of pulses, v(t) is the unit step function, the time period corresponding to the fault characteristic frequency is T0, η is the structural damping characteristic coefficient, ω0 corresponds to the resonant frequency of the bearing excitation, t i As the i-th random variable with a zero-mean uniform distribution, its standard deviation is within the range of 0.02T0. n(t) is the sum of random noise, harmonics, and other interference from the surrounding environment. The role of random variables in rotating machinery bearing fault diagnosis is to model and analyze the random characteristics of relevant parameters such as vibration signals and sensor data, thereby helping to determine whether the bearing is faulty and providing a diagnostic basis for the fault type.

[0062] Step 2: Extract the fault characteristic frequency in the envelope and diagnose the bearing fault type through the fault characteristic frequency.

[0063] Step 2.1: Extract the frequency point with the most prominent pulse characteristics in the envelope to obtain the peak value of the time domain envelope spectrum.

[0064] Step 2.1.1, the short-time Fourier transform function G of the discrete signal x(n) (n is the sampling point) in the time interval of length Nw / Fs STFT (i,ω k ) is calculated as:

[0065]

[0066] Among them, ω k is the discrete frequency, ω k =2πkΔf,k=0,...,N w -1, g[r] is the window function, r is the current window length, N w is the window length of the window function, F s is the sampling frequency, i is the time point corresponding to the current window length, ω k is the discrete frequency, x is the vibration signal, and L is the time shift between consecutive windows.

[0067] Sampling frequency F s The frequency resolution is as follows:

[0068]

[0069] When a bearing defect occurs, its moving parts repeat the same trajectory, generating a series of periodic pulses. Considering that pulses typically have a large bandwidth, there should be a frequency point where the time-frequency amplitude is most significant, and this frequency point should also exhibit periodic regularity.

[0070] Step 2.1.2, calculate the envelope spectrum value of each frequency point representing the short-time Fourier transform:

[0071]

[0072] Here, φ(ω) represents the average value of the short-time Fourier transform result at frequency ω, and ζ is the fault characteristic frequency of the signal. The periodicity of the bearing resonance varies for different fault types. Therefore, the repetition frequency of this high-frequency resonance, or the fault characteristic frequency, can be used to determine the type of bearing fault.

[0073] In step 2.1.3, the maximum value is used to represent the frequency point where the pulse feature in the fault signal is most prominent, that is, the peak value of the time domain envelope spectrum. In order to further represent the frequency point where the pulse feature in the fault signal is most prominent, the maximum value of the result of formula (8) can be taken, and its expression is:

[0074]

[0075] Among them, TFES(ω) is the envelope spectrum value of the frequency point, which is the peak value of the time domain envelope spectrum. The peak value of the time domain envelope spectrum has periodicity and regularity. ω is the frequency point, N w is the window length of the window function, r is the current window length, G STFT (i,ω k ) is the short-time Fourier transform function, φ(ω) represents the average value of the short-time Fourier transform result at the frequency point ω, ω k is the discrete frequency, Fs is the sampling frequency, and i is the time point corresponding to the current window length.

[0076] The pulse characteristics of the bearing fault can be indicated based on the short-time Fourier transform result of the frequency point corresponding to the peak of the time-frequency envelope spectrum, that is, This formula expresses that the frequency point corresponding to the maximum TFES value is used to represent the pulse characteristics of the bearing fault.

[0077] Step 2.2: extract the time domain envelope spectrum peak value in the envelope to form a fast time-frequency envelope spectrum peak diagram.

[0078] The fast time-frequency envelope spectrum peak diagram is used to represent the time-frequency envelope spectrum peak of the signal on the two-dimensional plane of frequency and frequency resolution. The first layer of the fast time-frequency envelope spectrum peak diagram algorithm decomposes the signal into a binary tree structure, the second layer decomposes the signal into a 1 / 3 tree structure, and the rest are obtained by analogy. The decomposition structure diagram is as follows Figure 11As shown. The fast time-frequency envelope spectrum peak diagram is as accurate as the time-frequency envelope spectrum peak in identifying the center frequency of the periodic pulse signal. Therefore, when demodulating the selected resonant frequency band, the fast time-frequency envelope spectrum can obtain more periodic pulse information. The fast time-frequency envelope spectrum detects and characterizes the non-stationarity of the signal and adaptively selects the optimal bandpass filter band as a preprocessing for envelope spectrum analysis. The process is to find the optimal solution for the combination of frequency and frequency resolution (determined by the window length) on the entire plane, so as to determine the frequency band position and interval of the periodic transient impact component. Different types of faults usually cause the system to have abnormal vibration or response in a specific frequency band. By analyzing the frequency band position and interval of these impact components, the characteristic signals associated with the specific fault mode can be identified, and it can also help determine the location of the fault in the system. In the noisy working environment of rotating machinery, this method can more accurately extract the fault characteristic frequency of the bearing and effectively diagnose the health status of the machinery.

[0079] Step 2.3: Calculate the short-time Fourier transform result of the frequency point corresponding to the peak value of the time domain envelope spectrum, which is the fault characteristic frequency.

[0080] Assume that the harmonic signal x1 is expressed as Then the result of its short-time Fourier transform is

[0081]

[0082] Where τ is the process variable and g(v) is the window function.

[0083] The peak value of the time-frequency envelope spectrum can resist the interference of harmonic signals, which can be deduced theoretically:

[0084] Substituting formula (6) into formula (5), we can get the following expression:

[0085]

[0086] This formula proves that the time-frequency envelope spectrum peak method proposed in this embodiment can resist the interference of harmonic signals.

[0087] Step 2.4: Envelope analysis is a very effective signal analysis technique for early bearing fault detection and diagnosis. The main challenge of envelope analysis is to find the most suitable frequency band for demodulation. To evaluate the ability of the time-frequency envelope spectrum peak to identify periodic pulse signals, the following formula is proposed:

[0088]

[0089] Where f0 represents the center frequency of the periodic pulse signal, and ER is the error rate between the center frequency of the periodic pulse signal identified by TFES and the inherent center frequency of the periodic pulse signal.

[0090] Formula (8) uses the error rate between the center frequency of the periodic pulse signal located by the time-frequency envelope spectrum peak and the inherent center frequency of the periodic pulse signal to verify the accuracy of periodic pulse signal recognition.

[0091] In this embodiment, after obtaining an obvious fault characteristic frequency, the specific fault type is diagnosed according to the fault identification method of the prior art, which will not be described in detail here.

[0092] In a specific implementation, this embodiment compares the time-frequency envelope spectrum peak analysis and the spectrum kurtosis analysis:

[0093] The constructed composite signal is represented in time domain as follows:

[0094]

[0095] Here, signal x1 is a 300Hz sinusoidal signal, and signal x2 is a periodic pulse signal with a center frequency of 200Hz. Signal x is the composite of signals x1 and x2. The sampling frequency is 1000Hz, and the sampling time is 2s. Figure 2 are the time domain waveform, short-time Fourier transform result, and spectral kurtosis result of the composite signal when the signal-to-noise ratio is 5dB, where: Figure 2 (a) is the time domain waveform, and (b) is the short-time Fourier transform result and spectral kurtosis result. Figure 3 are the time domain waveform, short-time Fourier transform result, and spectral kurtosis result of the composite signal when the signal-to-noise ratio is 0dB, where: Figure 3 (a) is the time domain waveform, and (b) is the short-time Fourier transform result and spectral kurtosis result. Figure 4 are the time domain waveform, short-time Fourier transform result, and spectral kurtosis result of the composite signal when the signal-to-noise ratio is -5dB, where: Figure 4 (a) is the time domain waveform, and (b) is the short-time Fourier transform result and spectral kurtosis result. Figure 2 、 3 As can be seen from Figure 4, as the noise intensity increases, the spectral kurtosis value decreases significantly. However, from the time-frequency representation of the short-time Fourier transform results, as the noise intensity increases, the periodicity of the periodic pulse signal is only slightly weakened and can still be identified. For a clearer effect, Figure 5 Draw separately Figure 2 、 3 ,4 is the time-frequency envelope of the short-time Fourier transform result at frequency f = 200 Hz, where Figure 5(a) is the time-frequency envelope of the composite signal at a frequency of f = 200Hz when the signal-to-noise ratio is 5dB, (b) is the time-frequency envelope of the composite signal at a frequency of f = 200Hz when the signal-to-noise ratio is 0dB, and (c) is the time-frequency envelope of the composite signal at a frequency of f = 200Hz when the signal-to-noise ratio is -5dB. Figure 5 It can be seen that as the signal-to-noise ratio decreases, the periodic characteristics at the center frequency of the periodic pulse signal fluctuate slightly, but it can still be identified. The proposed method is used to process the composite signal shown in formula (9). Figure 6 are the time domain waveform, short-time Fourier transform result, and time-frequency envelope spectrum peak result of the composite signal when the signal-to-noise ratio is 5dB, where: Figure 6 (a) is the time domain waveform and (b) is the short-time Fourier transform result and the time-frequency envelope spectrum peak result. Figure 7 are the time domain waveform, short-time Fourier transform result, and time-frequency envelope spectrum peak result of the composite signal when the signal-to-noise ratio is 0dB, where: Figure 7 (a) is the time domain waveform and (b) is the short-time Fourier transform result and the time-frequency envelope spectrum peak result. Figure 8 are the time domain waveform, short-time Fourier transform result, and time-frequency envelope spectrum peak result of the composite signal when the signal-to-noise ratio is -5dB, where: Figure 8 (a) is the time domain waveform and (b) is the short-time Fourier transform result and the time-frequency envelope spectrum peak result. Figure 6 、 7 The results of Figure 8 show that for the three different signal-to-noise ratios selected, the time-frequency envelope spectrum peak can very accurately identify the center frequency of the periodic pulse signal, and as the noise intensity increases, the size of the time-frequency envelope spectrum peak does not change much. In order to further evaluate the effectiveness of the time-frequency envelope spectrum peak technology, it verifies the accuracy of locating the center frequency of the periodic transient component in the multi-component signal under a larger signal-to-noise ratio range (from -10dB to 20dB). The results are as follows Figure 9 As shown in the figure, in the above cases, the error rate of the time-frequency envelope spectrum peak is within 1%. Therefore, it can be considered that the fast time-frequency envelope spectrum peak diagram has good noise immunity. It can accurately locate the center frequency of the periodic pulse signal in the presence of noise and other interference.

[0096] In a specific implementation, this embodiment compares the fast time-frequency envelope spectrum peak diagram and the fast spectrum kurtosis diagram:

[0097] The constructed composite signal is represented in time domain as follows:

[0098]

[0099] This embodiment uses several simulated signals to simulate the information that may exist in the acquisition. Signal A is a sine function with an amplitude of 0.3 and a frequency of 1000Hz, signal B is the sum of a single pulse with a center frequency of 2000Hz and a single pulse with a center frequency of 4000Hz, signal C is a periodic pulse with a center frequency of 3000Hz and a bandwidth of 150Hz, signal D is random noise with a signal-to-noise ratio of -7dB, and signal S(t) is a mixture of the above signals. The waveforms of the sine signal, sporadic pulse, periodic pulse signal, Gaussian white noise, mixed signal, and the short-time Fourier transform results of the mixed signal are shown in Figure 1. Figure 10 As shown. The sampling frequency of the simulation signal is 10000Hz, the sampling time is 1s, and the signal-to-noise ratio of the mixed signal S(t) is -7.8dB. The fast time-frequency envelope spectrum peak diagram method proposed in this embodiment is used to analyze S(t) and extract the periodic pulse signal C. The first level 0 in the fast time-frequency envelope spectrum peak diagram algorithm is the original signal to be analyzed. The second level 1 decomposes the signal into a binary tree structure, and the third level 1.6 decomposes the signal into a 1 / 3 tree structure; the fourth level 2 decomposes the signal into a binary tree structure, and the fifth level 2.6 decomposes the signal into a ternary tree structure. The remaining levels are similar, and the decomposition structure diagram is shown as follows. Figure 11 As shown. The original signal is used as the source to be classified and is named level 0. The frequency band is Δf∈[0,fs / 2], where fs is the sampling frequency. Divide level 0 into two parts, low frequency and high frequency, called level 1. The frequency bands of these two parts are [0,fs / 4] and [fs / 4,fs / 2] respectively. Divide level 0 into three parts, low frequency, medium frequency and high frequency, called level 1.6. The frequency bands of these three parts are [0,fs / 6], [fs / 6,fs / 3], and [fs / 3,fs / 2]. The second layer, level 1, decomposes the entire frequency band into 2 1 The third level 1.6 will decompose the entire frequency band into 2 1.6 Level K decomposes the entire frequency band into 2 k Therefore, level k divides level 0 into 2 k The boundary of the lowest frequency component corresponding to level k is [0,fs / 2 k+1 ]. Figure 12 is the result of fast time-frequency envelope spectrum peak diagram processing signal S(t), the time domain waveform of the filtered signal and the square envelope spectrum of the filtered signal, where, Figure 12(a) is the fast spectrum kurtosis diagram (where SKmax is the maximum spectrum kurtosis value, Bw is the bandwidth, and fc is the center frequency), (b) is the time domain waveform of the filtered signal, and (c) is the square envelope spectrum of the filtered signal. From the fast time-frequency envelope spectrum peak diagram, we can see that the frequency band with the largest time-frequency envelope spectrum peak is located at level 5, that is, the 20th frequency band from the left, with a frequency range of [2968.75Hz,3125Hz]. The center frequency of this band (f c ) is 3046Hz, bandwidth (Bw) is 156Hz, and the corresponding maximum value of the time-frequency envelope spectrum peak (TFESmax) is 6.8. Figure 12 In the time domain of the filtered signal, it can be seen that the periodic pulse interval t≈0.05s. Figure 12 The square envelope spectrum of the filtered signal also clearly shows the fault characteristic frequency of the periodic pulse (fe = 20 Hz) and its higher harmonics (nf e ).

[0100] As a comparison, this embodiment also uses a fast spectral kurtosis diagram based on spectral kurtosis to analyze the simulation signal S(t). Figure 13 .like Figure 13 The fast spectral kurtosis plot shown is misleading due to single-pulse interference. The frequency band with the maximum spectral kurtosis value (SKmax) is located at the frequency band with a center frequency of 3984 Hz at level 5. Periodic pulses and their characteristic frequencies are almost absent in the filtered time-domain signal and its envelope square spectrum. This example attempts to demodulate the second possible frequency band of the fast spectral kurtosis plot, namely [2968 Hz, 3125 Hz]. Figure 14 The time domain signal and its square envelope spectrum after filtering in this frequency band are shown. The characteristics of the periodic pulse can be identified from the figure. Figure 12 Compared with the decomposition results of the fast time-frequency envelope spectrum peak diagram in , it can be clearly seen that the periodic impact of the time domain signal after filtering of the fast time-frequency envelope spectrum peak diagram is more obvious, and the fault characteristic frequency (fe) in its square envelope spectrum is also more significant.

[0101] This embodiment experimentally verifies the bearing fault diagnosis method, and the specific process is as follows:

[0102] The vibration signal used in this experiment comes from the outer ring of a variable speed bearing. The vibration signal acquisition device includes a motor, an AC drive to control the speed, an accelerometer to collect vibration data, and an incremental encoder to measure the shaft speed. Based on the structural parameters of the bearing, the fault characteristic frequency of the bearing outer ring is known to be f o =3.57f r (f r The sampling frequency in the experiment was set to 200,000 Hz. The vibration data was from a faulty bearing outer race with a defect, and its rotation frequency increased from 13.3 Hz to 26.3 Hz within 10 seconds. Figure 15 The waveform of the vibration signal and its short-time Fourier transform result are plotted. The approximate distribution of the vibration signal frequency can be observed in the short-time Fourier transform result.

[0103] The signal is processed using the fast time-frequency envelope spectrum peak diagram method. Figure 16 As shown in (a), the maximum value of the time-frequency envelope spectrum peak appears in the first frequency band of level 5, the center frequency of this band is 1562.5Hz, and the bandwidth is 3125Hz. The time domain waveform of the signal filtered out of this band is as follows Figure 16 As shown in (b) in the figure. It can be seen that the waveform of the filtered component has less noise than the original vibration signal waveform. Since the experimental data was taken from a test bench under variable speed conditions, the bearing speed is constantly changing, that is, the fault characteristic frequency of the bearing outer ring is also changing. Therefore, the accuracy of the result can be judged by observing the short-time Fourier transform result of the square envelope spectrum of the filtered signal component, as shown in Figure 16 As shown in (c) in the figure. The fault characteristic frequency (f o ), twice the rotation frequency (2f r ) and its difference (f o -2f r ) can be clearly seen in the figure, and they are consistent with the results calculated with known parameters. This means that the fast time-frequency envelope spectrum peak diagram can be used to diagnose the fault of the outer ring of the variable speed bearing. Subsequently, the signal results are processed using the fast spectrum kurtosis diagram as shown in the figure. Figure 17 As shown in (a) in the figure. The component with the largest spectral kurtosis value in the fast spectral kurtosis diagram is located in the second frequency band of level 1. The center frequency of this frequency band is 75000Hz and the bandwidth is 50000Hz. The short-time Fourier transform results of the time domain waveform and square envelope spectrum of the filtered component are plotted on Figure 17 In (b) and (c). Figure 17 In (c), only the rotation frequency can be vaguely identified, but there is no frequency related to the fault characteristic frequency of the bearing outer race. Therefore, this proves that when demodulating the selected resonant frequency band, the fast time-frequency envelope spectrum peak diagram can obtain more and more effective periodic pulse information.

[0104] Example 2:

[0105] A second embodiment of the present invention provides a bearing fault diagnosis system based on time-frequency envelope spectrum peak analysis, including:

[0106] a signal acquisition module configured to acquire a time-frequency spectrum of a vibration signal to be diagnosed, preprocess the time-frequency spectrum, and obtain a preprocessed envelope;

[0107] A fast time-frequency envelope spectrum peak diagram module is configured to construct a fast time-frequency envelope spectrum peak diagram and decompose the preprocessed envelope using the fast time-frequency envelope spectrum peak diagram;

[0108] A fault diagnosis module is configured to extract a fault characteristic frequency in the envelope and diagnose the bearing fault type according to the fault characteristic frequency;

[0109] Among them, the fault diagnosis module includes an envelope analysis module, which is configured to extract the fault characteristic frequency in the fast time-frequency envelope spectrum peak diagram. The specific steps include: extracting the frequency point with the most prominent pulse characteristics in the fast time-frequency envelope spectrum peak diagram to obtain the time domain envelope spectrum peak; calculating the short-time Fourier transform result of the frequency point corresponding to the time domain envelope spectrum peak, which is the fault characteristic frequency.

[0110] Example 3:

[0111] A third embodiment of the present invention provides a medium having a program stored thereon. When the program is executed by a processor, the steps of the bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis as described in the first embodiment of the present invention are implemented.

[0112] Example 4:

[0113] Embodiment 4 of the present invention provides a device, including a memory, a processor, and a program stored in the memory and runnable on the processor. When the processor executes the program, the steps in the bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis as described in Embodiment 1 of the present invention are implemented.

[0114] The steps involved in the above embodiments 2, 3, and 4 correspond to those in the method embodiment 1. For detailed implementation, please refer to the relevant description of embodiment 1. The term "computer-readable storage medium" should be understood to mean a single medium or multiple media that includes one or more instruction sets; it should also be understood to include any medium that can store, encode, or carry an instruction set for execution by a processor and cause the processor to perform any method of the present invention.

[0115] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0116] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis, characterized in that: The following steps are involved: Obtaining the time-frequency spectrum of the vibration signal to be diagnosed, preprocessing the time-frequency spectrum, and obtaining a preprocessed envelope; Construct a fast time-frequency envelope spectrum peak map, and use the fast time-frequency envelope spectrum peak map to decompose the preprocessed envelope; The first level 0 of the fast time-frequency envelope spectrum peak diagram is the original signal to be analyzed, the second level 1 decomposes the signal into a binary tree structure, the third level 1.6 decomposes the signal into a 1 / 3 tree structure; the fourth level 2 decomposes the signal into a binary tree structure, the fifth level 2.6 decomposes the signal into a ternary tree structure, and the rest of the levels are similar; the original signal is the source to be classified, named level 0, and the frequency band is ,in is the sampling frequency; level 0 is divided into low frequency and high frequency parts, called level 1, and the frequency bands are and ; Divide level 0 into three parts: low frequency, medium frequency and high frequency, called level 1.6, the frequency band is 、 、 ; The second level 1 decomposes the entire frequency band into 2 1 The third level 1.6 decomposes the entire frequency band into 2 1.6 Level k decomposes the entire frequency band into 2 k the law of parts; Level k splits level 0 into two k Part, the boundary of the lowest frequency component corresponding to level k is ; Extract the fault characteristic frequency from the fast time-frequency envelope spectrum peak diagram and diagnose the bearing fault type through the fault characteristic frequency; The specific steps of extracting the fault characteristic frequency from the fast time-frequency envelope spectrum peak diagram include: extracting the frequency point with the most prominent pulse characteristics in the fast time-frequency envelope spectrum peak diagram to obtain the time domain envelope spectrum peak; calculating the short-time Fourier transform result of the frequency point corresponding to the time domain envelope spectrum peak, which is the fault characteristic frequency; The specific process of extracting the frequency point with the most prominent pulse characteristics in the envelope and obtaining the peak value of the time domain envelope spectrum is as follows: Calculate the envelope spectrum value of each frequency point of short-time Fourier transform; The maximum value is used to represent the frequency point where the pulse feature in the fault signal is most prominent, that is, the peak value of the time domain envelope spectrum; The calculation formula for the time domain envelope spectrum peak is: in, is the envelope spectrum value of the frequency point, is the frequency point, is the window length of the window function, r is the current window length, is the short-time Fourier transform function, Indicates the short-time Fourier transform result at the frequency point The average value of is the discrete frequency, F s is the sampling frequency, and i is the time point corresponding to the current window length.

2. The bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis according to claim 1 is characterized in that: The time-frequency spectrum is preprocessed to remove the mean value of each frequency point in the signal time-frequency spectrum.

3. The bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis according to claim 1 is characterized in that: Short-time Fourier transform function The calculation formula is: ， in, is the discrete frequency, is the window function, r is the current window length, is the window length of the window function, F s is the sampling frequency, i is the time point corresponding to the current window length, x is the vibration signal, and L is the time shift between consecutive windows.

4. The bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis according to claim 1, characterized in that: The accuracy of periodic pulse signal recognition is verified by using the error rate between the center frequency of the periodic pulse signal located by the time-frequency envelope spectrum peak and the inherent center frequency of the periodic pulse signal.

5. A bearing fault diagnosis system based on time-frequency envelope spectrum peak analysis, characterized in that: include: a signal acquisition module configured to acquire a time-frequency spectrum of a vibration signal to be diagnosed, preprocess the time-frequency spectrum, and obtain a preprocessed envelope; A fast time-frequency envelope spectrum peak diagram module is configured to construct a fast time-frequency envelope spectrum peak diagram and decompose the preprocessed envelope using the fast time-frequency envelope spectrum peak diagram; The first level 0 of the fast time-frequency envelope spectrum peak diagram is the original signal to be analyzed, the second level 1 decomposes the signal into a binary tree structure, the third level 1.6 decomposes the signal into a 1 / 3 tree structure; the fourth level 2 decomposes the signal into a binary tree structure, the fifth level 2.6 decomposes the signal into a ternary tree structure, and the rest of the levels are similar; the original signal is the source to be classified, named level 0, and the frequency band is ,in is the sampling frequency; level 0 is divided into low frequency and high frequency parts, called level 1, and the frequency bands are and ; Divide level 0 into three parts: low frequency, medium frequency and high frequency, called level 1.6, the frequency band is 、 、 ; The second level 1 decomposes the entire frequency band into 2 1 The third level 1.6 decomposes the entire frequency band into 2 1.6 Level k decomposes the entire frequency band into 2 k the law of parts; Level k splits level 0 into two k Part, the boundary of the lowest frequency component corresponding to level k is ; The fault diagnosis module is configured to extract the fault characteristic frequency in the envelope and diagnose the bearing fault type through the fault characteristic frequency. The fault diagnosis module includes an envelope analysis module, which is configured to extract the fault characteristic frequency in the fast time-frequency envelope spectrum peak diagram. The specific steps include: extracting the frequency point with the most prominent pulse feature in the fast time-frequency envelope spectrum peak diagram to obtain the time domain envelope spectrum peak; calculating the short-time Fourier transform result of the frequency point corresponding to the time domain envelope spectrum peak, which is the fault characteristic frequency; The specific process of extracting the frequency point with the most prominent pulse characteristics in the envelope and obtaining the peak value of the time domain envelope spectrum is as follows: Calculate the envelope spectrum value of each frequency point of short-time Fourier transform; The maximum value is used to represent the frequency point where the pulse feature in the fault signal is most prominent, that is, the peak value of the time domain envelope spectrum; The calculation formula for the time domain envelope spectrum peak is: in, is the envelope spectrum value of the frequency point, is the frequency point, is the window length of the window function, r is the current window length, is the short-time Fourier transform function, Indicates the short-time Fourier transform result at the frequency point The average value of is the discrete frequency, F s is the sampling frequency, and i is the time point corresponding to the current window length.

6. A computer-readable storage medium, characterized in that A plurality of instructions are stored therein, and the instructions are suitable for being loaded by a processor of a terminal device and executing the bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis according to any one of claims 1 to 4.

7. A terminal device, characterized in that: It includes a processor and a computer-readable storage medium, the processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded by the processor and executing the bearing fault diagnosis method based on time-frequency envelope spectrum peak analysis described in any one of claims 1-4.

Images