Bearing fault diagnosis method based on generalized time-frequency extraction transformation

Through the generalized time-frequency extraction and transformation method, the problem of insufficient resolution in traditional time-frequency analysis in bearing fault diagnosis is solved, and more accurate fault impact feature capture and noise suppression is achieved, which improves the accuracy of bearing fault diagnosis.

CN119004053AInactive Publication Date: 2024-11-22CHINA UNIV OF MINING & TECH
View PDF 0 Cites 5 Cited by

Patent Information

Application Number
CN202411038219.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional time-frequency analysis methods have insufficient resolution when dealing with complex non-stationary bearing vibration signals, and cannot accurately capture fault impact characteristics, especially in group delay feature estimation and time-frequency energy distribution.

Method used

Using a method based on generalized time-frequency extraction transformation, a generalized time-frequency matrix is ​​constructed through short-time Fourier transforms of different Gaussian window functions, the generalized group delay is calculated and generalized time-frequency extraction transformation is performed, and the energy of combined time-frequency distribution is extracted, so as to accurately Describe the time-frequency distribution of fault impact characteristics and judge the bearing fault type.

Benefits of technology

It realizes more accurate fault diagnosis, improves time-frequency aggregation, reduces noise interference, and enhances the reliability and accuracy of diagnostic results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004053A_ABST
    Figure CN119004053A_ABST
Patent Text Reader

Abstract

The invention discloses a bearing fault diagnosis method based on generalized time-frequency extraction transformation. The method specifically comprises the following steps: 1, acquiring a vibration signal of a fault bearing; 2, performing short-time Fourier transform of different Gaussian window functions on the obtained fault bearing vibration signals to obtain time-frequency distribution, taking the time-frequency distribution as elements of a generalized time-frequency matrix, constructing a generalized time-frequency matrix B, calculating generalized group delay estimation according to the B, performing a generalized time-frequency extraction transform algorithm based on the generalized group delay estimation, and obtaining a generalized time-frequency extraction transform algorithm based on the generalized group delay estimation; extracting energy of the combined time-frequency distribution at the generalized group delay estimation position by adopting a generalized time-frequency extraction transformation algorithm; 3, extracting the time interval of the impact characteristics in the combined time-frequency distribution result of the steps; and 4, judging the fault type of the bearing based on the time interval. The method has higher time-frequency resolution, and transient impact characteristics caused by bearing faults can be captured more accurately. The method also has good anti-noise performance, and can effectively extract fault signals in a complex noise environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of machine failure, and in particular relates to a bearing failure diagnosis method based on generalized time-frequency extraction transformation. Background Art

[0002] In industrial production, bearings are core components of rotating machinery, and their operating status is directly related to the stability and safety of the entire mechanical system. However, due to the influence of various factors such as long-term operation, load changes, and poor lubrication, bearings are prone to wear, cracks, looseness and other faults. If these faults are not discovered and handled in time, they may cause equipment shutdown, production interruption and even safety accidents. Therefore, it is particularly important to perform timely and accurate fault diagnosis on bearings. As one of the important parameters reflecting the operating status of bearings, vibration signals contain rich fault information. By monitoring and analyzing the vibration signals of bearings, early warning and diagnosis of bearing faults can be achieved, thereby avoiding major losses.

[0003] Traditional bearing fault diagnosis methods are mainly based on time domain analysis or frequency domain analysis. These methods perform well when processing stationary signals, but when faced with non-stationary, nonlinear vibration signals generated by bearing faults, it is often difficult to accurately capture fault characteristics. As an important development in the field of signal processing, the time-frequency analysis method can describe signals in both the time domain and the frequency domain, effectively revealing the time-varying characteristics of non-stationary signals. In bearing fault diagnosis, the time-frequency analysis method can more accurately capture the transient impact characteristics caused by the fault, such as impact frequency, impact energy, etc., thereby improving the accuracy and reliability of fault diagnosis. In addition, the time-frequency analysis method can also perform a detailed analysis of the local characteristics of the signal, which helps to identify early weak fault signals and provides strong support for predictive maintenance of bearings.

[0004] Although time-frequency analysis methods have performed well in bearing fault diagnosis, traditional time-frequency analysis methods such as short-time Fourier transform and wavelet transform still have problems such as insufficient resolution and cross-term interference when processing complex non-stationary signals. When processing transient signals such as bearing fault impacts, it is impossible to accurately estimate the group delay characteristics of the signal and thus it is impossible to obtain the time-frequency energy distribution aggregated along the time direction. Summary of the invention

[0005] Purpose of the invention: In order to solve the problems existing in the above-mentioned prior art, the present invention provides a bearing fault diagnosis method based on generalized time-frequency extraction transformation.

[0006] Technical solution: The present invention provides a bearing fault diagnosis method based on generalized time-frequency extraction transformation, comprising the following steps:

[0007] Step 1: Obtain the vibration signal of the faulty bearing;

[0008] Step 2: Apply short-time Fourier transform of different Gaussian window functions to the faulty bearing vibration signal obtained in step 1 to obtain the time-frequency distribution. The time-frequency distribution is used as the element of the generalized time-frequency matrix to construct the generalized time-frequency matrix B. The generalized group delay estimate is calculated based on B. Based on the generalized group delay estimation, a generalized time-frequency extraction transform algorithm is used to extract the energy of the combined time-frequency distribution at the generalized group delay estimation.

[0009] Step 3: extracting the time interval of the impact feature from the combined time-frequency distribution result calculated in step 2;

[0010] Step 4: Determine the bearing fault type based on the time interval.

[0011] Furthermore, the expression of the elements of the generalized time-frequency matrix in step 2 is:

[0012]

[0013] Where h(t) is the time domain Gaussian window function in different transformation forms, h(t) = (-j) k-1 g (k-1) (t), j is the imaginary unit, g(t) is the Gaussian window function, g (k-1) (t) represents the k-1 order differential of the function g(t), x(τ) represents the vibration signal of the faulty bearing, and * represents taking the conjugate; , where t represents the time variable, ω represents the frequency variable, N is the order of the generalized time-frequency matrix B, k = 1, 2, ..., N, and the elements of the generalized time-frequency matrix are combined into the generalized time-frequency matrix B:

[0014]

[0015] Furthermore, the expression of the generalized group delay estimate in step 2 is As shown below:

[0016]

[0017] Among them, σ is the Gaussian window function parameter, Indicates taking the imaginary part of the function, B 1,k is the algebraic cofactor of the generalized time-frequency matrix B.

[0018] Furthermore, the expression of the generalized time-frequency extraction transformation algorithm in step 2 is:

[0019]

[0020] Among them, δ(·) is the Dirac function, D is the operator matrix, and D 1,k Represents the algebraic co-factor of the operator matrix D. The expression of D is as follows:

[0021]

[0022] Wherein, g(0) is the Gaussian window function at t=0.

[0023] Furthermore, when a generalized time-frequency extraction transform algorithm is used to extract the energy of the combined time-frequency distribution at the generalized group delay estimate, the retention time is equal to the time-frequency energy at the generalized group delay estimate, and the time-frequency energy at other positions is set to zero.

[0024] Further, the step 4 is specifically:

[0025] Step 4.1: Calculate the fault characteristic frequencies of the bearing under different fault types according to the bearing structural parameters;

[0026] Step 4.2: Calculate the time interval Δt between two adjacent pulse components under different fault types according to the fault characteristic frequency in step 4.1;

[0027] Step 4.3: Compare the time interval extracted in step 3 with Δt in step 4.2 to determine the fault type of the bearing.

[0028] Furthermore, the following formula is used in step 4.1 to calculate the fault characteristic frequency of the bearing under different fault types:

[0029]

[0030] Among them, f BPFO is the characteristic frequency of outer race fault, f BPFI is the characteristic frequency of the inner race fault, f BSF is the characteristic frequency of rolling element failure, f FTF is the characteristic frequency of cage failure, N b and d represent the number and diameter of rolling elements respectively, D and They represent the pitch diameter and contact angle of the bearing, respectively, r The main axis rotation frequency.

[0031] Beneficial effects: The generalized time-frequency extraction transformation proposed in the present invention can directly calculate the explicit expression of the transient signal high-order group delay estimation, thereby accurately describing the time-frequency distribution of the fault impact characteristics, and has better time-frequency aggregation. The time interval of the bearing fault impact can be clearly and intuitively obtained from the time-frequency distribution of the generalized time-frequency extraction transformation, and the transient impact characteristics caused by the bearing fault can be more accurately captured, so that the bearing fault type can be more accurately determined; at the same time, the precise reconstruction characteristics of the generalized time-frequency extraction transformation can accurately reconstruct the time domain signal of the fault in the impact characteristics, and can realize the time-frequency filtering of the signal, reduce noise interference, verify the diagnosis results, and improve the accuracy of bearing fault diagnosis. The present invention provides a more reliable technical means for bearing fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of the present invention;

[0033] Figure 2 The time domain waveform diagram of the simulation signal of the bearing outer ring fault, where (a) is the time domain waveform diagram of the fault signal without noise, and (b) is the time domain waveform diagram of the fault signal with a signal-to-noise ratio of -5dB;

[0034] Figure 3 The time-frequency distribution results of different time-frequency analysis methods for the simulation signal, among which (a) is the time-frequency distribution result of the short-time Fourier transform, (b) is the time-frequency distribution result of the time-rearrangement synchronous compression transform, (c) is the time-frequency distribution result of the transient extraction transform, and (d) is the time-frequency distribution result of the generalized time-frequency extraction transform;

[0035] Figure 4 Time domain waveform diagram of the original signal, reconstructed signal and signal time-frequency slices of the simulation signal;

[0036] Figure 5 is the envelope spectrum of the simulated signal, where (a) is the envelope spectrum of the original signal and (b) is the envelope spectrum of the reconstructed signal;

[0037] Figure 6 Figure 1 is a diagram of a faulty bearing, where (a) is a diagram of a bearing with an outer ring fault, and (b) is a partial magnified diagram of an outer ring fault;

[0038] Figure 7 This is the time domain waveform of the bearing fault vibration signal;

[0039] Figure 8 The time-frequency distribution results of different time-frequency analysis methods for actual bearing fault vibration signals; (a) is the time-frequency distribution result of short-time Fourier transform, (b) is the time-frequency distribution result of time-rearranged synchronous compression transform, (c) is the time-frequency distribution result of transient extraction transform, and (d) is the time-frequency distribution result of generalized time-frequency extraction transform;

[0040] Fig. 9 It is the time domain waveform diagram of the original signal, reconstructed signal and signal time-frequency slice of the actual bearing fault vibration signal;

[0041] Fig.10 It is the envelope spectrum of the actual bearing fault vibration signal; (a) is the envelope spectrum of the original signal, and (b) is the envelope spectrum of the reconstructed signal. DETAILED DESCRIPTION

[0042] 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.

[0043] When the faulty bearing is running, its local defects contact the mating surface and generate transient impact. This impact appears in the vibration signal in the form of a transient pulse. When the bearing speed remains unchanged, the transient pulses in the vibration signal will appear periodically, and the frequency of its periodic appearance is the traditionally defined characteristic frequency of the bearing fault. By detecting the characteristic frequency of the bearing fault, the type of bearing fault can be determined. The vibration signal of the faulty bearing will show periodic transient pulses and have the characteristics of an amplitude modulated signal. Around the bearing resonance frequency, the rapidly changing amplitude modulation characteristics of the vibration signal will cause the appearance of many complex sidebands, so it is very difficult to extract the fault characteristic frequency from the vibration signal spectrum. In fact, each pulse component is reflected in the time-frequency diagram in the form of a broadband. Based on this feature, the present invention will accurately identify the time interval of the fault impact characteristics in the time-frequency distribution through a high-precision time-frequency analysis method, thereby performing bearing fault diagnosis.

[0044] like Figure 1 As shown, the present invention provides a bearing fault diagnosis method based on generalized time-frequency extraction transformation, comprising the following steps:

[0045] The first step is to obtain the vibration signal of the faulty bearing.

[0046] The detection method based on vibration signals has attracted wide attention due to its non-invasiveness, real-time performance and high accuracy. As a direct reflection of the operating status of the bearing, the vibration signal contains rich fault information. The present invention realizes effective monitoring and diagnosis of bearing faults by collecting the vibration response signal of the faulty bearing.

[0047] The second step is to calculate the characteristic frequency of bearing fault.

[0048] First, according to the bearing structural parameters, the fault characteristic frequency of the bearing under different fault forms is calculated. The calculation formula is as follows:

[0049]

[0050] In the formula, f BPFOis the characteristic frequency of outer race fault, f BPFI is the characteristic frequency of the inner race fault, f BSF is the characteristic frequency of rolling element failure, f FTF is the characteristic frequency of cage failure, N b and d represent the number and diameter of rolling elements respectively, D and They represent the pitch diameter and contact angle of the bearing, respectively, r The main axis rotation frequency.

[0051] Then, the time interval between two adjacent pulse components is calculated based on the characteristic frequency of the bearing fault:

[0052]

[0053] Where f is the characteristic frequency of bearing fault, including: f BPFO , f BPFI , f BSF and f FTF .

[0054] The third step is time-frequency feature analysis and fault diagnosis.

[0055] The present invention adopts a generalized time-frequency extraction transformation to calculate the time-frequency distribution of the signal to be analyzed. For the fast-varying transient signal of bearing fault impact, the traditional time-frequency analysis method cannot obtain the ideal time-frequency distribution effect, and cannot directly obtain the information that can characterize the impact energy characteristics from the time-frequency distribution. The time-frequency distribution result of the generalized time-frequency extraction transformation has higher time-frequency aggregation, which is suitable for processing transient signals such as fault impact, and can more accurately obtain the time interval of the impact characteristics.

[0056] First, the generalized time-frequency extraction transform time-frequency distribution of the analysis signal is calculated. A generalized time-frequency matrix for calculating the generalized group delay estimation is constructed. The elements of the generalized time-frequency matrix are the time-frequency distributions obtained by short-time Fourier transform based on different Gaussian window functions. The calculation formula is as follows:

[0057]

[0058] Where h(t) is the time domain Gaussian window function in different transformation forms, h(t) = (-j) k-1 g (k-1) (t), j is the imaginary unit, g(t) is the Gaussian window function, g (k-1) (t) represents the k-1 order differential of the function g(t), * represents taking the conjugate; where t represents the time variable, ω represents the frequency variable, x(τ) is the time domain signal, that is, the vibration signal of the faulty bearing, and the N-order generalized time-frequency matrix B can be composed of 2N-1 elements, that is, Then at this time k = 1, 2,…, N.

[0059] The generalized time-frequency matrix is ​​constructed as follows:

[0060]

[0061] In the formula, the matrix B is a symmetric matrix, that is, B = B T The generalized group delay estimate display expression can be calculated based on the above generalized time-frequency matrix, and the calculation formula is as follows:

[0062]

[0063] Where σ is the Gaussian window function parameter, Indicates taking the imaginary part of the function, B 1,k is the algebraic cofactor of the generalized time-frequency matrix B.

[0064] The basic principle of the generalized time-frequency extraction transform is to extract the energy of the combined time-frequency distribution at the generalized group delay estimate through the Dirac function extraction operator, thereby obtaining a sparser time-frequency distribution result. Therefore, the generalized time-frequency extraction transform algorithm is as follows:

[0065]

[0066] In the formula, is the generalized group delay estimate in formula (5), δ(·) is the Dirac function, and D is the operator matrix, which is composed of the high-order derivatives of the Gaussian window function and is composed as follows:

[0067]

[0068] In the formula, is a Gaussian window function, and the matrix D is a symmetric matrix, that is, D = D T , whose N-order form consists of 2N-1 elements. Therefore, in the combined time-frequency distribution constructed by the operator matrix, only the time-frequency energy at the time equal to the generalized group delay is retained, and the time-frequency energy at other positions is set to zero, thereby obtaining a more sparse and concentrated time-frequency energy distribution.

[0069] Then, the time interval of the impact feature in the transformed time-frequency distribution result is extracted by generalized time-frequency, and compared with the theoretical time interval calculated in step 2. Compare and determine the type of bearing fault.

[0070] Step 4: Verify the fault diagnosis results.

[0071] The present invention uses the inverse transform of the generalized time-frequency extraction transform to reconstruct the signal, filter out the noise interference in the signal from the time-frequency domain, and reconstruct the time-domain waveform of the characteristic component. The envelope spectrum of the reconstructed signal is calculated, and the fault characteristic frequency is identified from the envelope spectrum to verify the accuracy of the proposed method.

[0072] First, the generalized transient extraction inverse transform method is used to reconstruct the fault characteristic components. The reconstruction algorithm of the generalized time-frequency extraction transform is as follows:

[0073]

[0074] Then, calculate the envelope spectrum of the reconstructed signal. The upper envelope of the vibration signal can be obtained by Hilbert transform:

[0075]

[0076] In the formula, x(t) is the original signal, and Hb(·) represents the Hilbert transform. The obtained signal is the upper envelope of the original signal, that is, x envolope (t)≥0. The envelope spectrum can be obtained by calculating the frequency spectrum of the envelope signal, and the accuracy of the fault diagnosis method of the present invention can be verified by the fault characteristic frequency in the envelope spectrum.

[0077] Embodiment 1 of the present invention

[0078] The first step is to obtain the vibration signal of the faulty bearing.

[0079] This embodiment constructs a bearing fault signal based on the vibration response principle of a defective bearing. Each time the bearing rolling element passes through a defective position, a transient impact is generated, including a large number of transient pulse components, as shown in the following formula:

[0080]

[0081] In the formula, β and f reson are the bearing system damping coefficient and resonant frequency, N and t k Respectively represent the number and time of transient impact, ε(t) represents the unit step signal, A(t) is the amplitude function, and its expression is as follows:

[0082]

[0083] In the formula, f r is the main shaft rotation frequency, u is the amplitude modulation coefficient, u=0 represents the bearing outer ring fault, and u>0 represents the bearing inner ring fault. In this embodiment, a simulation signal of the bearing outer ring fault is constructed, and the bearing outer ring fault parameters are shown in Table 1.

[0084] Table 1

[0085] <![CDATA[f r ]]> β <![CDATA[f reson ]]> <![CDATA[f BPFO ]]> 45Hz 1000 5000Hz 135Hz

[0086] Assuming that the working condition is the rotor moving at a constant speed, the moment of transient impact due to outer ring fault can be determined as follows:

[0087]

[0088] In the formula, f BPFO is the characteristic frequency of the outer ring bearing fault in Table 1. Substituting the data in Table 1 into the above formula, the time domain waveform of the constructed signal is obtained as follows: Figure 2 As shown in (a), the environment and working condition noise are simulated. -5dB Gaussian white noise is added to the simulation signal. The time domain waveform of the noisy signal is as follows: Figure 2 As shown in (b) in the figure, it is difficult to clearly distinguish the fault features contained in the signal from the noise signal.

[0089] The second step is to calculate the fault characteristic frequency.

[0090] According to the bearing structural parameters, the fault characteristic frequency of the bearing under different fault forms is calculated. The calculation formula is as follows:

[0091]

[0092] In the formula, f BPFO is the characteristic frequency of outer race fault, f BPFI is the characteristic frequency of the inner race fault, f BSF is the characteristic frequency of rolling element failure, f FTF is the characteristic frequency of cage failure, N b and d represent the number and diameter of rolling elements respectively, D and They represent the pitch circle diameter and contact angle of the bearing respectively.

[0093] The simulation signal in the previous step is the outer ring fault of the rolling bearing. The time interval between two adjacent pulse components is calculated based on the characteristic frequency of the outer ring fault of the bearing:

[0094]

[0095] According to the characteristic frequency of the bearing outer ring fault, the interval between adjacent impacts is calculated as:

[0096] The third step is time-frequency feature analysis and fault diagnosis.

[0097] Different time-frequency analysis methods are used to deal with Figure 2 The time-frequency distribution of the noise signal in (b) is as follows: Figure 3 As shown. Figure 3 As can be seen from (a) in the figure, the time-frequency distribution result of the short-time Fourier transform has the worst resolution and the most serious energy diffusion, and can only determine the number of transient pulses. Figure 3 (b) in Figure 2 shows that the time-rearranged synchronous compression transform does improve the clarity of the time-frequency distribution results, but the energy is still relatively dispersed. The transient extraction transform is relatively sparse, which greatly improves the aggregation of the time-frequency distribution. Figure 3 As shown in (c) in the figure. Figure 3As can be seen from (d) in the figure, the energy of the time-frequency distribution result of the generalized time-frequency extraction transform is relatively concentrated, and the time-frequency resolution is relatively improved. At the same time, the energy of the extraction transform is more concentrated, which is more conducive to the extraction and separation of transient pulses and fault judgment. At the same time, in the time-frequency distribution result, the time interval between the pulse components (i.e., 1 / f BPFO =7.4ms) to accurately identify the fault type. Table 2 below gives the Rayleigh entropy values ​​of the above time-frequency analysis results. Rayleigh entropy has always been used as a quantitative indicator to measure the aggregation of time-frequency distribution. The smaller the value, the more concentrated the time-frequency distribution. From the corresponding Rayleigh entropy value, it can be judged that the generalized time-frequency extraction transform has the best time-frequency distribution.

[0098] Table 2

[0099] Time-frequency analysis method Rayleigh Entropy Short-time Fourier transform 8.3765 Synchronous compression transformation in time direction 5.3201 Transient Extraction Transform 3.3284 Generalized Time-Frequency Extraction Transform 2.7624

[0100] Step 4: Verify the fault diagnosis results.

[0101] The generalized time-frequency extraction transform does not need to select the reconstruction energy bandwidth parameter when reconstructing the signal. Therefore, its reversibility can be used to perform time-frequency filtering on the signal to eliminate the noise in the signal. The simulation signal and its time-frequency distribution reconstructed signal are as follows: Figure 4 As shown. By comparison, it can be found that the periodic transient pulse components in the reconstructed signal are more obvious than those in the original signal. Figure 3 The normalized time-frequency slice at 5000 Hz of the time-frequency distribution in (d) is used for comparison. It can be found that the time-frequency distribution of the proposed time-frequency analysis method can accurately characterize the occurrence time of each impact, thereby verifying the effectiveness of using the time-frequency analysis method for fault diagnosis.

[0102] In order to fully illustrate the effectiveness of the time-frequency analysis method in bearing failure, Figure 5 The envelope spectra of the simulated signal and the reconstructed signal are given in . Figure 5 It can be observed from (a) that only weak fault characteristic frequencies are found in the envelope spectrum of the simulation signal. Figure 5 In (b), the characteristic frequency of the outer race fault and its harmonics in the reconstructed signal envelope spectrum are very obvious, proving that the pulses caused by the fault appear periodically. This shows that the proposed time-frequency analysis method can effectively improve the bearing fault diagnosis performance under strong noise conditions by utilizing its reconstruction characteristics.

[0103] Embodiment 2 of the present invention

[0104] The first step is to obtain the vibration signal of the faulty bearing.

[0105] The effectiveness of the algorithm proposed in this invention is verified by collecting bearing fault signals through real fault bearing experiments. The fault of the rolling bearing is the outer ring fault of the wire cutting, such as Figure 6As shown in the figure. During the test, when the rolling element passes through the outer ring defect, a periodic pulse vibration signal is generated. Figure 7 The time domain waveform of the vibration signal shown in the figure shows that the signal contains a large number of background noise harmonics, and it is impossible to distinguish whether the periodic component is a modulation signal or a pulse signal. Therefore, it is difficult to diagnose the bearing fault from the time domain waveform.

[0106] The second step is to calculate the characteristic frequency of bearing fault.

[0107] The structural parameters of the rolling bearing used in the experiment are shown in Table 3. The impact time interval can be calculated based on the bearing structural parameters and the characteristic fault characteristic frequency calculation formula.

[0108] Table 3

[0109] parameter Pitch diameter Rolling element diameter Number of rolling elements Contact Angle Numeric 20mm 4.76mm 8 0°

[0110] For bearing outer ring faults, when the rolling element passes through the fault point, an impact pulse will be excited. In this experiment, the shaft rotation frequency is kept at 45Hz; therefore, the characteristic frequency of the rolling element passing through the outer ring is calculated to be 137.16Hz through the bearing fault characteristic frequency calculation formula and the bearing parameters in Table 3, and the impact time interval is 1 / f BPFO =7.3ms.

[0111] The third step is time-frequency feature analysis and fault diagnosis.

[0112] The vibration signal was analyzed using time-frequency analysis methods including short-time Fourier transform, time-rearranged synchronous compression transform, transient extraction transform and generalized time-frequency extraction transform. The time-frequency distribution results of different time-frequency analysis methods are shown in the figure. Figure 8 In order to more easily observe the time-frequency distribution of the impulse component, the time-frequency diagram only shows the frequency band containing the periodic pulse. The pulse component frequency band can be determined by the spectral kurtosis of the impulse signal. Figure 8 As can be seen from (a) in the figure, the time-frequency distribution of the short-time Fourier transform is the worst, the energy diffusion is the most serious, and it is even impossible to determine the number of impact components. Figure 8 (b) in Figure 1 shows that the time-rearranged synchronous compression transform does improve the clarity of the time-frequency distribution results. It can be roughly determined that the signal contains 6 impulse components, but the fault information is limited. The transient extraction transform is relatively sparse, which greatly improves the aggregation of the time-frequency distribution. Figure 8 As shown in (c) in 8. In (d) in 8, the energy of the time-frequency distribution result of the generalized time-frequency extraction transformation is more concentrated, and the energy of the extraction transformation is more concentrated, which is more conducive to the extraction and separation of the impulse component and fault judgment. At the same time, from the time-frequency distribution results, the fault type can be accurately identified according to the time interval between the pulse components. Table 4 shows the Rayleigh entropy value of the above time-frequency analysis results.

[0113] Table 4

[0114] Time-frequency analysis method Rayleigh Entropy Short-time Fourier transform 6.3590 Synchronous compression transformation in time direction 3.0226 Transient Extraction Transform 0.9409 Generalized Time-Frequency Extraction Transform 0.1644

[0115] Since the generalized time-frequency extraction transform has excellent reconstruction characteristics and does not require reconstruction of energy bandwidth parameters, its reversibility can be used to perform time-frequency filtering on the signal to eliminate noise in the vibration signal. The original vibration signal and its time-frequency distribution reconstructed signal are as follows: Fig. 9 As shown in Figure 2, by comparison, it can be found that the impact pulse component in the reconstructed signal is more obvious than the original signal. It is easier to observe the periodic pulse in the reconstructed signal. Figure 8 The normalized time-frequency slice at 9665 Hz of the time-frequency distribution in (d) is compared with the reconstructed signal. It can be found that the time-frequency distribution of the proposed time-frequency analysis method can accurately characterize the occurrence time of each impact, thereby verifying the effectiveness of fault diagnosis through time-frequency analysis method.

[0116] To fully verify that the reconstructed signal contains more fault information than the original signal, Fig.10 The envelope spectra of the original signal and the reconstructed signal are given in . Fig.10 In (a), it can be observed that the outer race fault characteristic frequency and its harmonics 10 are found in the envelope spectrum of the reconstructed signal. In (b), only weak fault characteristic frequencies are found in the envelope spectrum of the original signal. In summary, the proposed time-frequency analysis method can effectively improve the fault diagnosis performance under strong noise conditions by utilizing its reconstruction characteristics.

[0117] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

Claims

1. A bearing fault diagnosis method based on generalized time-frequency extraction transform, characterized in that: The steps include: Step 1: Obtain the vibration signal of the faulty bearing; Step 2: Apply short-time Fourier transform of different Gaussian window functions to the faulty bearing vibration signal obtained in step 1 to obtain the time-frequency distribution. The time-frequency distribution is used as the element of the generalized time-frequency matrix to construct the generalized time-frequency matrix B. The generalized group delay estimate is calculated based on B. Based on the generalized group delay estimation, a generalized time-frequency extraction transform algorithm is used to extract the energy of the combined time-frequency distribution at the generalized group delay estimation. Step 3: extracting the time interval of the impact feature from the combined time-frequency distribution result calculated in step 2; Step 4: Determine the bearing fault type based on the time interval.

2. A bearing fault diagnosis method based on generalized time-frequency extraction transformation according to claim 1, characterized in that: The expression of the elements of the generalized time-frequency matrix in step 2 is: Where h(t) is the time domain Gaussian window function in different transformation forms, h(t) = (-j) k-1 g (k-1) (t), j is the imaginary unit, g(t) is the Gaussian window function, g (k-1) (t) represents the k-1 order differential of the function g(t), x(τ) is the vibration signal of the faulty bearing, and * represents taking the conjugate; where t represents the time variable, ω represents the frequency variable, N is the order of the generalized time-frequency matrix B, k = 1, 2, ..., N, and the elements of the generalized time-frequency matrix are combined into the generalized time-frequency matrix B:

3. A bearing fault diagnosis method based on generalized time-frequency extraction transformation according to claim 2, characterized in that: The expression for the generalized group delay estimate in step 2 is As shown below: Among them, σ is the Gaussian window function parameter, Indicates taking the imaginary part of the function, B 1,k is the algebraic cofactor of the generalized time-frequency matrix B.

4. A bearing fault diagnosis method based on generalized time-frequency extraction transformation according to claim 2, characterized in that: The expression of the generalized time-frequency extraction transformation algorithm in step 2 is: Among them, δ(·) is the Dirac function, D is the operator matrix, and D 1,k Represents the algebraic co-factor of the operator matrix D. The expression of D is as follows: Wherein, g(0) is the Gaussian window function at t=0.

5. The bearing fault diagnosis method based on generalized time-frequency extraction transformation according to claim 1 is characterized in that: When the generalized time-frequency extraction transform algorithm is used to extract the energy of the combined time-frequency distribution at the generalized group delay estimate, the retention time is equal to the time-frequency energy at the generalized group delay estimate, and the time-frequency energy at other positions is set to zero.

6. The bearing fault diagnosis method based on generalized time-frequency extraction transformation according to claim 1 is characterized in that: The step 4 is specifically as follows: Step 4.1: Calculate the fault characteristic frequencies of the bearing under different fault types according to the bearing structural parameters; Step 4.2: Calculate the time interval Δt between two adjacent pulse components under different fault types according to the fault characteristic frequency in step 4.1; Step 4.3: Compare the time interval extracted in step 3 with Δt in step 4.2 to determine the fault type of the bearing.

7. A bearing fault diagnosis method based on generalized time-frequency extraction transformation according to claim 6, characterized in that: In step 4.1, the following formula is used to calculate the fault characteristic frequency of the bearing under different fault types: Among them, f BPFO is the characteristic frequency of outer race fault, f BPFI is the characteristic frequency of the inner race fault, f BSF is the characteristic frequency of rolling element failure, f FRF is the characteristic frequency of cage failure, N b and d represent the number and diameter of rolling elements respectively, D and They represent the pitch diameter and contact angle of the bearing, respectively, r The main axis speed.

Citation Information

Cited By

  • Mechanical transient fault diagnosis method and equipment based on group delay ridge optimization

    CN119475106A

  • Drag-free control system micro-thruster fault diagnosis method based on weight embedding and heuristic information

    CN120103817A

  • Drag-free control system microthruster fault diagnosis method based on weight embedding and heuristic information

    CN120103817B

  • Vibration diagnosis method and system for abrasion of main shaft bearing of turboshaft engine

    CN120352148A

  • Bearing fault diagnosis method based on time odd symmetry extraction transformation

    CN120761037A