Motor bearing fault diagnosis method under strong noise interference

Abstract

The invention aims to provide a motor bearing fault diagnosis method under strong noise interference. The method comprises the following steps: A, acquiring an acceleration vibration signal of a motor bearing by using a piezoelectric acceleration sensor; b, taking the minimum envelope spectrum peak factor Ec as a fitness function, optimizing a center frequency omega d and a balance factor alpha in a VME algorithm model by using a WOA algorithm model to obtain a WOA-VME model, and decomposing the collected acceleration vibration signal to obtain an expected modal signal; c, taking the minimum envelope spectrum peak factor Ec as a fitness function, optimizing the size f of a filter in the SMHD algorithm model by using the WOA algorithm model, and processing the expected modal signal by using the WOA-SMHD model to obtain an enhanced expected modal signal; and D, extracting motor bearing fault features from the enhanced expected mode signal, and comparing the motor bearing fault features with the motor bearing fault frequency database to obtain a diagnosis result. According to the invention, motor bearing fault feature extraction of strong background noise can be well realized.

Description

Technical Field

[0001] This invention relates to the field of motor bearing fault diagnosis technology, and specifically to a method for diagnosing motor bearing faults under strong noise interference. Background Technology

[0002] Bearing fault signals exhibit nonlinear and non-stationary characteristics, and fault features are difficult to extract under strong background noise. To address this issue, researchers often employ variational mode decomposition (VMD) methods for bearing fault diagnosis. Shi Jie et al. [1] The Dynamic Model Decomposition (VMD) was optimized using the Fishing Optimization Algorithm (CFOA), and marginal spectrum analysis was performed on the decomposed modal components to successfully determine the fault type and severity of the bearing. (Zhong Xianyou et al.) [2] A method combining Continuous Variational Mode Decomposition (SVMD) and Maximum Correlation Kujicic Deconvolution (MCKD) is proposed to enhance the periodic impact component of bearings, effectively extracting fault features of the inner and outer rings of rolling bearings even in disturbed environments. However, the VMD method faces challenges in selecting the penalty factor and the number of decomposition layers. Therefore, researchers have proposed a Variational Mode Extraction (VME) method. [3] This method does not require a preset number of decomposition layers, has lower computational complexity than VMD, and has better processing capabilities.

[0003] The Sparse Maximum Harmonic Noise Ratio Deconvolution (SMHD) method is a signal enhancement method based on sparse representation and harmonic noise ratio optimization. It maximizes the ratio of harmonics to noise in the signal and extracts sparse impulse components from the signal through deconvolution techniques, effectively enhancing the signal under strong background noise. (Tang Guiji et al.) [4] The VME method is used to extract specific signal components from vibration signals, and these components are then subjected to SMHD processing. Envelope spectrum analysis is performed on the resulting deconvolutioned signal to extract bearing damage characteristic frequencies. Although these methods have achieved good results, the choice of center frequency and balance factor in the VME method, as well as the filter size in the SMHD, significantly affects the final accuracy of the algorithm. Summary of the Invention

[0004] This invention aims to provide a method for fault diagnosis of motor bearings under strong noise interference. This method integrates the VME and SMHD algorithms, and further optimizes the key parameters using the WOA algorithm, enabling it to effectively extract fault features from motor bearings in strong background noise. Experimental results show that under noise interference, the method clearly reveals multi-harmonic fault features and accurately identifies the outer ring crack fault type of the bearing.

[0005] The technical solution of the present invention is as follows:

[0006] The method for diagnosing motor bearing faults under strong noise interference includes the following steps:

[0007] A. Use a piezoelectric accelerometer to collect the acceleration vibration signal of the motor bearing;

[0008] B. Using the minimum envelope peak factor Ec as the fitness function, the center frequency ω in the VME algorithm model is optimized using the WOA algorithm model. d The equilibrium factor α is used to decompose the collected acceleration vibration signal using the parameter-optimized VME model to obtain the desired modal signal.

[0009] C. Using the minimum envelope spectrum peak factor Ec as the fitness function, the filter size f in the SMHD algorithm model is optimized using the WOA algorithm model. The desired modal signal is then processed using the parameter-optimized SMHD algorithm model to further enhance the fault impact component in the mode, thereby obtaining the enhanced desired modal signal.

[0010] D. Extract motor bearing fault features from the enhanced desired mode signal and compare them with the motor bearing fault frequency database to obtain diagnostic results.

[0011] In steps B and C, the center frequency ω of the VME in the WOA algorithm model is... d The optimization process for the filter size f, along with the balance factor α and SMHD, is as follows:

[0012] Set the upper and lower limits of the value range and dimensions of each parameter to be optimized, and perform the optimization process as follows to finally obtain the three optimal parameters, which are then input into the VME algorithm model and the SMHD algorithm model as the initial parameters of these two models.

[0013] Optimization process:

[0014] Whales update their location by randomly searching for prey using the following formula:

[0015]

[0016] In the formula, D is the distance vector between the whale and its prey; t is the current iteration number; X(t) is the position vector of the whale in the t-th iteration; X r (t) is the position vector of a random whale in the current group; A and B are coefficient vectors, where A = 2a·r1-a, B = 2r2, r1 and r2 are random vectors in [0,1], and a is the convergence factor, which decays with the increase of the number of iterations;

[0017] Suppose X b (t) represents the optimal position of the current individual in the group. The remaining whales in the group surround the optimal individual using the following formula:

[0018]

[0019] In the formula, k is a constant used to define the shape of the logarithmic spiral; h is a random number in the range [0,1].

[0020] In steps B and C, the formula for calculating the minimum envelope spectrum peak factor Ec is as follows:

[0021]

[0022] In the formula, Y i for Envelope spectrum amplitude, f i ' represents the fault frequency of the vibration signal, and μ is set to 4.

[0023] In steps B and C, the population size of the WOA algorithm model is 50, and the maximum number of iterations is 50.

[0024] In step B, the center frequency ω of the VME algorithm d The value range is [0.02πf]. s ,0.98πf s ], f s The sampling frequency is given, and the balance factor α ranges from [200, 500].

[0025] In step C, the filter size f of the SMHD algorithm has a range of [120, 2400].

[0026] The VME algorithm model is as follows:

[0027] ① Decompose the input signal f(t) into two parts:

[0028] f(t) = u d (t)+f r (t) (4)

[0029] In the formula, u d (t) represents the desired mode, f r (t) represents the residual signal;

[0030] The extracted modes satisfy three conditions: (1) u d (t) needs to be at its center frequency ω d Nearby contraction; (2)u d (t) and f r The spectral overlap between (t) needs to be minimized; (3)u d (t) and f r (t) can completely reconstruct f(t);

[0031] The constraints of formula (4) are as follows:

[0032]

[0033] In the formula, α is a balance factor used to control the modal bandwidth. J1 is the first penalty function used to ensure the smoothness of the modes, and J2 is the second penalty function used to ensure that the error between the reconstructed signal and the original signal is minimized.

[0034] The expressions for J1 and J2 are:

[0035]

[0036] In the formula, δ represents the Dirac distribution, and * represents convolution. Let β(t) be the selected frequency domain filter;

[0037] ② After converting the time-domain signal into a frequency-domain signal using Fourier transform, the update expressions for the center frequency and the desired mode are:

[0038]

[0039] In the formula, n is the number of iterations, λ is the Lagrange multiplier, and its update equation is:

[0040]

[0041] In the formula, τ is the update parameter of the VME algorithm.

[0042] The processing steps of the VME algorithm model include:

[0043] ① Initialization and Let n = 0;

[0044] ② Let n = n + 1, and execute the algorithm;

[0045] ③ For all frequency domain signals with ω≥0, update ω using formulas (8) and (9) respectively. d ,

[0046] ④ Set the precise value ε > 0, and judge based on the convergence condition: "If the convergence condition is not met, return to step ② to continue iterating; if the convergence condition is met, stop the loop. The formula for the convergence condition is:

[0047]

[0048] The processing procedure of the SMHD algorithm model is as follows:

[0049] (1) Input the desired mode signal x(t) after VME decomposition, and filter it to obtain signal y1;

[0050] (2) Sparsify the signal y1 to obtain the sparse signal y′1;

[0051] (3) Calculate the autocorrelation matrix A of x(t), and the correlation between the signal x and the filtered signal y. i The cross-correlation matrix b is used to update the filter coefficients;

[0052] (4) Calculate the harmonic noise ratio of the envelope of the sparse signal y′1, compare the value with the threshold to estimate the new period. The threshold is updated according to the change of the kurtosis of the filtered signal, so as to estimate the new period to be used in the next iteration; the initial threshold is the average value of the original signal.

[0053] (5) If the harmonic noise ratio of the sparse signal envelope cannot satisfy the periodic value of x(t), then repeat the above process and iterate according to i = i + 1, and finally output the filtered signal y′. i Filtered signal y′ i This is to enhance the desired modal signal.

[0054] In step (2), the calculation formula for sparsification is as follows:

[0055]

[0056] In the formula, σ is the sparsity threshold;

[0057] In step (3), the update formula for the filter coefficients is:

[0058]

[0059] In the formula, A is the autocorrelation matrix of the signal x(t), f is the inverse filter, j = 1, 2, ..., L, L is the filter length, n = 1, 2, ..., N, N is the length of the signal x(t);

[0060] In step (4), the formula for calculating the harmonic noise ratio of the envelope of the sparse signal y′1 is as follows:

[0061]

[0062] In the formula, T0 is the period of the desired modal signal being loaded.

[0063] The specific process of step D is as follows:

[0064] The filtered signal y′ iAfter Hilbert transform, an analytical signal is obtained. The absolute value of the analytical signal is subtracted from its mean value to obtain its envelope spectrum. The envelope spectrum is compared with the motor bearing fault frequency database to determine whether the frequency range in the envelope spectrum includes the theoretical bearing fault characteristic frequency and its integer multiples. If so, the corresponding fault prompt information is output; otherwise, a normal prompt information is output.

[0065] The algorithm names corresponding to the abbreviations in Chinese and English in this invention are as follows:

[0066] VMD: Variational Mode Decomposition;

[0067] VME: Variational Mode Extraction;

[0068] MCKD: Maximum Correlated Kurtosis Deconvolution;

[0069] SMHD: Sparse Maximum Harmonics-to-Noise Ratio Deconvolution (SMHD);

[0070] WOA: Whale Optimization Algorithm.

[0071] The advantages of this invention compared to existing technologies are:

[0072] The method of this invention combines VME and SMHD, and considering that the performance of VME and SMHD is affected by parameters, the whale optimization algorithm is used to optimize the center frequency of VME, balance factor and filter size of SMHD. This can successfully extract bearing fault features covered by strong background noise and achieve more accurate bearing fault diagnosis. Attached Figure Description

[0073] Figure 1 The time-domain waveform diagram of the faulty bearing experimental signal in Example 2;

[0074] Figure 2 The envelope spectrum of the faulty bearing experimental signal in Example 2;

[0075] Figure 3 The waveform of the faulty bearing experimental signal after adding noise is shown in Example 2.

[0076] Figure 4The image shows the envelope spectrum of the faulty bearing experimental signal after adding noise, as shown in Example 2.

[0077] Figure 5 The fitness value curve of the VME optimized by WOA in Example 2;

[0078] Figure 6 The waveform of the desired mode signal after WOA-VME decomposition in Example 2 is shown in the time domain.

[0079] Figure 7 The image shows the envelope spectrum of the desired mode signal after WOA-VME decomposition in Example 2.

[0080] Figure 8 The fitness value curve of WOA-optimized SMHD in Example 2 is shown.

[0081] Figure 9 The following is a time-domain waveform diagram of the signal after WOA-SMHD processing in Example 2;

[0082] Figure 10 This is the signal envelope spectrum after WOA-SMHD processing in Example 2. Detailed Implementation

[0083] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0084] Example 1

[0085] The method for diagnosing motor bearing faults under strong noise interference includes the following steps:

[0086] A. Use a piezoelectric accelerometer to collect the acceleration vibration signal of the motor bearing;

[0087] B. Using the minimum envelope peak factor Ec as the fitness function, the center frequency ω in the VME algorithm model is optimized using the WOA algorithm model. d The equilibrium factor α is used to decompose the collected acceleration vibration signal using the parameter-optimized VME model to obtain the desired modal signal.

[0088] The VME algorithm model is as follows:

[0089] ① Decompose the input signal f(t) into two parts:

[0090] f(t) = u d (t)+f r (t) (4)

[0091] In the formula, u d (t) represents the desired mode, f r (t) represents the residual signal;

[0092] The extracted modes satisfy three conditions: (1) u d (t) needs to be at its center frequency ω d Nearby contraction; (2)u d (t) and f r The spectral overlap between (t) needs to be minimized; (3)u d (t) and f r (t) can completely reconstruct f(t);

[0093] The constraints of formula (1) are as follows:

[0094]

[0095] In the formula, α is a balance factor used to control the modal bandwidth. J1 is the first penalty function used to ensure the smoothness of the modes, and J2 is the second penalty function used to ensure that the error between the reconstructed signal and the original signal is minimized.

[0096] The expressions for J1 and J2 are:

[0097]

[0098]

[0099] In the formula, δ represents the Dirac distribution, and * represents convolution. Let β(t) be the selected frequency domain filter, and let β(t) be its impulse response.

[0100] ② After converting the time-domain signal into a frequency-domain signal using Fourier transform, the update expressions for the center frequency and the desired mode are:

[0101]

[0102] In the formula, n is the number of iterations, λ is the Lagrange multiplier, and its update equation is:

[0103]

[0104] In the formula, τ is the update parameter of the VME algorithm.

[0105] The processing steps of the VME algorithm model include:

[0106] ① Initialization and Let n = 0;

[0107] ② Let n = n + 1, and execute the algorithm;

[0108] ③ For all frequency domain signals with ω≥0, update ω using formulas (5) and (6) respectively. d ,

[0109] ④ Set the precise value ε > 0, and judge based on the convergence condition: "If the convergence condition is not met, return to step ② to continue iterating; if the convergence condition is met, stop the loop. The formula for the convergence condition is:

[0110]

[0111] C. Using the minimum envelope spectrum peak factor Ec as the fitness function, the filter size f in the SMHD algorithm model is optimized using the WOA algorithm model. The desired modal signal is then processed using the parameter-optimized SMHD algorithm model to further enhance the fault impact component in the mode, thereby obtaining the enhanced desired modal signal.

[0112] The processing procedure of the SMHD algorithm model is as follows:

[0113] (1) Input the desired mode signal x(t) after VME decomposition, and filter it to obtain signal y1;

[0114] (2) Sparsify the signal y1 to obtain the sparse signal y′1;

[0115] (3) Calculate the autocorrelation matrix A of x(t), and the correlation between the signal x and the filtered signal y. i The cross-correlation matrix b is used to update the filter coefficients;

[0116] (4) Calculate the harmonic noise ratio of the envelope of the sparse signal y′1, compare the value with the threshold to estimate the new period. The threshold is updated according to the change of the kurtosis of the filtered signal, so as to estimate the new period to be used in the next iteration; the initial threshold is the average value of the original signal.

[0117] (5) If the harmonic noise ratio of the sparse signal envelope cannot satisfy the periodic value of x(t), then repeat the above process and iterate according to i = i + 1, and finally output the filtered signal y′. i Filtered signal y′ i This is to enhance the desired modal signal.

[0118] In step (2), the calculation formula for sparsification is as follows:

[0119]

[0120] In the formula, σ is the sparsity threshold;

[0121] In step (3), the update formula for the filter coefficients is:

[0122]

[0123] In the formula, A is the autocorrelation matrix of the signal x(t), f is the inverse filter, j = 1, 2, ..., L, L is the filter length, n = 1, 2, ..., N, N is the length of the signal x(t);

[0124] In step (4), the formula for calculating the harmonic noise ratio of the envelope of the sparse signal y′1 is as follows:

[0125]

[0126] In the formula, T0 is the period of the desired modal signal being loaded.

[0127] D. Extract motor bearing fault features from the enhanced desired mode signal and compare them with the motor bearing fault frequency database to obtain diagnostic results.

[0128] In steps B and C, the center frequency ω of the VME in the WOA algorithm model is... d The optimization process for the filter size f, along with the balance factor α and SMHD, is as follows:

[0129] Set the upper and lower limits of the value range and dimensions of each parameter to be optimized, and perform the optimization process as follows to finally obtain the three optimal parameters above, which are then input into the VME algorithm model and the SMHD algorithm model as the initial parameters of these two models.

[0130] Optimization process:

[0131] Whales update their location by randomly searching for prey using the following formula:

[0132]

[0133] In the formula, D is the distance vector between the whale and its prey; t is the current iteration number; X(t) is the position vector of the whale in the t-th iteration; X r (t) is the position vector of a random whale in the current group; A and B are coefficient vectors, where A = 2a·r1-a, B = 2r2, r1 and r2 are random vectors in [0,1], and a is the convergence factor, which decays with the increase of the number of iterations;

[0134] Suppose X b (t) represents the optimal position of the current individual in the group. The remaining whales in the group surround the optimal individual using the following formula:

[0135]

[0136] In the formula, k is a constant used to define the shape of the logarithmic spiral; h is a random number in the range [0,1].

[0137] The formula for calculating the minimum envelope spectral peak factor Ec is as follows:

[0138]

[0139] In the formula, Y i for Envelope spectrum amplitude, f i ' represents the fault frequency of the vibration signal, and μ is set to 4.

[0140] The population size of the WOA algorithm model is 50, and the maximum number of iterations is 50.

[0141] In step B, the center frequency ω of the VME algorithm d The value range is [0.02πf]. s ,0.98πf s ], f s The sampling frequency is given, and the balance factor α ranges from [200, 500].

[0142] In step C, the filter size f of the SMHD algorithm has a range of [120, 2400].

[0143] The specific process of step D is as follows:

[0144] The filtered signal y′ i After Hilbert transform, an analytical signal is obtained. The absolute value of the analytical signal is subtracted from its mean value to obtain its envelope spectrum. The envelope spectrum is compared with the motor bearing fault frequency database to determine whether the frequency range in the envelope spectrum includes the theoretical bearing fault characteristic frequency and its integer multiples. If so, the corresponding fault prompt information is output; otherwise, a normal prompt information is output.

[0145] Example 2

[0146] Using the Western Reserve University bearing fault dataset as an example, outer ring fault data was selected, corresponding to a fault diameter of 0.007 inches, a motor load of 1 HP, and a speed of 1797 r / min. The experimental data sampling frequency was 12 kHz, sourced from the drive-end accelerometer, with the outer ring fault located at the 6 o'clock position. The fault detection test was performed using the method of Example 1, as follows:

[0147] Step 1: Use an accelerometer to measure the outer ring of the faulty bearing to obtain the acceleration vibration signal; its time-domain waveform and envelope spectrum are shown below. Figure 1 , 2 As shown. Furthermore, to simulate the high background noise under actual working conditions, -11dB Gaussian white noise was added to the vibration signal as the experimental signal. Its time-domain waveform and envelope spectrum are shown below. Figure 3 , 4 As shown. Figure 4 and Figure 1 In comparison, it can be seen that the impact components of the faulty bearing have been masked by the strong background noise. Figure 7 It is also impossible to directly observe the impact frequency components.

[0148] Step 2: Use the WOA algorithm to optimize the VME parameters and determine the optimal parameter combination [ω]. d The population size of the WOA algorithm is 50, the maximum number of iterations is 50, and the center frequency ω in the VME algorithm is... d The value range is [754, 36926], the balance factor α ranges from [200, 500], and the optimization result obtained is ω. d =3898, α=397. Its optimized fitness value curve is shown below. Figure 5 As shown.

[0149] Step 3: Set the center frequency ω in the VME algorithm d The value is 3898, and the balance factor α is 397. The experimental signal is decomposed using VME, and the time-domain waveform and envelope spectrum of the desired mode after decomposition are shown below. Figure 6 , 7 As shown.

[0150] Step 4: Based on the theoretical formula for calculating the characteristic frequency of rolling bearing failure and the specific bearing parameters of the faulty bearing SKF6205 in Table 1, the multiple relationship between the characteristic frequency of the faulty bearing and the rotational frequency is calculated, as shown in Table 2.

[0151] Table 1 Specific parameters of the faulty bearing

[0152] bearing model bearing pitch diameter / mm Rolling element diameter / mm Number of rolling elements Contact angle / (°) SKF6205 39.04 7.94 9 0

[0153] Table 2 Fault Frequency and Frequency Multiple

[0154] Inner ring fault Outer ring fault Rolling element failure cage failure 5.415 3.585 2.357 0.398

[0155] Based on the rotational frequency of the shaft where the faulty bearing is located (1797 rpm / 60 = 29.95 Hz), the calculated fault frequency of the outer ring is 107.37 Hz. Figure 7 The frequency is close to 108Hz, so this frequency is a first harmonic. But from... Figure 7 Only the first harmonic can be seen in the signal, and the status of other harmonics cannot be seen. Therefore, the desired modal signal needs to be processed by SMHD.

[0156] Step 5: The WOA algorithm is used to optimize the SMHD parameters and determine the optimal filter size f. The WOA algorithm has a population size of 50 and a maximum number of iterations of 50. Its optimized fitness curve is shown below. Figure 8 As shown.

[0157] Step 6: Process the desired modal signal using the WOA-SMHD method. The processed time-domain waveform and envelope spectrum are shown below. Figure 9 , 10 As shown in the diagram, by observing the processed envelope spectrum, the fault frequency of 107.37Hz and its harmonics are clearly visible, thus indicating that the bearing fault is an outer ring fault.

[0158] In summary, the motor bearing fault diagnosis method under strong noise interference of the present invention can successfully extract bearing fault features masked by strong background noise. The combination of VME and SMHD can achieve more accurate bearing fault diagnosis; the WOA algorithm can realize the parameter selection of VME.

[0159] The above description is merely the preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for diagnosing motor bearing faults under strong noise interference, characterized in that, Includes the following steps: A. Use a piezoelectric accelerometer to collect the acceleration vibration signal of the motor bearing; B. Using the minimum envelope peak factor Ec as the fitness function, the center frequency ω in the VME algorithm model is optimized using the WOA algorithm model. d The equilibrium factor α is used to decompose the collected acceleration vibration signal using the parameter-optimized VME model to obtain the desired modal signal. C. Using the minimum envelope spectrum peak factor Ec as the fitness function, the filter size f in the SMHD algorithm model is optimized using the WOA algorithm model. The desired modal signal is then processed using the parameter-optimized SMHD algorithm model to further enhance the fault impact component in the mode, thereby obtaining the enhanced desired modal signal. D. Extract motor bearing fault features from the enhanced desired mode signal and compare them with the motor bearing fault frequency database to obtain diagnostic results.

2. The method for diagnosing motor bearing faults under strong noise interference as described in claim 1, characterized in that: In steps B and C, the center frequency ω of the VME in the WOA algorithm model is... d The optimization process for the filter size f, along with the balance factor α and SMHD, is as follows: Set the upper and lower limits of the value range and dimensions of each parameter to be optimized, and perform the optimization process as follows to finally obtain the three optimal parameters, which are then input into the VME algorithm model and the SMHD algorithm model as the initial parameters of these two models. Optimization process: Whales update their location by randomly searching for prey using the following formula: In the formula, D is the distance vector between the whale and its prey; t is the current iteration number; X(t) is the position vector of the whale in the t-th iteration; X r (t) is the position vector of a random whale in the current group; A and B are coefficient vectors, where A = 2a·r1-a, B = 2r2, r1 and r2 are random vectors in [0,1], and a is the convergence factor, which decays with the increase of the number of iterations; Suppose X b (t) represents the optimal position of the current individual in the group. The remaining whales in the group surround the optimal individual using the following formula: In the formula, k is a constant used to define the shape of the logarithmic spiral; h is a random number in the range [0,1].

3. The method for diagnosing motor bearing faults under strong noise interference as described in claim 2, characterized in that: In steps B and C, the formula for calculating the minimum envelope spectrum peak factor Ec is as follows: In the formula, Y i for Envelope spectrum amplitude, f i ' represents the fault frequency of the vibration signal, and μ is set to 4.

4. The method for diagnosing motor bearing faults under strong noise interference as described in claim 2, characterized in that: In steps B and C, the population size of the WOA algorithm model is 50, and the maximum number of iterations is 50. In step B, the center frequency ω of the VME algorithm d The value range is [0.02πf]. s ,0.98πf s ], f s The sampling frequency is given, and the balance factor α ranges from [200, 500]. In step C, the filter size f of the SMHD algorithm has a range of [120, 2400].

5. The method for diagnosing motor bearing faults under strong noise interference as described in claim 1, characterized in that: The VME algorithm model is as follows: ① Decompose the input signal f(t) into two parts: f(t)=u d (t)+f r (t) (4) In the formula, u d (t) represents the desired mode, f r (t) represents the residual signal; The extracted modes satisfy three conditions: (1) u d (t) needs to be at its center frequency ω d Nearby contraction; (2)u d (t) and f r The spectral overlap between (t) needs to be minimized; (3)u d (t) and f r (t) can completely reconstruct f(t); The constraints of formula (4) are as follows: In the formula, α is a balance factor used to control the modal bandwidth. J1 is the first penalty function used to ensure the smoothness of the modes, and J2 is the second penalty function used to ensure that the error between the reconstructed signal and the original signal is minimized. The expressions for J1 and J2 are: In the formula, δ represents the Dirac distribution, and * represents convolution. Let β(t) be the selected frequency domain filter, and let β(t) be its impulse response. ② After converting the time-domain signal into a frequency-domain signal using Fourier transform, the update expressions for the center frequency and the desired mode are: In the formula, n is the number of iterations, λ is the Lagrange multiplier, and its update equation is: In the formula, τ is the update parameter of the VME algorithm.

6. The method for diagnosing motor bearing faults under strong noise interference as described in claim 2, characterized in that: The processing steps of the VME algorithm model include: ① Initialization and Let n = 0; ② Let n = n + 1, and execute the algorithm; ③ For all frequency domain signals with ω≥0, update ω using formulas (8) and (9) respectively. d , ④ Set the precise value ε > 0, and judge based on the convergence condition. If the convergence condition is not met, return to step ② to continue iterating; if the convergence condition is met, stop the loop. The formula for the convergence condition is:

7. The method for diagnosing motor bearing faults under strong noise interference as described in claim 1, characterized in that: The processing procedure of the SMHD algorithm model is as follows: (1) Input the desired mode signal x(t) after VME decomposition, and filter it to obtain signal y1; (2) Sparsify the signal y1 to obtain the sparse signal y1′; (3) Calculate the autocorrelation matrix A of x(t), and the correlation between the signal x and the filtered signal y. i The cross-correlation matrix b is used to update the filter coefficients; (4) Calculate the harmonic noise ratio of the envelope of the sparse signal y1′, compare the value with the threshold to estimate the new period. The threshold is updated according to the change of the kurtosis of the filtered signal, so as to estimate the new period to be used in the next iteration; the initial threshold is the average value of the original signal. (5) If the harmonic noise ratio of the sparse signal envelope cannot satisfy the periodic value of x(t), then repeat the above process and iterate according to i = i + 1, and finally output the filtered signal y. i ′, filtered signal y i ′ is the enhanced desired modal signal.

8. The method for diagnosing motor bearing faults under strong noise interference as described in claim 7, characterized in that: In step (2), the calculation formula for sparsification is as follows: In the formula, σ is the sparsity threshold; In step (3), the update formula for the filter coefficients is: In the formula, A is the autocorrelation matrix of the signal x(t), f is the inverse filter, j = 1, 2, ..., L, L is the filter length, n = 1, 2, ..., N, N is the length of the signal x(t); In step (4), the formula for calculating the harmonic noise ratio of the envelope of the sparse signal y1′ is as follows: In the formula, T0 is the period of the desired modal signal being loaded.

9. The method for diagnosing motor bearing faults under strong noise interference as described in claim 7, characterized in that: The specific process of step D is as follows: The filtered signal y′ i After Hilbert transform, the analytic signal is obtained. The absolute value of the analytic signal is subtracted from the mean value to obtain its envelope spectrum. The envelope spectrum is compared with the motor bearing fault frequency database to determine whether the frequency range in the envelope spectrum includes the theoretical bearing fault characteristic frequency and their respective integer multiples. If yes, output the corresponding fault message; otherwise, output a normal message.

Images