Flexible thin-wall bearing fault diagnosis method based on CPO optimization MCKD
By introducing CPO optimization in the MCKD algorithm, the key parameters in the MCKD algorithm are synchronized, and the fault diagnosis problem in the background of flexible thin -walled bearing is solved, and efficient fault feature extraction and diagnostic effects are achieved.
Patent Information
- Application Number
- CN202411330710.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-24
- Publication Date
- 2025-01-28
AI Technical Summary
Under the interference of strong impact background, the problem of extracting weak fault features of flexible thin-wall bearings, as well as the problem of difficult to determine the filter length L, displacement number M and deconvolution period T in the MCKD algorithm, resulting in poor fault diagnosis effect of flexible thin-wall bearings. good.
Adopt the MCKD algorithm based on CPO optimization, synchronize the key parameter L, M and T in MCKD through the CPO of the group intelligent optimization algorithm, and then realize the outer circle failure diagnosis of flexible thin -walled bearing, and effectively filter out the modulation in the signal to remove the modulation in the signal. Frequency and modulation edge band.
The external circle fault diagnosis of flexible thin -walled bearing is successfully achieved. The optimization process only needs 8 iterations, and can effectively filter noise and hybrid frequency, which improves the accuracy and efficiency of diagnosis.
Smart Images
Figure SMS_1 
Figure SMS_2 
Figure SMS_5
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of flexible thin-walled bearing fault diagnosis, and in particular to a flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD. Background Art
[0002] As one of the core components of the harmonic reducer, the flexible thin-walled bearing is similar in structure to the conventional rolling bearing, and is also composed of four parts: rolling element, cage, inner raceway and outer raceway. The wall thickness of its inner and outer raceways is very thin. When they are used as independent parts, they are circular, but once they are installed on the camshaft, they are forced to deform into an ellipse. When this elliptical bearing rotates, it will cause cyclic alternating stress to act on the thin wall, making it prone to failure and generating fault shocks. At the same time, the alternating changes in the long and short axes will produce strong background shock components superimposed on the fault shocks, which greatly increases the difficulty of FTB fault diagnosis. Therefore, studying the fault diagnosis method of flexible thin-walled bearings and monitoring and diagnosing their operating status are of great significance for the reliable and continuous operation of high-end mechanical equipment such as robots.
[0003] The research on bearing fault diagnosis technology has a long history and has achieved many results. Harting proposed the resonance demodulation technology in 1978. By analyzing the high-frequency vibration excited by the periodic pulse force generated by bearing defects, the bearing fault is diagnosed based on the characteristics of the spectrum graph using filtering, envelope detection and Fourier transform. As the core method for rolling bearing and gear fault diagnosis, envelope analysis has successively developed demodulation technologies such as Hilbert transform and generalized detection filtering, but its application scope is limited because of the need to manually set parameters. To cope with complex scenarios, researchers have proposed new methods such as spectral kurtosis (SK) and maximum correlated kurtosis deconvolution (MCKD). The former is difficult to detect transient shocks under strong noise, and the latter is more resistant to noise, but it is difficult to set parameters such as filter length, displacement number and deconvolution period. Miao et al. improved MCKD and successfully applied it to locomotive bearing fault diagnosis by iteratively approximating the real fault period through autocorrelation calculation. However, if the filter length and displacement number are set improperly, the diagnostic effect will be affected. Chen Hui, Liu Xingjiao and others used the control variable method to determine these parameters and effectively identified the faults of flexible thin-walled bearings, but this method is sensitive to speed changes and cannot optimize the three parameters of MCKD at the same time.
[0004] Therefore, an effective diagnostic method is urgently needed to address the difficulty in extracting weak fault features of flexible thin-walled bearings under strong impact background interference, as well as the difficulty in determining the filter length L, displacement number M and deconvolution period T of the MCKD algorithm. Summary of the invention
[0005] The present invention aims to provide a flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD, which successfully realizes the outer ring fault diagnosis of the flexible thin-walled bearing and can effectively filter out the modulation frequency and modulation sideband in the signal.
[0006] The technical solution of the present invention is as follows:
[0007] The flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD comprises the following steps:
[0008] A. Obtain one-dimensional vibration data of the flexible thin-walled bearing during rotation;
[0009] B. Set the upper and lower bounds of optimization, the number of cycles, the number of crested porcupine populations, initialize the position of the crested porcupine population, and control the minimum size of the population:
[0010] C. Execute the MCKD algorithm, take the envelope spectrum peak factor as the fitness, calculate the fitness value, and update the current optimal position of each crested porcupine;
[0011] D. Execute CPO optimization: When the predator is far away, use the first and second defense strategies; when the predator is close, use the third and fourth defense strategies;
[0012] E. If the global optimal solution is found, exit the CPO loop and output the optimal parameter combination of L, M, and T; otherwise, loop through steps C and D until the optimal solution is obtained;
[0013] F. Substitute the optimal parameter combination of L, M, and T into MCKD to obtain the CPO-MCKD optimization algorithm, and use it to process the one-dimensional vibration data of the flexible thin-walled bearing. Compare the theoretical value of the fault characteristic frequency of FTB with the frequency value corresponding to the peak spectrum line in the envelope spectrum of the filtered signal to identify the fault type and location.
[0014] The position of the crested porcupine population is initialized using the following formula (1):
[0015] X i =L+r×(UL),i=1,2,L,N′ (1)
[0016] Where, X i is the i-th candidate solution in the search space, L and U are the upper and lower bounds of the search range, the elements in r are random numbers between 0 and 1, and N′ is the population size;
[0017] The following formula (2) is used to control the minimum size of the population:
[0018]
[0019] In the formula, T, Tmax Represent the number of cycles and the maximum number of cycles respectively, t is the current fitness function value, % is the remainder or modulus operator, N min is the minimum number of individuals in the newly generated population.
[0020] The formula of the first defense strategy is shown in the following formula (3):
[0021]
[0022] In the formula, is the optimal solution for the kth iteration, τ1 is a random number based on normal distribution, τ2 is a random value in the interval [0,1], represents the position of the predator at the kth iteration, which can be calculated by the following formula (4):
[0023]
[0024] In the formula, r is a random number between [1, N].
[0025] The formula of the second defense strategy is shown in the following formula (5):
[0026]
[0027] In the formula, r1 and r2 are random numbers between [1,N], and τ3 is a random value generated between 0 and 1.
[0028] The formula of the third defense strategy is shown in the following formula (6):
[0029]
[0030] In the formula, δ is a parameter determined by formula (7) and used to control the search direction, γ k is the defense factor determined by formula (8), is the odor diffusion factor determined by formula (9); U1 is a vector used to simulate the three situations that the third defense strategy may encounter, and its value is 0 or 1 or a combination of the two;
[0031]
[0032] In the formula, represents the fitness value of the ith individual at iteration k, ε is a minimum value that avoids division by zero, ζ contains a vector of randomly generated values between 0 and 1, λ is a variable that randomly generates numbers between 0 and 1, T is the current iteration number, T max is the maximum number of iterations.
[0033] The formula of the fourth defense strategy is shown in the following formula (10):
[0034]
[0035] In the formula, is the optimal solution for the kth iteration, represents the position of the predator corresponding to the i-th individual at the k-th iteration, α is the convergence speed factor, τ4 is a random value in the interval [0,1], is the average influence of the crested porcupine on the i-th predator, defined by the following formula (11);
[0036]
[0037] In the formula, is the mass of the i-th predator at the k-th iteration, f(g) represents the fitness function, is the velocity value of the i-th individual at the k-th and k+1-th iterations respectively, and τ6 is a random vector with a value in the interval [0,1].
[0038] Compared with the prior art, the present invention has the following advantages:
[0039] The method of the present invention proposes for the first time to use CPO in the swarm intelligence optimization algorithm to solve the problem that the key parameters in MCKD (filter length L, displacement number M, deconvolution period T) are difficult to determine simultaneously. The CPO algorithm is used to realize the synchronous optimization of the key parameters of MCKD, and the optimization process can be completed in only 8 iterations.
[0040] The method of the present invention uses the MCKD algorithm obtained by CPO optimization to successfully realize the outer ring fault diagnosis of flexible thin-walled bearings, and can effectively filter out the modulation frequency and modulation sidebands in the signal. In contrast, the asynchronous optimization MCKD algorithm can also realize FTB fault diagnosis, but the initial segment of its filtered signal time domain waveform will be distorted, and the modulation frequency and modulation sidebands in the corresponding envelope spectrum cannot be effectively filtered out. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 This is the vibration waveform and spectrum of a normal flexible thin-walled bearing;
[0042] Figure 2 The time domain waveform and spectrum of the bearing vibration signal of the flexible thin-walled bearing outer ring fault;
[0043] Figure 3 Iteration curve graph for fitness value;
[0044] Figure 4 To optimize the time domain waveform of the MCKD filter signal;
[0045] Figure 5 To optimize the envelope spectrum of the MCKD filtered signal;
[0046] Figure 6 To asynchronously optimize the MCKD filtered signal and its envelope spectrogram;
[0047] Figure 7 It is a comparison chart of the method of the present invention, the Grey Wolf Optimization (GWO) algorithm and the Particle Swarm Optimization (PSO) algorithm. DETAILED DESCRIPTION
[0048] The present invention is described in detail below with reference to the accompanying drawings and embodiments.
[0049] Example 1
[0050] The flexible thin-walled bearing fault diagnosis method based on CPO optimization MCKD of this embodiment includes the following steps:
[0051] The flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD comprises the following steps:
[0052] A. Obtain one-dimensional vibration data of the flexible thin-walled bearing during rotation;
[0053] B. Set the upper and lower bounds of optimization, the number of cycles, the number of crested porcupine populations, initialize the position of the crested porcupine population, and control the minimum size of the population:
[0054] The position of the crested porcupine population is initialized using the following formula (1):
[0055] X i =L+r×(UL),i=1,2,L,N′ (1)
[0056] Where, X i is the i-th candidate solution in the search space, L and U are the upper and lower bounds of the search range, the elements in r are random numbers between 0 and 1, and N′ is the population size;
[0057] The following formula (2) is used to control the minimum size of the population:
[0058]
[0059] In the formula, T, T max Represent the number of cycles and the maximum number of cycles respectively, t is the current fitness function value, % is the remainder or modulus operator, N min is the minimum number of individuals in the newly generated population.
[0060] C. Execute the MCKD algorithm, take the envelope spectrum peak factor as the fitness, calculate the fitness value, and update the current optimal position of each crested porcupine;
[0061] D. Execute CPO optimization: When the predator is far away, use the first and second defense strategies; when the predator is close, use the third and fourth defense strategies;
[0062] The formula of the first defense strategy is shown in the following formula (3):
[0063]
[0064] In the formula, is the optimal solution for the kth iteration, τ1 is a random number based on normal distribution, τ2 is a random value in the interval [0,1], represents the position of the predator at the kth iteration, which can be calculated by the following formula (4):
[0065]
[0066] In the formula, r is a random number between [1, N].
[0067] The formula of the second defense strategy is shown in the following formula (5):
[0068]
[0069] In the formula, r1 and r2 are random numbers between [1,N], and τ3 is a random value generated between 0 and 1.
[0070] The formula of the third defense strategy is shown in the following formula (6):
[0071]
[0072] In the formula, δ is a parameter determined by formula (7) and used to control the search direction, γ k is the defense factor determined by formula (8), is the odor diffusion factor determined by formula (9); U1 is a vector used to simulate the three situations that the third defense strategy may encounter, and its value is 0 or 1 or a combination of the two;
[0073]
[0074] In the formula, represents the fitness value of the ith individual at iteration k, ε is a minimum value that avoids division by zero, ζ contains a vector of randomly generated values between 0 and 1, λ is a variable that randomly generates numbers between 0 and 1, T is the current iteration number, T max is the maximum number of iterations.
[0075] The formula of the fourth defense strategy is shown in the following formula (10):
[0076]
[0077] In the formula, is the optimal solution for the kth iteration, represents the position of the predator corresponding to the i-th individual at the k-th iteration, α is the convergence speed factor, τ4 is a random value in the interval [0,1], is the average influence of the crested porcupine on the i-th predator, defined by the following formula (11);
[0078]
[0079] In the formula, is the mass of the i-th predator at the k-th iteration, f(g) represents the fitness function, is the velocity value of the i-th individual at the k-th and k+1-th iterations respectively, and τ6 is a random vector with a value in the interval [0,1].
[0080] E. If the global optimal solution is found, exit the CPO loop and output the optimal parameter combination of L, M, and T; otherwise, loop through steps C and D until the optimal solution is obtained;
[0081] F. Substitute the optimal parameter combination of L, M, and T into MCKD to obtain the CPO-MCKD optimization algorithm, and use it to process the one-dimensional vibration data of the flexible thin-walled bearing. Compare the theoretical value of the fault characteristic frequency of FTB with the frequency value corresponding to the peak spectrum line in the envelope spectrum of the filtered signal to identify the fault type and location.
[0082] Example 2 Experimental Analysis
[0083] 3.1 Test conditions
[0084] The test data comes from the acceleration sensor detection. The outer ring fault of the tested flexible thin-walled bearing is simulated by electrospark machining a rectangular pit with a length and width of 1 mm. The specific parameters are shown in Table 1.
[0085] Table 1 Basic test parameters
[0086]
[0087] 3.2 Vibration signal analysis
[0088] In the test, the motor speed was set to 1050r / min (i.e., the rotation frequency f n =17.5Hz), the vibration signals of the normal and outer ring faulty bearing FTB were measured respectively as follows Figure 1 and Figure 2 shown.
[0089] Depend on Figure 1(a) It can be seen that the vibration acceleration signal of the normal FTB is seriously interfered by noise, making its waveform characteristics blurred. Figure 1 (b) It can be seen that the double frequency 2f appears in the spectrum n and its higher harmonic components (2f n 、4f n 、6f n The reason is that the long and short axes of the FTB camshaft undergo two alternating changes every time it rotates. In summary, the vibration signal of a normal FTB contains an impact component caused by the alternating change of the long and short axes, which is completely different from the impact component that only appears when a common rolling bearing fails, greatly increasing the difficulty of FTB fault diagnosis.
[0090] Figure 2 is the vibration acceleration signal of the outer ring fault FTB, Figure 2 It can be seen that there are obvious impact components and random noise in the signal, and its amplitude is significantly higher than Figure 4 According to formula (8), the outer ring fault characteristic frequency f of FTB can be calculated eo The value range of is about 171.5-187.0Hz. Figure 2 In the spectrum of (b), no frequency components with particularly large amplitudes are found in this range. This is because f eo By other frequency components (2f n ) modulation.
[0091] 3.3 Fault diagnosis of flexible thin-walled bearing outer ring
[0092] Based on the method of Example 1, CPO is first used to optimize the three parameters [L, M, T] in the MCKD algorithm. The fitness function is solved in the frequency range [15,800], and its iteration curve is as follows: Figure 3 shown.
[0093] Depend on Figure 3 It can be seen that at the 8th iteration, the dimensionless fitness value E c The maximum value 15.5122 is obtained, and the value remains unchanged after iterating. At this time, the global optimal parameter combination [L o ,T o ,M o ] is [403,110,5]. Based on these optimization results, the filter length of the MCKD algorithm is set to 403, the deconvolution period is 110, and the displacement number is 5, and the CPO-optimized MCKD algorithm can be obtained. Figure 2 The FTB outer ring fault signal in is processed, and the result is as follows Figure 4-5 shown.
[0094] Figure 4 This is the time domain waveform of the filtered signal after the optimized MCKD algorithm is processed. Figure 2 (a) Figure 4 The impact component in is more obvious. The envelope spectrum of the filtered signal is as follows Figure 5 shown. Figure 5 It can be seen that the frequencies corresponding to the three peak spectrum lines are: 179.668Hz, 359.375Hz, 539.062Hz, which correspond to the outer ring fault characteristic frequencies f eo and its 2 times frequency 2f eo and 3 times the frequency 3f eo , and the noise between them is also filtered out. This shows that the proposed optimization algorithm can not only accurately extract the outer ring fault characteristics of FTB, but also has a good filtering effect on random noise and modulation frequency.
[0095] 3.4 Comparison with other methods
[0096] In order to illustrate the superiority of the proposed algorithm for FTB fault diagnosis, the asynchronous optimization MCKD method of the following reference [1] is used for comparison. Figure 2 The outer ring fault signal in is processed, and the obtained filtered signal and its envelope spectrum are as follows Figure 6 shown.
[0097] References [1]: Chen Hui, Fault feature extraction of flexible thin-walled bearings based on spectral kurtosis and MCKD [D]. Guangzhou: South China University of Technology, 2019
[0098] Depend on Figure 6 (a) It can be seen that asynchronous optimization of MCKD is adopted [1] The impact characteristics of the obtained filtered signal are also obvious, but the initial segment signal is relatively Figure 2 The original signal in has a certain degree of distortion. The corresponding envelope spectrum is as follows Figure 6 As shown in (b), Figure 6 It can be seen that in addition to the outer ring fault frequency f eo and its harmonic components (2f eo 、3f eo 、4f eo , 5f eo In addition to the double frequency 2f n In addition, from Figure 6 In (b), it can be clearly seen that there is a gap equal to twice the rotation frequency 2f on both sides of the outer ring fault frequency and its harmonic components. n The modulation side frequency components are the cause of the distortion of the initial filtering signal.
[0099] Example 3
[0100] The method of Example 1 of the present invention is compared with the prior art Grey Wolf Optimization (GWO) algorithm and Particle Swarm Optimization (PSO) algorithm. The three algorithms all use the peak factor of the envelope spectrum as the fitness evaluation criterion, the initial population is set to 20, and the number of iterations is set to 20. Figure 7 As shown in the figure, CPO surpasses GWO and PSO in terms of convergence speed and accuracy. Specifically, the PSO algorithm obtains the optimal solution for the first time in the 5th iteration, and its fitness value (Ec) is slightly lower than that of CPO and GWO. The GWO algorithm also reaches the optimal solution in the 9th iteration, and its fitness value (Ec) is also lower than that of CPO. The parameters of the MCKD algorithm are optimized by the CPO algorithm, and the optimal parameter combination [403,110,5] is determined and applied to the MCKD process.
Claims
1. A flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD, characterized in that: The following steps are involved: A. Obtain one-dimensional vibration data of the flexible thin-walled bearing during rotation; B. Set the upper and lower bounds of optimization, the number of cycles, the number of crested porcupine populations, initialize the position of the crested porcupine population, and control the minimum size of the population: C. Execute the MCKD algorithm, take the envelope spectrum peak factor as the fitness, calculate the fitness value, and update the current optimal position of each crested porcupine; D. Execute CPO optimization: When the predator is far away, use the first and second defense strategies; when the predator is close, use the third and fourth defense strategies; E. If the global optimal solution is found, exit the CPO loop and output the optimal parameter combination of L, M, and T; otherwise, loop through steps C and D until the optimal solution is obtained; F. Substitute the optimal parameter combination of L, M, and T into MCKD to obtain the CPO-MCKD optimization algorithm, and use it to process the one-dimensional vibration data of the flexible thin-walled bearing. Compare the theoretical value of the fault characteristic frequency of FTB with the frequency value corresponding to the peak spectrum line in the envelope spectrum of the filtered signal to identify the fault type and location.
2. The flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD as claimed in claim 1 is characterized in that: The position of the crested porcupine population is initialized using the following formula (1): X i =L+r×(U-L),i=1,2,L,N′ (1) Where, X i is the i-th candidate solution in the search space, L and U are the upper and lower bounds of the search range, the elements in r are random numbers between 0 and 1, and N′ is the population size; The following formula (2) is used to control the minimum size of the population: In the formula, T, T max Represent the number of cycles and the maximum number of cycles respectively, t is the current fitness function value, % is the remainder or modulus operator, N min is the minimum number of individuals in the newly generated population.
3. The flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD as claimed in claim 1, characterized in that: The formula of the first defense strategy is shown in the following formula (3): In the formula, is the optimal solution for the kth iteration, τ1 is a random number based on normal distribution, τ2 is a random value in the interval [0,1], represents the position of the predator at the kth iteration, which can be calculated by the following formula (4): In the formula, r is a random number between [1, N].
4. The flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD as claimed in claim 1, characterized in that: The formula of the second defense strategy is shown in the following formula (5): In the formula, r1 and r2 are random numbers between [1,N], and τ3 is a random value generated between 0 and 1.
5. The flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD as claimed in claim 1, characterized in that: The formula of the third defense strategy is shown in the following formula (6): In the formula, δ is a parameter determined by formula (7) and used to control the search direction, γ k is the defense factor determined by formula (8), is the odor diffusion factor determined by formula (9); U1 is a vector used to simulate the three situations that the third defense strategy may encounter, and its value is 0 or 1 or a combination of the two; In the formula, represents the fitness value of the ith individual at iteration k, ε is a minimum value that avoids division by zero, ζ contains a vector of randomly generated values between 0 and 1, λ is a variable that randomly generates numbers between 0 and 1, T is the current iteration number, T max is the maximum number of iterations.
6. The flexible thin-walled bearing fault diagnosis method based on CPO optimized MCKD as claimed in claim 1, characterized in that: The formula of the fourth defense strategy is shown in the following formula (10): In the formula, is the optimal solution for the kth iteration, represents the position of the predator corresponding to the i-th individual at the k-th iteration, α is the convergence rate factor, τ4 is a random value in the interval [0,1], and F i k is the average influence of the crested porcupine on the i-th predator, defined by the following formula (11); In the formula, is the mass of the i-th predator at the k-th iteration, f(g) represents the fitness function, is the velocity value of the i-th individual at the k-th and k+1-th iterations respectively, and τ6 is a random vector with a value in the interval [0,1].
Citation Information
Cited By
Controller fault diagnosis method and device based on machine learning
CN122411394A