Rolling bearing fault diagnosis method based on IMVO optimized MCKD denoising

Abstract

The application discloses a rolling bearing fault diagnosis method based on IMVO optimized MCKD denoising, solves the problems of the limitations of the parameter selection of the existing MCKD algorithm often depending on the artificial experience setting and the slow iteration speed of the MVO algorithm caused by the influence of TDR change, and the technical solution points are as follows: collecting vibration signals in the running process of a wind turbine rolling bearing, adaptively improving the TDR of the MVO algorithm, obtaining an improved multi-universe optimization algorithm IMVO, taking the Gini coefficient and the envelope spectrum peak factor as the fitness function for iterative training, optimizing the optimal FIR filter parameters (L, M) of the maximum correlation kurtosis deconvolution MCKD, and obtaining the filtered signal after MCKD denoising, analyzing the signal after MCKD deconvolution through the envelope spectrum, identifying the fault characteristic frequency components in the spectrum, and performing the fault diagnosis of the bearing state, so that the rolling bearing fault diagnosis method based on the IMVO optimized MCKD denoising can effectively eliminate interference, is fast and high in accuracy.

Description

Technical Field

[0001] This invention relates to bearing fault diagnosis technology, and in particular to a rolling bearing fault diagnosis method based on IMVO-optimized MCKD noise reduction. Background Technology

[0002] Currently, in the power generation process of wind turbine units, rolling bearings have a significant impact on the overall safe operation of the power generation equipment and the economic benefits of power generation capacity. Under actual working conditions, rolling bearings are often subjected to corrosion from harsh working environments, causing changes in performance and leading to faults such as abnormal vibration, spalling, pitting, and wear, and even serious accidents such as wind turbine breakage, endangering personal safety. Therefore, research on rolling bearing fault diagnosis is particularly important.

[0003] The minimum entropy deconvolution (MED) denoising method, under the kurtosis principle, has been introduced into the field of rolling bearing fault diagnosis, achieving certain results. However, the effectiveness of MED is greatly reduced when noise completely overwhelms the impact signal. Therefore, based on MED, maximum correlation kurtosis deconvolution (MCKD) is proposed for rolling bearing fault diagnosis in wind turbines. Iteratively, the correlation kurtosis value of the fault signal is maximized, thus eliminating noise interference. However, the deconvolution effect of the MCKD algorithm depends on the selection of filter length and displacement number.

[0004] The selection of parameters in existing MCKD algorithms often relies on human experience. Only by considering the influence of filter length L and shift number M on the deconvolution effect of MCKD and setting reasonable parameter values ​​can the impulse information hidden in the fault signal be accurately extracted, thereby identifying important frequency components.

[0005] The MVO algorithm primarily utilizes the variables Wormhole Existence Rate (WEP) and Wormhole Travel Distance Rate (TDR) to randomly exchange matter between black holes and white holes, maintaining a relatively stable cosmic state and preserving the optimal cosmic position. While the MVO algorithm has the advantage of requiring fewer computational parameters, it also has certain drawbacks. The wormhole travel distance rate (TDR) decreases slowly during the iterative algorithm process, so changes in TDR can affect the overall iteration speed of the MVO program. Summary of the Invention

[0006] The purpose of this invention is to provide a rolling bearing fault diagnosis method based on IMVO-optimized MCKD noise reduction, which can effectively eliminate interference, and is fast and accurate.

[0007] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0008] A rolling bearing fault diagnosis method based on IMVO-optimized MCKD denoising includes the following steps:

[0009] S1. The sensor collects vibration signals during the operation of the rolling bearing of the wind turbine.

[0010] S2. The adaptive improved MVO algorithm wormhole travel distance rate TDR is used to obtain the improved multiverse optimization algorithm IMVO;

[0011] S3. Using the Gini coefficient and envelope peak factor as the comprehensive index GE as the fitness function, iterative training is performed to optimize and obtain the optimal FIR filter parameters (L,M) for the maximum correlation kurtosis deconvolution MCKD.

[0012] The S4 and MCKD algorithms maximize the correlation kurtosis through deconvolution, calculate the deconvolution period parameter T by using the sampling frequency and the fault feature frequency, and obtain the filtered signal after MCKD noise reduction.

[0013] S5. By analyzing the envelope spectrum of the deconvolved signal from MCKD, identify the fault characteristic frequency components in the spectrum and perform fault diagnosis of the bearing condition.

[0014] As a preferred option, step S3 specifically involves:

[0015] S31. Initialize the universe matrix U position: number of universes s, number of matter variables in each universe d. Since each object in each universe corresponds to one (L,M), d=2. Set the upper and lower bounds of the object and the maximum number of iterations H.

[0016] S32. Set the comprehensive index GE as the fitness function of IMVO-MCKD. The function expression of the comprehensive index is GE=GI×Ec, where GI represents the Gini coefficient and Ec represents the envelope peak factor.

[0017] S33. Calculate the cosmic expansion rate NI(Ui) and select a white hole by arranging individual cosmic entities;

[0018] S34. Update the cosmic position based on the roulette wheel mechanism and calculate the wormhole existence rate (WEP) and the improved wormhole travel distance rate (TDR), the expressions of which are:

[0019]

[0020]

[0021] In the formula, WEP min The minimum WEP value is set to 0.2. max The maximum value of WEP is set to 1; h represents the current iteration number; p represents the development precision and is set to 6.

[0022] S35. Determine whether the current value GE is the global optimum. If it is, the program terminates and outputs the optimal solution (L,M). Otherwise, the iteration continues until it is satisfied and terminates.

[0023] As a preferred option, step S4 specifically involves:

[0024] The MCKD algorithm maximizes the correlation kurtosis through deconvolution, highlighting the periodic impulse sequence that is submerged in noise in the fault signal;

[0025] Input the optimal parameters (L,M), calculate the deconvolution period parameter T using the sampling frequency and the fault characteristic frequency, and obtain the filtered signal after MCKD denoising;

[0026] The expression for the correlation kurtosis is:

[0027]

[0028] In the formula: M is the displacement number; T is the period of the pulse impulse signal; y n (n = 0, 1, 2, ..., N) is the fault pulse signal sequence.

[0029] In summary, the present invention has the following beneficial effects:

[0030] An improved MVO algorithm (IMVO) is used to optimize the two-dimensional parameter vector combination (L,M) of the maximum correlation kurtosis deconvolution, achieving adaptive parameter optimization within the search interval. This allows for the acquisition of fault information after MCKD filtering and noise reduction. The algorithm requires fewer computational parameters, has a fast search rate, and high optimization accuracy. It effectively eliminates noise interference during signal transmission. Envelope spectrum analysis is used to identify fault characteristic frequencies, fully eliminating background noise interference in the acquired signal. Further analysis of the spectrum frequency information extracts fault characteristic information, enabling fault diagnosis of bearing condition. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the fault diagnosis process of this method;

[0032] Figure 2 The time-domain waveform of the inner ring fault being tested;

[0033] Figure 3 An iterative graph of the parameter optimization algorithm;

[0034] Figure 4 This is the time-domain waveform after deconvolution;

[0035] Figure 5 This is the envelope spectrum after deconvolution. Detailed Implementation

[0036] The present invention will be further described in detail below with reference to the accompanying drawings.

[0037] According to one or more embodiments, a rolling bearing fault diagnosis method based on IMVO-optimized MCKD denoising is disclosed, such as... Figure 1 As shown, it includes the following steps:

[0038] S1. The sensor collects vibration signals during the operation of the rolling bearing of the wind turbine.

[0039] S2. The adaptive improved MVO algorithm wormhole travel distance rate TDR is used to obtain the improved multiverse optimization algorithm IMVO.

[0040] S3. Using the Gini coefficient and envelope peak factor as the comprehensive index GE as the fitness function, iterative training is performed to optimize and obtain the optimal FIR filter parameters (L,M) for the maximum correlation kurtosis deconvolution MCKD. Specifically:

[0041] S31. Initialize the universe matrix U position: number of universes s, number of matter variables in each universe d. Since each object in each universe corresponds to one (L,M), d=2. Set the upper and lower bounds of the object and the maximum number of iterations H.

[0042] S32. Set the comprehensive index GE as the fitness function of IMVO-MCKD. The function expression of the comprehensive index is GE=GI×Ec, where GI represents the Gini coefficient and Ec represents the envelope peak factor.

[0043] S33. Calculate the cosmic expansion rate NI(Ui) and select a white hole by arranging the individual universes.

[0044] S34. Update the cosmic position based on the roulette wheel mechanism and calculate the wormhole existence rate (WEP) and the improved wormhole travel distance rate (TDR), the expressions of which are:

[0045]

[0046]

[0047] In the formula, WEP min The minimum WEP value is set to 0.2. max is the maximum value of WEP, set to 1; h represents the current iteration number; p represents the development precision, set to 6.

[0048] S35. Determine whether the current value GE is the global optimum. If it is, the program terminates and outputs the optimal solution (L,M). Otherwise, the iteration continues until it is satisfied and terminates.

[0049] The S4 and MCKD algorithms maximize the correlation kurtosis through deconvolution. They calculate the deconvolution period parameter T using the sampling frequency and fault feature frequency, and obtain the filtered signal after MCKD denoising. Specifically:

[0050] The MCKD algorithm maximizes the correlation kurtosis through deconvolution, highlighting the periodic impulse sequences in the fault signal that are submerged by noise.

[0051] Input the optimal parameters (L,M), calculate the deconvolution period parameter T using the sampling frequency and fault characteristic frequency, and obtain the filtered signal after MCKD denoising.

[0052] The expression for the correlation kurtosis is:

[0053]

[0054] In the formula: M is the displacement number; T is the period of the pulse impulse signal; y n (n = 0, 1, 2, ..., N) is the fault pulse signal sequence.

[0055] S5. By analyzing the envelope spectrum of the deconvolved signal from MCKD, identify the fault characteristic frequency components in the spectrum and perform fault diagnosis of the bearing condition.

[0056] This method first adaptively improves the Wormhole Travel Distance Rate (TDR) of the MVO algorithm to enhance its space exploration capabilities, resulting in an improved Multiverse Optimization Algorithm (IMVO). Then, the IMVO algorithm is applied to adaptively determine the optimal FIR filter parameters (L, M) of the MCKD algorithm, avoiding the limitations of relying on prior knowledge and achieving filtering and noise reduction of bearing fault signals. Finally, envelope spectrum demodulation is applied to the deconvolutioned signal of the MCKD algorithm to reduce interference from unimportant frequency components, enabling feature extraction and fault diagnosis of rolling bearing fault signals in wind turbines. The improved TDR function is used to complete the improved Multiverse Optimization Algorithm (IMVO), accelerating the iteration speed of the optimization algorithm. The IMVO algorithm is also used to adaptively optimize the filter length L and displacement number M of the MCKD algorithm, reducing errors caused by manual parameter selection and improving the deconvolution effect of the MCKD algorithm.

[0057] To illustrate this clearly, here is an example:

[0058] The experimental data were obtained from rolling bearing failure data provided by the Bearing Data Center at Case Western Reserve University, USA. This data is available on the center's website: http: / / www.eecs.case.edu / laboratory / bearing / download.htm. The experimental bearing was an SKF6205 model, and the tested bearing had an inner ring failure. The damage diameter was 0.021 inches, achieved through electrical discharge machining. The sampling frequency was set at 12000Hz, the experimental speed at 1730 r / min, and the motor load at 3 horsepower. Based on theoretical formulas, the rotational frequency fr = 29Hz, the inner ring failure frequency fi = 156Hz, and the deconvolution period T = 77. The error between the actual failure frequency and the theoretical failure frequency is generally within 2%. 4096 data points were selected for analysis. (See attached...) Figure 2 The time-domain waveform diagram for an inner ring fault is shown below. Figure 2 As shown, information related to the fault characteristic frequency cannot be directly observed from the figure; the periodic impact component is almost completely submerged by noise. This method is applied to analyze the inner-ring fault signal. The parameters of IMVO are initialized: the number of universes U is set to 30, the maximum number of iterations H = 50, the search interval for the filter length L is [100, 300], and the search interval for the displacement number M is [1, 7]. The IMVO algorithm is used to optimize the parameters of the MCKD algorithm. (See attached...) Figure 3 The iterative graph for the parameter optimization algorithm shows that IMVO reaches its optimal solution (L,M) = (264,7) after 7 iterations with a GE value of 0.0281, MVO reaches its optimal solution after 45 iterations with a GE value of 0.3302, and PSO reaches its optimal solution after 48 iterations with a GE value of 0.8253. Therefore, IMVO has a smaller fitness function value, resulting in faster iteration speed and higher search accuracy. Input the optimal MCKD parameters, and attach... Figure 4 The deconvolutioned time-domain waveform, compared with the original fault signal, shows that noise has been effectively filtered out, and the impact signal components are clearly visible. Envelope spectrum analysis of the MCKD-denoised signal is applied, and fault characteristic frequency information is observed through demodulation. Figure 5 To obtain the envelope spectrum after deconvolution, from Figure 5 The observed frequency is 29.29 Hz and its harmonics, and the inner ring fault characteristic frequency is 155.27 Hz and its harmonics. The actual observed values ​​are very close to the theoretical values. Based on the frequency information, the bearing fault can be determined to be an inner ring fault. In summary, this method can be used for fault diagnosis of rolling bearings in wind turbine units.

[0059] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A rolling bearing fault diagnosis method based on IMVO-optimized MCKD denoising, characterized in that, It includes the following steps: S1. The sensor collects vibration signals during the operation of the rolling bearing of the wind turbine. S2. The adaptive improved MVO algorithm wormhole travel distance rate TDR is used to obtain the improved multiverse optimization algorithm IMVO; S3. Using the Gini coefficient and envelope peak factor as the comprehensive index GE as the fitness function, iterative training is performed to optimize and obtain the optimal FIR filter parameters (L,M) for the maximum correlation kurtosis deconvolution MCKD. The S4 and MCKD algorithms maximize the correlation kurtosis through deconvolution, calculate the deconvolution period parameter T by using the sampling frequency and the fault feature frequency, and obtain the filtered signal after MCKD noise reduction. S5. By analyzing the envelope spectrum of the deconvolved signal from MCKD, identify the fault characteristic frequency components in the spectrum and perform fault diagnosis of the bearing condition. Step S3 is as follows: S31. Initialize the universe matrix U position: number of universes s, number of matter variables in each universe d. Since each object in each universe corresponds to one (L,M), d=2. Set the upper and lower bounds of the object and the maximum number of iterations H. S32. Set the comprehensive index GE as the fitness function of IMVO-MCKD. The function expression of the comprehensive index is GE=GI×Ec, where GI represents the Gini coefficient and Ec represents the envelope peak factor. S33. Calculate the cosmic expansion rate NI(Ui) and select a white hole by arranging individual cosmic entities; S34. Update the cosmic position based on the roulette wheel mechanism and calculate the wormhole existence rate (WEP) and the improved wormhole travel distance rate (TDR), the expressions of which are: ； ； In the formula, WEP min The minimum WEP value is set to 0.

2. max The maximum value of WEP is set to 1; h represents the current iteration number; p represents the development precision, set to 6; S35. Determine whether the current value GE is the global optimum. If it is, the program terminates and outputs the optimal solution (L,M). Otherwise, the iteration continues until it is satisfied and terminates.

2. The rolling bearing fault diagnosis method based on IMVO-optimized MCKD denoising as described in claim 1, characterized in that, Step S4 is as follows: The MCKD algorithm maximizes the correlation kurtosis through deconvolution, highlighting the periodic impulse sequences in the fault signal that are submerged by noise. Input the optimal parameters (L,M), calculate the deconvolution period parameter T using the sampling frequency and the fault characteristic frequency, and obtain the filtered signal after MCKD denoising; The expression for the correlation kurtosis is: ； In the formula: M is the displacement number; T is the period of the pulse impulse signal; y n (n = 0, 1, 2, ..., N) is the fault pulse signal sequence.

Images