CN111238812A

CN111238812A - Self-diagnosis method for rolling bearing retainer fault

Info

Publication number: CN111238812A
Application number: CN202010059189.7A
Authority: CN
Inventors: 闫柯; 康伟; 朱永生; 洪军; 袁倩倩; 刘煜炜; 任智军
Original assignee: Xian Jiaotong University
Current assignee: Xian Jiaotong University
Priority date: 2020-01-19
Filing date: 2020-01-19
Publication date: 2020-06-05
Anticipated expiration: 2040-01-19
Also published as: CN111238812B

Abstract

A self-diagnosis method for the failure of rolling bearing cage includes such steps as quickly filtering vibration signals by high-precision quick filter set, choosing optimal signal according to the spectral kurtosis of mean-square envelope autocorrelation signal of each filtered signal, calculating its envelope spectrum, automatically choosing the first M-order actual failure characteristic frequency according to the theoretical failure characteristic frequency of cage, automatically finding out the threshold meeting probability requirement, and calculating the global test index and failure incidence rate of cyclic stability.

Description

Self-diagnosis method for rolling bearing retainer fault

Technical Field

The invention relates to the technical field of fault diagnosis and signal processing analysis, in particular to a self-diagnosis method for a rolling bearing retainer fault.

Background

The cage is used as an important part of the rolling bearing, and the normal operation of the rolling bearing is seriously influenced by the fault of the cage. In order to diagnose the fault of the retainer in time, the diagnosis can be completed by analyzing the vibration characteristics of the retainer. However, the cage fault signal is usually weak and easily submerged in the background noise, and it is necessary to perform relevant signal processing to reduce noise interference. In addition, the conventional diagnosis usually requires expert decision, the diagnosis result is closely related to the expert experience, and there is a problem of time cost. Therefore, it is necessary to develop a self-diagnosis method of cage failure.

The high-precision fast filter bank comprises a plurality of layers of filter banks, each layer of filter bank comprises a plurality of filters with different center frequencies and the same resolution, and multi-band filtering of vibration signals can be achieved. The filter bank is constructed based on a group of quasi-analytic low-pass filters and high-pass filters, and an 1/3-binary tree structure form is used, so that the precision can be improved while the filtering is fast. The bearing fault diagnosis method based on the rapid kurtosis graph is a classic fault feature extraction method, the filter bank is used when signals are decomposed, proper filtered signals are selected by directly utilizing the spectral kurtosis of the filtered signals to extract fault features, and the method can effectively extract the fault features when the fault features are not interfered. However, since the kurtosis is extremely sensitive to the impact characteristics, once abnormal impact interference exists in the vibration signal, a weak fault signal of the retainer is difficult to extract. The autocorrelation analysis is used for quantitatively determining the similarity degree of the original signal and the time delay signal thereof, and can be used for extracting the periodic characteristics in the signal and overcoming the influence of interference noise such as abnormal impact. But this method alone is susceptible to introducing periodic interference in the signal. In addition, after the rolling bearing retainer fault characteristic signal is extracted, the diagnosis of the rolling bearing retainer fault characteristic signal is usually identified by experts, which is not favorable for real-time online diagnosis. In order to overcome the problem, the theoretical fault characteristic frequency of the retainer can be used for identification, but due to the influence of the slip factor, the actual fault characteristic frequency is often not equal to the theoretical fault characteristic frequency, and the self-diagnosis is difficult to effectively complete. And because the acquired signal is interfered by noise, the effect of directly carrying out retainer fault self-diagnosis on the frequency spectrum of the original signal is poor. Therefore, it is necessary to enhance the weak cage fault signal in the signal and then automatically identify the actual fault characteristic frequency to self-diagnose the cage fault.

Disclosure of Invention

In order to overcome the defects of the prior art, the invention aims to provide a self-diagnosis method for the faults of the rolling bearing retainer, vibration signals are quickly filtered through a high-precision quick filter group, an optimal signal is selected according to the spectral kurtosis of the mean square envelope autocorrelation signal of each filtered signal, the envelope spectrum of the optimal signal is calculated, the actual fault characteristic frequency of the previous M orders is automatically selected according to the theoretical fault characteristic frequency of the retainer, the threshold meeting the probability requirement is automatically found out according to the statistical characteristic of the frequency spectrum, and finally the circular stability global test index and the fault occurrence rate are calculated, so that the problem that the fault of the rolling bearing retainer depends on the decision of an expert during fault diagnosis is solved, and the self-diagnosis of the faults of the rolling bearing.

In order to achieve the purpose, the technical scheme of the invention is as follows:

a self-diagnosis method for rolling bearing retainer faults comprises the following specific steps:

step1, acquiring a vibration acceleration signal s (t) of the rolling bearing;

step2, filtering the signals by using a high-precision fast filter group with a structure of 1/3-binary tree form to obtain N layers of filtered signals s_ni(t), n denotes an nth layer filter bank, i denotes an ith filter of the nth layer filter bank, and i is 2, …, 2ⁿ-1；

Step3, filtering the signals s obtained from Step2 according to the spectral kurtosis_ni(t) selecting the optimum signal, in particular the pair signal s_ni(t) performing mean square envelope demodulation, calculating the spectral kurtosis of the autocorrelation function of the demodulated signal, comparing the magnitudes, wherein the corresponding signal with the maximum value is the optimal signal, and then extracting the mean square envelope spectrum of the signaly(f)；

Step4, and using theoretical fault characteristic frequency f of rolling bearing retainer_theoryCentered, using a narrow band [ b ]_lb_h]Finding out the frequency corresponding to the maximum spectral line of the mean square envelope spectrum y (f), determining the frequency as the 1 st order actual fault characteristic frequency, then finding out the 2 nd order actual fault characteristic frequency near the double actual fault characteristic frequency by utilizing the narrow bands with the same bandwidth, finally determining the center frequency for the mth order actual fault characteristic frequency according to the previous M-1 order actual fault characteristic frequency, and finding out the mth order fault characteristic frequency by utilizing the narrow bands with the same bandwidth, wherein M is 1, 2, …, M, and M is the maximum frequency multiple for extracting the fault characteristic frequency;

step5, calculating a periodic statistical chart of the envelope spectrum y (f) obtained at Step3 to determine a threshold value a meeting the probability P; then using the amplitude y (f) of the characteristic frequency of the previous M-order fault_m) Calculating the fault occurrence rate ppf with a threshold value a, wherein the calculation formula is shown in formula (1), meanwhile, automatically finding out spectral lines z (f) exceeding the threshold value by comparing the size of the envelope spectrum y (f) with the threshold value a, K is 1, …, K is the maximum spectral line number exceeding the threshold value, and calculating the cyclostationary global test index c according to the spectral lines, wherein the calculation formula is shown in formula (2):

and Step6, comparing the cyclostationary global test index c obtained in the Step5 with the fault occurrence rate ppf, if c is smaller than ppf, diagnosing that the rolling bearing retainer is normal, otherwise, diagnosing that the rolling bearing retainer is fault.

The invention has the beneficial effects that: the method has the advantages that the original signals are filtered through the high-precision rapid filter bank, the filtering of multiple frequency bands is rapidly achieved, the optimal signals are screened by utilizing the mean square envelope autocorrelation spectrum kurtosis, the interferences such as abnormal impact are weakened, the potential fault characteristics of the retainer are enhanced, the threshold meeting the requirements is automatically calculated through the histogram, the circular stability global test index and the fault occurrence rate are automatically calculated and compared, and the diagnosis of the rolling bearing retainer fault is given when the circular stability global test index is smaller than the fault occurrence rate. The method is beneficial to automatically diagnosing the faults of the retainer of the rolling bearing and is suitable for fault automatic diagnosis systems of intelligent bearings and the like.

Drawings

FIG. 1 is a flow chart of the method steps of the present invention.

Fig. 2 is a raw vibration signal in an embodiment of the present invention.

Fig. 3 is a spectrum of fig. 2 obtained by mean square envelope analysis in accordance with the present invention.

Fig. 4 is a spectrum of fig. 2 according to the present invention obtained by the method of the present invention.

Detailed Description

The invention is described in detail below with reference to the drawings and the detailed description.

As shown in fig. 1, a self-diagnosis method for rolling bearing cage failure comprises the following specific steps:

the specific parameters are as follows: 1) the test bed consists of a driving electric main shaft and a mechanical shaft, the rolling ball bearing is ZYS B7008C type, the contact angle is 0 degree, the outer diameter of the bearing is 68mm, the inner diameter of the bearing is 40mm, the diameter of the rolling bodies is 7.138mm, and the number of the rolling bodies is 18; 2) the fault type of the retainer is that the pocket cross beam is broken and communicated; 3) the sampling frequency in the acquisition embodiment is 51.2KHz, and the rotating speed of the rotating shaft is 1500 r/min.

Step1, acquiring a vibration acceleration signal at a base of a fault end of a mechanical shaft retainer by using a three-shaft acceleration sensor, selecting a radial z-shaft acceleration signal for analysis, as shown in fig. 2, judging the retainer fault of a rolling bearing through a time domain difficultly, analyzing the retainer fault by using a more effective mean square envelope analysis method during the rolling bearing analysis, and obtaining an envelope spectrum as shown in fig. 3, wherein the characteristic frequency and the frequency multiplication of the retainer fault corresponding to a dot-dash line are not obvious;

step2, filtering the vibration signal by using a 5-layer high-precision fast filter group with a structure of 1/3 binary tree form to obtain a series of filtered signals s_ni(t), n representsAn nth layer filter bank, i denotes an ith filter of the nth layer filter bank, and i is 2, …, 2ⁿ-1；

Step3, filtering the signals s obtained from Step2 according to the spectral kurtosis_ni(t) selecting the optimum signal, in particular the pair signal s_ni(t) performing mean square envelope demodulation, calculating the spectral kurtosis of the autocorrelation function of the demodulated signal, and comparing the magnitudes, wherein the maximum spectral kurtosis value is 19.9, the center frequency of the corresponding filter is 13600Hz, and the frequency bandwidth is 1600Hz, and the corresponding signal is the optimal signal. Then, extracting a mean square envelope spectrum y (f) of the signal, wherein as shown in fig. 4, the retainer fault characteristic frequency and frequency multiplication thereof corresponding to the dot-dash line are obvious, and the retainer fault characteristic is effectively extracted;

step4, taking the first 3-order characteristic frequency to carry out fault diagnosis, taking the theoretical fault characteristic frequency 10.8477Hz of the rolling bearing retainer as the center, finding out the frequency corresponding to the maximum spectral line by utilizing the narrow band with the bandwidth 0.1 times of the theoretical characteristic frequency to obtain the 1 st-order actual fault characteristic frequency of 10.8458Hz, then finding out the 2 nd-order actual fault characteristic frequency near the double actual fault characteristic frequency by utilizing the narrow band with the same bandwidth to be 21.7893Hz, finally determining the center frequency for the 3 rd-order actual fault characteristic frequency according to the first two-order actual fault characteristic frequencies, and finding out the 3 rd-order fault characteristic frequency of 32.6351Hz by utilizing the narrow band with the same bandwidth;

step5, calculating a periodic statistical chart of the envelope spectrum y (f) obtained in Step3, and determining that the threshold value meeting the probability of 90% is 0.0020; the fault occurrence rate ppf is then calculated as 0.8453 according to equation (1) using the amplitude of the first 3 order fault signature frequencies and the threshold. Meanwhile, spectral lines exceeding the threshold value are automatically found out by comparing the envelope spectrum with the threshold value, and the cyclostationary global test index is calculated by using the formula (2) and is 0.4345.

And Step6, comparing the cyclostationary global test index 0.4345 obtained in the Step5 with the fault occurrence rate 0.8453, wherein the cyclostationary global test index is smaller than the fault occurrence rate, and self-diagnosing that the rolling bearing retainer has faults and conforms to the retainer faults existing in the test bearing.

While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.

Claims

1. A self-diagnosis method for a rolling bearing cage fault is characterized in that: the method comprises the following specific steps:

step1, acquiring a vibration acceleration signal s (t) of the rolling bearing;

Step3, filtering the signals s obtained from Step2 according to the spectral kurtosis_ni(t) selecting the optimum signal, in particular the pair signal s_ni(t) performing mean square envelope demodulation, calculating the spectral kurtosis of an autocorrelation function of a demodulated signal, comparing the magnitudes, wherein the corresponding signal with the maximum value is the optimal signal, and then extracting the mean square envelope spectrum y (f) of the signal;

step4, and using theoretical fault characteristic frequency f of rolling bearing retainer_theoryCentered, using a narrow band [ b ]_lb_h]Finding out the frequency corresponding to the maximum spectral line of the mean square envelope spectrum y (f), determining the frequency as the 1 st order actual fault characteristic frequency, then finding out the 2 nd order actual fault characteristic frequency near the double actual fault characteristic frequency by utilizing the narrow bands with the same bandwidth, finally determining the center frequency for the mth order actual fault characteristic frequency according to the previous m-1 order actual fault characteristic frequency, and finding out the mth order fault characteristic frequency by utilizing the narrow bands with the same bandwidth, wherein m is 1, 2, …M, M is the maximum frequency multiplication number of the extracted fault characteristic frequency;

step5, calculating a periodic statistical chart of the envelope spectrum y (f) obtained at Step3 to determine a threshold value a meeting the probability P; then using the amplitude y (f) of the characteristic frequency of the previous M-order fault_m) Calculating fault incidence ppf with a threshold value a, wherein the calculation formula is shown as a formula (1), and simultaneously automatically finding out spectral lines z (f) exceeding the threshold value by comparing the envelope spectrum y (f) with the threshold value a_k) K is 1, …, K is the maximum spectral line number exceeding the threshold, and then the cyclostationary global test index c is calculated according to the maximum spectral line number, and the calculation formula is shown as formula (2):