A rolling bearing fault diagnosis method based on exponentially weighted synchronous averaging

Abstract

The invention discloses a rolling bearing fault diagnosis method based on exponential weighted synchronous averaging. The method comprises the following steps: firstly, performing Hilbert transform on the collected rolling bearing vibration signal, performing modulo and de-averaging operations, obtaining an envelope signal of the vibration signal, segmenting the envelope signal according to the fault cycle, constructing a quasi-average signal, solving the autocorrelation function of each segment of the quasi-average signal, further calculating the envelope harmonic noise ratio, assigning a weight to each segment according to the size of the envelope harmonic noise ratio, reconstructing the signal using the weighted quasi-average signal, and finally performing spectrum analysis on the filtered signal. The fault characteristic frequency and its harmonic frequency can be clearly observed from the spectrum, and the rolling bearing fault can be finally identified. The reconstructed signal obtained by the invention has obvious time domain impact, and its spectrum resolution is higher than that of the original synchronous averaging method, and the envelope harmonic noise ratio is effectively used as an indicator to enhance the noise reduction effect.

Description

Technical Field

[0001] The invention relates to the technical field of mechanical equipment fault diagnosis, and in particular to a rolling bearing fault diagnosis method based on exponential weighted synchronous averaging. Background Art

[0002] Vibration signal analysis is one of the main methods for monitoring and diagnosing mechanical equipment. Damage and fault information during the operation of mechanical equipment can often be reflected in the vibration signal. Since mechanical equipment contains a large number of rotating parts such as rolling bearings and gears, these parts have a symmetrical physical structure and their kinematic laws are also periodic. Therefore, when performing uniform rotational motion, their vibration response will show significant periodicity. In addition, there is non-periodic background noise in the environment where the monitoring system is located. Therefore, the rolling bearing vibration signal collected by the condition monitoring system can be regarded as a signal composed of the superposition of periodic vibration response and background noise. If the background noise is strong, the periodic vibration response signal will be buried in the noise, resulting in the failure information cannot be accurately captured.

[0003] Time domain synchronous averaging is a common periodic feature enhancement technology. This method divides the original measurement signal into time domain periodic segments according to the period, sums each segment and then averages it. Since the background noise in the signal is not periodic and the mean is zero, the noise component in the obtained signal will be greatly attenuated, thereby achieving the goal of suppressing background noise and highlighting the relevant vibration components. Time domain synchronous averaging is a denoising method with low computational complexity and good effect, but it has the following disadvantages: since this method averages the signal segments after superposition, the result only contains a signal with a period length, so it no longer has the periodicity of the original signal and the periodic shock can no longer be observed; after Fourier transforming the result signal, the signal spectrum resolution is very low, and it is difficult to smoothly observe the harmonic frequency of the fault shock; in addition, if the noise mean is not zero, the effect of the time domain averaging method will be reduced. Summary of the invention

[0004] In order to overcome the above-mentioned existing shortcomings, the purpose of the present invention is to provide a rolling bearing fault diagnosis method based on exponentially weighted synchronous averaging, which can better eliminate background noise and enhance rolling bearing fault characteristics.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A rolling bearing fault diagnosis method based on exponentially weighted synchronous averaging comprises the following steps:

[0007] Step 1: Use an acceleration sensor to collect the vibration signal of the rolling bearing, perform mean value processing on the vibration signal, and record the processed vibration signal as x(n);

[0008] Step 2: Perform Hilbert transform on the signal x(n) according to formula (1) to obtain its analytical signal Then, according to formula (2), modulus and mean removal operations are performed to obtain the envelope signal e(n) of the vibration signal;

[0009]

[0010]

[0011] Among them, τ is the offset, mean(·) is the mean operation;

[0012] Step 3: Divide the envelope signal e(n) into N segments according to the fault period T. The number of sampling points in each envelope signal period is T. s Each envelope signal and the following M segments together constitute a quasi-average signal a i (n);

[0013] a i (n) = e(n + iT s ), n∈(0, MT s ), i∈(0,NM) (3)

[0014] Step 4: Quasi-average signal a for each segment i (n) Obtain the autocorrelation function r according to formula (4) i (τ);

[0015]

[0016] Step 5: Calculate the envelope harmonic noise ratio EHNR of each segment of the quasi-average signal according to formula (5): i ;

[0017]

[0018] Among them, τ max The offset corresponding to the maximum value of the autocorrelation function, r i (τ max ) is the maximum value of the autocorrelation function, indicating the total harmonic energy, r i (0) is the position of the zero point of the autocorrelation function, which represents the total energy of the envelope signal;

[0019] Step 6: Substitute the quasi-average signal a for each segment i (n) According to formula (6), the average signal of each segment is obtained by synchronous averaging, and the exponential weight w of each segment is calculated according to formula (7) using the envelope harmonic noise ratio of each segment i ;

[0020]

[0021]

[0022] Step 7: According to formula (8), weights are assigned to the average signals of each segment for reconstruction to obtain the reconstructed signal of the exponentially weighted synchronous averaging method:

[0023]

[0024] Step 8: Reconstruct the signal The frequency spectrum is obtained by Fourier transform, and the fault characteristic frequency and its harmonic frequency can be clearly observed on the spectrum, and the rolling bearing fault is finally identified.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] a) Compared with the traditional time-domain synchronous averaging method, the waveform of the result signal (reconstructed signal) obtained by the present invention has periodicity, which can better observe the fault impact.

[0027] b) Compared with the traditional time-domain synchronous averaging method, the results of the present invention are subjected to spectrum analysis, and the spectral line resolution is higher and the recognition is better.

[0028] c) The present invention uses the envelope harmonic-to-noise ratio as an indicator to perform exponential weighting on each signal segment, so that the signal segment with significant fault impact components has a higher proportion, thereby producing a better noise reduction effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 Schematic diagram of a bearing test bench according to an embodiment.

[0030] Figure 2 Flow chart of the method of the present invention.

[0031] Figure 3 is the vibration signal x(n) of the embodiment.

[0032] Figure 4 is the envelope signal e(n) of the vibration signal of the embodiment.

[0033] Figure 5 is the reconstructed signal obtained after being processed by the exponentially weighted synchronous averaging method

[0034] Figure 6 is the spectrum of the reconstructed signal obtained after processing by the exponentially weighted synchronous averaging method.

[0035] Figure 7 It is the spectrum obtained by directly enveloping the vibration signal and then performing Fourier transform.

[0036] Figure 8The waveform of the signal is obtained after the vibration signal envelope is denoised by traditional synchronous averaging.

[0037] Fig. 9 The spectrum of the signal is obtained after the vibration signal envelope is denoised by traditional synchronous averaging. DETAILED DESCRIPTION

[0038] The present invention is described in detail below in conjunction with embodiments and drawings.

[0039] The embodiment adopts a bearing test bench, such as Figure 1 As shown in the figure, the experimental platform consists of a driving motor, an experimental bearing, a support bearing and a support shaft, among which the experimental bearing is a bearing with an outer ring fault. The system movement mode is direct drive by the driving motor, which drives the support shaft to rotate, and then drives the inner ring of the experimental bearing to move. The acceleration sensor is fixed on the outside of the experimental bearing to measure the vibration signal of the experimental bearing.

[0040] The experimental bearing in the test bench is diagnosed using the exponentially weighted synchronous averaging method. The vibration signal measured by the acceleration sensor is analyzed and compared with the time domain synchronous averaging method and the envelope demodulation method.

[0041] like Figure 2 As shown, a rolling bearing fault diagnosis method based on exponential weighted synchronous averaging includes the following steps:

[0042] Step 1: Use the acceleration sensor to collect the rolling bearing vibration signal and perform mean processing on the vibration signal. The sampling frequency is 25600 Hz and the signal duration is 1.28 s. The processed vibration signal x(n) is as follows: Figure 3 As shown;

[0043] Step 2: Perform Hilbert transform on the signal x(n) according to formula (1) to obtain its analytical signal Then, according to formula (2), the modulus and mean removal operations are performed to obtain the envelope signal e(n) of the vibration signal, as shown in Figure 4 As shown;

[0044]

[0045]

[0046] Among them, τ is the offset, mean(·) is the mean operation;

[0047] Step 3: Divide the envelope signal e(n) into N = 1.28 / 0.0093 = 138 segments according to the fault period T = 0.0093s. The number of sampling points in each envelope signal period is T. s =25600×1.28 / 138=237, each envelope signal and the following M=30 segments together constitute a quasi-average signal ai (n);

[0048] a i (n) = e(n + iT s ), n∈(0, MT s ), i∈(0,NM) (3)

[0049] Step 4: Quasi-average signal a for each segment i (n) Obtain the autocorrelation function r according to formula (4) i (τ);

[0050]

[0051] Step 5: Calculate the envelope harmonic noise ratio EHNRi of each segment of the quasi-average signal according to formula (5);

[0052]

[0053] Among them, τ max The offset corresponding to the maximum value of the autocorrelation function, r i (τ max ) is the maximum value of the autocorrelation function, indicating the total harmonic energy, r i (0) is the position of the zero point of the autocorrelation function, which represents the total energy of the envelope signal;

[0054] Step 6: Substitute the quasi-average signal a for each segment i (n) According to formula (6), the average signal of each segment is obtained by synchronous averaging, and the exponential weight w of each segment is calculated according to formula (7) using the envelope harmonic noise ratio of each segment i ;

[0055]

[0056]

[0057] Step 7: According to formula (8), weights are assigned to the average signals of each segment for reconstruction to obtain the reconstructed signal of the exponentially weighted synchronous averaging method: like Figure 5 As shown in the figure, it can be found that there are a lot of obvious periodic impact components in the signal;

[0058]

[0059] Step 8: Verification Figure 5 The relationship between the medium-cycle shock and the fault characteristics, and the reconstruction signal The spectrum is obtained by Fourier transform, such as Figure 6 As shown, the characteristic frequency of the outer ring fault and its harmonics can be clearly observed on the spectrum, and the outer ring fault of the rolling bearing can be finally identified.

[0060] like Figure 7 As shown, Figure 7 The spectrum is obtained by directly enveloping the vibration signal and then performing Fourier transform. It can be seen that the spectrum contains a large number of irrelevant frequency interference components, and the characteristic frequencies of each order of fault are not obvious; Figure 8 As shown, Figure 8 This is the time domain waveform obtained after the vibration signal envelope is de-noised by traditional synchronous averaging. It can be seen that the signal only contains one cycle, and it is difficult to observe the periodic fault impact; Fig. 9 As shown, Fig. 9 for Figure 8 From the corresponding spectrum signal, it can be seen that the frequency resolution is low and it is difficult to accurately identify the characteristic frequency. From the above analysis, it can be seen that the method of the present invention is more advantageous than the traditional synchronous averaging method and the traditional envelope demodulation method, and provides a good tool for rolling bearing fault diagnosis.

Claims

1. A rolling bearing fault diagnosis method based on exponentially weighted synchronous averaging, characterized in that: The following steps are involved: Step 1: Use an acceleration sensor to collect the vibration signal of the rolling bearing, perform mean value processing on the vibration signal, and record the processed vibration signal as x(n); Step 2: Perform Hilbert transform on the signal x(n) according to formula (1) to obtain its analytical signal Then, according to formula (2), modulus and mean removal operations are performed to obtain the envelope signal e(n) of the vibration signal; Among them, τ is the offset, mean(·) is the mean operation; Step 3: Divide the envelope signal e(n) into N segments according to the fault period T. The number of sampling points in each envelope signal period is T. s Each envelope signal and the following M segments together constitute a quasi-average signal a i (n); a i (n)=e(n+iT s ),n∈(0,MT s ),i∈(0,N-M) (3) Step 4: Quasi-average signal a for each segment i (n) Obtain the autocorrelation function r according to formula (4) i (τ); Step 5: Calculate the envelope harmonic noise ratio EHNR of each segment of the quasi-average signal according to formula (5): i ; Among them, τ max The offset corresponding to the maximum value of the autocorrelation function, r i (τ max ) is the maximum value of the autocorrelation function, indicating the total harmonic energy, r i (0) is the position of the zero point of the autocorrelation function, which represents the total energy of the envelope signal; Step 6: Substitute the quasi-average signal a for each segment i (n) According to formula (6), the average signal of each segment is obtained by synchronous averaging, and the exponential weight w of each segment is calculated according to formula (7) using the envelope harmonic noise ratio of each segment i ; Step 7: According to formula (8), weights are assigned to the average signals of each segment for reconstruction to obtain the reconstructed signal of the exponentially weighted synchronous averaging method: Step 8: Reconstruct the signal The frequency spectrum is obtained by Fourier transform, and the fault characteristic frequency and its harmonic frequency can be clearly observed on the spectrum, and the rolling bearing fault is finally identified.

Images