Fault diagnosis method for rolling bearing data enhancement spectrum amplitude modulation

Abstract

The invention provides a rolling bearing fault diagnosis method, which comprises the following steps of: firstly, performing Fourier transform on a bearing vibration signal to obtain an amplitude A (f) and a phase phi (f) of an original signal, keeping the phase information phi (f) unchanged, and endowing different weight indexes MO to the amplitude A (f) to obtain A (f) MO (MO is more than or equal to-0.5 and less than or equal to 1.5); then combining the spectrum amplitudes A (f) MO under different MO with the original phase phi (f) to obtain an editing spectrum, performing inverse Fourier transform on the editing spectrum to obtain a series of correction signals, and performing improved wavelet transform processing on the square envelope of the correction signals to further reduce noise interference; and finally, calculating a logarithm maximum square envelope spectrum of the noise reduction signal, and performing normalization processing (eliminating signal amplitude influence) to obtain data enhanced spectrum amplitude modulation. And the fault frequency is calculated through the given data, and diagnosis is completed.

Description

Technical Field

[0001] The present invention relates to a fault diagnosis method, in particular to a rolling bearing data enhanced spectrum amplitude modulation fault diagnosis method. Background Art

[0002] Rolling bearings are widely used in rotating machinery and are among the most prone to failure. Their operating environments are relatively complex and harsh, and failures can cause serious accidents and even casualties. Statistics show that 40%-45% of rotating machinery failures are due to rolling bearing failure. Therefore, the importance of rolling bearing fault detection is self-evident. Precisely because of the complexity of these operating conditions and the interference of various noises, fault detection is extremely difficult, and early fault identification is difficult. Therefore, identifying subtle faults under complex conditions is of paramount importance.

[0003] Among existing rolling bearing fault detection methods, spectral amplitude modulation (SAM), a classic nonlinear filtering method, is widely used in the identification of bearing fault characteristic frequencies. This method performs a Fourier transform on the acquired vibration signal to obtain its amplitude and phase spectra. The phase spectrum is then assigned different weighting indices (magnitude orders, MOs). The resulting edited spectrum is then inverse Fourier transformed to produce a series of corrected signals. Finally, the squared spectra of the corrected signals are Fourier transformed to obtain the squared envelope spectrum, which provides fault characteristic information. SAM is both efficient and computationally simple, achieving significant results in fault diagnosis. However, its effectiveness is significantly reduced because it amplifies both the fault component and the interference component, significantly reducing its effectiveness. Building on this, this paper proposes a data-enhanced spectral amplitude modulation method. By applying a modified wavelet transform to the squared envelope of the corrected signal, it reduces the influence of interference components and further enhances the fault periodicity component. Summary of the Invention

[0004] The main technical problem to be solved by the present invention is that when the vibration signal contains high-intensity noise interference, the accuracy of SAM in identifying the fault characteristic frequency is greatly reduced. Therefore, the present invention provides a rolling bearing fault diagnosis method. The method first performs Fourier transform on the bearing vibration signal to obtain the amplitude A(f) and phase Φ(f) of the original signal, keeps the phase information Φ(f) unchanged, and assigns different weight exponents MO to the amplitude A(f) to obtain A(f). MO , where -0.5≤MO≤1.5; then the spectrum amplitude A(f) under different MO MOCombined with the original phase Φ(f), the edited spectrum is obtained, which is then inverse Fourier transformed to obtain a series of corrected signals. The square envelope of the corrected signal is processed by improved wavelet transform to further reduce noise interference. Finally, the logarithmic maximum square envelope spectrum of the noise reduction signal is calculated and normalized (to eliminate the influence of signal amplitude) to obtain data-enhanced spectrum amplitude modulation. The fault frequency is calculated using the given data to complete the diagnosis.

[0005] The technical solution adopted by the present invention to solve the above technical problems is: a fault diagnosis method for rolling bearing data enhancement spectrum amplitude modulation, comprising the following steps:

[0006] Step 1: First, define x(t) as the original input signal. Then, transform the vibration signal from the time domain to the frequency domain through Fourier transform. According to the formula X(f) = FT{x(t)} = A(f)e jΦ(f) Get its phase Φ(f) and amplitude A(f).

[0007] Step 2: Phase remains unchanged, through the formula x m (t,MO)=IFT{A(f) MO e jΦ (f)}, different weight indices are assigned to the amplitude spectrum to adjust the amplitude in different frequency bands, and then the original phase is combined with the modulated amplitude to obtain the edited spectrum, and finally the inverse Fourier transform is performed to obtain the corrected signal.

[0008] Step 3: Select a suitable MO, with a value of -0.5≤MO≤1.5 and a step size of 0.1. The advantage of this is that the size of MO can be adjusted to diagnose faults in different situations and effectively shield interference components.

[0009] Step 4: Pass The square envelope of the corrected signal is processed by improved wavelet transform.

[0010] Step 5: Pass Perform wavelet decomposition.

[0011] Step 6: Give σ n definition

[0012] Step 7: Use the formula log{max[SES{x m (t,MO)}]}=|IWT{|A{x m (t,MO)}| 2 Generate a modified logarithmic maximum square envelope spectrum instead of the original square envelope spectrum.

[0013] Step 8: Pass Standardize the SES.

[0014] Compared with the traditional method, this method processes the squared envelope of the correction signal through improved wavelet transform to reduce the influence of interference components and further enhance the fault cycle components. By highlighting the signal characteristics and strengthening the feature information, this method can minimize the interference information, thereby improving the overall robustness. Therefore, the method of the present invention can effectively improve the accuracy of rolling bearing fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of the implementation of the method of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

[0016] The present invention will be described in detail below with reference to the accompanying drawings.

[0017] As Figure 1 shown, the present invention relates to a rolling bearing fault diagnosis method, and the specific implementation steps of this method are as follows:

[0018] Step 1: First, define x(t) as the original input signal, and convert the vibration signal from the time domain to the frequency domain through Fourier transform to obtain its phase Φ(f) and amplitude A(f), as shown in Equation (1). X(f) = FT{x(t)} = A(f)e jΦ(f) (1) where j is the imaginary unit and FT is the Fourier transform.

[0019] Step 2: Keep the phase unchanged, assign different weight exponents to the amplitude spectrum to adjust the amplitudes in different frequency bands, then combine the original phase with the modulated amplitude to obtain the edited spectrum, and finally perform inverse Fourier transform to obtain the correction signal. This non-linear filtering process is shown in Equation (2). x m (t, MO) = IFT{A(f) MO e jΦ(f)} (2)

[0020] Step 3: Select a suitable MO, whose value ranges from -0.5 ≤ MO ≤ 1.5 with a step size of 0.1. The advantage of this is that the size of MO can be adjusted for fault diagnosis in different situations and effectively shield interference components. The effects of the MO value on the signal are mainly the following three: When MO < 0, it will amplify the relatively small amplitude parts in the original signal, and this range is most suitable when the noise interference is small; when 0 < MO < 1, this range will weaken the components with larger amplitudes and strengthen the parts with smaller amplitudes, but still retain the relative levels of each component; when MO > 1, the components with larger amplitudes will overwhelm the components with smaller amplitudes, and this range is most suitable when the noise interference is large.

[0021] Step 4: Perform improved wavelet transform on the square envelope of the corrected signal, where the improved wavelet threshold function expression is as shown in formula (3). where ω n is the wavelet coefficient of the vibration signal decomposed at the nth layer; n is the threshold of each layer’s wavelet coefficient. n |≥λ n In order to simulate the real environment, a decaying sine function is given to reduce noise; when |ω n |<λ n When , the noise ratio is large, so the exponential noise reduction method is used. u is the threshold function adjustment factor, which is determined by calculating the mean and standard deviation of the wavelet coefficients and then adding or subtracting a multiple of the standard deviation from the mean. The value range is (0, 1], and ξ is a constant of 0.02.

[0022] Step 5: Let x(t) be the vibration signal and perform wavelet decomposition. The threshold corresponding to each layer of wavelet coefficients is calculated as follows (4). where σ n is the noise standard deviation of the nth wavelet coefficient; N is the signal length.

[0023] Step 6: Give σ n The definition is shown in (5). (5) The threshold value decreases as the decomposition scale increases, which satisfies the characteristic that the wavelet coefficient of the vibration signal decreases as the decomposition scale increases, and has a better noise reduction effect.

[0024] Step 7: Use the modified logarithmic maximum square envelope spectrum to replace the original square envelope spectrum, as shown in formula (6). log{max[SES{x m (t,MO)}]}=|IWT{|A{x m (t,MO)}| 2}| (6) The IWT is the improved wavelet transform.

[0025] Step 8: Finally, the SES is normalized between [0, 1] for explicit comparison, as in (7).

Claims

1. A fault diagnosis method for rolling bearing data enhancement spectrum amplitude modulation, characterized in that: The following steps are involved: Step 1: First, define x(t) as the original input signal. Convert the vibration signal from the time domain to the frequency domain through Fourier transform to obtain its phase Φ(f) and amplitude A(f), as shown in formula (1). X(f)=FT[x(t)}=A(f)e jΦ(f) (1) Where j is the imaginary unit, FT is the Fourier transform, Step 2: The phase remains unchanged, and different weight indices are assigned to the amplitude spectrum to adjust the amplitude in different frequency bands. The original phase is then combined with the modulated amplitude to obtain the edited spectrum. Finally, an inverse Fourier transform is performed to obtain the corrected signal. This nonlinear filtering process is shown in formula (2). x m (t,MO)=IFT{A(f) MO e jΦ(f) } (2) Step 3: Select a suitable MO, with a value of -0.5≤MO≤1.5 and a step size of 0.

1. Step 4: Perform improved wavelet transform on the square envelope of the corrected signal, where the improved wavelet threshold function expression is as shown in formula (3): where ω n is the wavelet coefficient of the vibration signal decomposed at the nth layer; n is the threshold value of each layer’s wavelet coefficients, u is the threshold function adjustment factor, which is determined by calculating the mean and standard deviation of the wavelet coefficients and then adding or subtracting a multiple of the standard deviation from the mean value. The value range is (0,1], ξ is a constant of 0.02, Step 5: Let x(t) be the vibration signal and perform wavelet decomposition. The threshold calculation formula (4) corresponding to each layer of wavelet coefficients is as follows: where σ n is the standard deviation of the noise of the nth wavelet coefficient; N is the signal length, Step 6: Give σ n The definition is as shown in (5), (5) The threshold value decreases as the decomposition scale increases, which satisfies the characteristic that the wavelet coefficient of the vibration signal decreases as the decomposition scale increases, and has a better noise reduction effect. Step 7: Use the modified logarithmic maximum square envelope spectrum to replace the original square envelope spectrum, as shown in formula (6). log{max[SES{x m (t,MO)}]}=|IWT{|A{x m (t,MO)}| 2 }| (6) Where IWT is improved wavelet transform, Step 8: Finally, the SES is normalized between [0, 1] for explicit comparison, as shown in (7),

Images