A method for denoising mechanical seal acoustic emission signals based on new LCD-wavelet threshold

Through the combination of local feature scale decomposition and new wavelet threshold, the problem of noise interference in rotary mechanical sealed acoustic emission signal detection is solved, effective signal noise reduction and fault diagnosis are achieved, and signal purity and diagnostic accuracy are improved.

CN114858430BActive Publication Date: 2024-11-15NINGBO INST OF TECH ZHEJIANG UNIV ZHEJIANG
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202210414301.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-20
Publication Date
2024-11-15
Estimated Expiration
2042-04-20

AI Technical Summary

Technical Problem

In the detection of rotating mechanical sealed acoustic emission signals, the acoustic emission signals are easily disturbed by noise, which leads to difficulty in signal identification. The existing hard threshold and soft threshold functions have discontinuities and derivative discontinuities in noise reduction processing, which is difficult to effectively handle than The wavelet coefficient of the threshold value causes the noise component reduction trend inconsistent.

Method used

Using local feature scale decomposition (LCD) combined with new wavelet threshold method, the signal noise reduction process is constructed, and pure components are screened through the acquisition of mechanically sealed acoustic emission signals, local feature scale decomposition, mutual correlation coefficient screening and new wavelet threshold noise reduction processing. Reconstruction overcomes the disadvantages of hard threshold and soft threshold.

Benefits of technology

It effectively reduces noise interference in mechanically sealed acoustic emission signals, improves signal purity, enhances the accuracy of status monitoring and fault diagnosis, and reduces the calculation amount of signal processing and the variance of noise components.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114858430B_ABST
    Figure CN114858430B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for reducing the noise of mechanical seal acoustic emission signals based on a new threshold of LCD-wavelet. The method decomposes the acoustic emission signal X(t) by local characteristic scale decomposition to obtain a number of ISC components arranged from high to low frequency; calculates the mutual correlation coefficient between each ISC component obtained by the local characteristic scale decomposition and the original signal; constructs a noisy component screening rule based on the combination of spectrum analysis and mutual correlation coefficient, and screens out pure components and noisy components from the ISC components; combines and reconstructs the screened noisy components, and performs noise reduction on the combined and reconstructed noisy components; reconstructs the screened pure signal components and the combined noisy components after noise reduction to obtain a signal after noise reduction by the new threshold of LCD-wavelet. The present invention can effectively solve the technical problem that the acoustic emission signal of the mechanical seal has many interference components in a low-speed unstable state, which is not conducive to state monitoring and fault diagnosis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a mechanical seal acoustic emission signal noise reduction method based on a new LCD-wavelet threshold, and belongs to the field of mechanical fault diagnosis. Background Art

[0002] Compared with commonly used vibration signals, acoustic emission signals have unique advantages in the early detection of friction faults in rotating machinery due to their sensitive response, wide frequency band and rich frequency components. However, in the detection of acoustic emission signals of rotating machinery seals, acoustic emission signals are often interfered by actual noise or even annihilated, making it difficult to identify useful acoustic emission signals. Therefore, when analyzing the collected acoustic emission signals, it is necessary to reduce the noise.

[0003] In the actual application of signal denoising, the hard threshold function is discontinuous in the entire wavelet domain and has discontinuities. It only processes wavelet coefficients less than the threshold, while those greater than the threshold are not processed, resulting in a large variance in the denoised signal. The soft threshold function is continuous in the wavelet domain. Although there is no problem of discontinuities, its derivative is indeed discontinuous. In practical applications, the signal often needs to be processed by first-order derivatives or even higher-order derivatives. In addition, when using the soft threshold function to denoise, those wavelet coefficients greater than the threshold are compressed at a constant value, which is actually inconsistent with the trend that the noise component gradually decreases as the wavelet coefficient increases. Summary of the invention

[0004] Aiming at the shortcomings and defects of the existing technology, a signal denoising method based on the combination of LCD and new wavelet threshold is provided for the preprocessing of mechanical seal acoustic emission signals. It can effectively solve the technical problem that the mechanical seal has more interference components in the low-speed unstable state, which is not conducive to state monitoring and fault diagnosis.

[0005] A method for reducing noise of mechanical seal acoustic emission signals based on LCD-wavelet new threshold, comprising the following steps:

[0006] (1) Collecting mechanical seal acoustic emission signals through a mechanical seal acoustic emission signal collection test platform;

[0007] (2) Decomposing the mechanical seal acoustic emission signal X(t) by local characteristic scale decomposition to obtain several ISC components arranged from high to low frequency;

[0008] (3) Calculate the mutual correlation coefficient between each ISC component obtained by local feature scale decomposition and the original signal;

[0009] (4) By performing FFT spectrum analysis on the ISC components, a noisy component screening rule based on the combination of spectrum analysis and mutual correlation coefficient is constructed to screen out pure components and noisy components from the ISC components;

[0010] (5) Using signal reconstruction technology, the filtered noisy components are combined and reconstructed, and the noise-containing components of the combined reconstruction are denoised using an improved wavelet new threshold denoising method;

[0011] (6) The filtered pure signal components and the denoised combined noisy components are reconstructed to obtain the LCD-wavelet new threshold denoised signal.

[0012] Furthermore, in step (1), the mechanical seal acoustic emission signal collection test platform includes a mechanical seal test bench, a variable frequency starter, an acoustic emission signal collector, a mechanical seal test piece, an air supply system, and an industrial computer, so that the mechanical seal acoustic emission signal can be collected more completely for subsequent noise reduction processing.

[0013] Furthermore, in step (2), the specific method for obtaining the ISC component is:

[0014] (2-1) Take all extreme points of x(t), use adjacent extreme points to divide x(t) into intervals, and within the intervals divided by the extreme points, use linear transformation to transform x(t);

[0015] (2-2) Assuming that the original signal is composed of the baseline signal Lt and the residual signal P1(t), the two are separated from the original signal x(t), and the separated P1(t) is judged using the ISC component criterion;

[0016] (2-3) ISC1(t) is removed from the original signal x(t), and the calculation is repeated n times. Each time, an ISC component that satisfies the component criterion can be obtained.

[0017] The LCD decomposition method separates the ISC components from the original signal in order from high to low frequency, decomposing the signal into components with different frequency scales, thereby fully mining the information.

[0018] Furthermore, in step (2), the expression satisfying the ISC component is:

[0019]

[0020] Where: ISC i (t) is the ISC component obtained after each cycle, r n is the final residual component;

[0021] The residual component rn in the decomposition result reflects the change trend of the signal. By eliminating the residual component, a certain degree of noise reduction effect can be achieved. In addition, compared with the signal decomposition method using spline difference, such as empirical mode decomposition (EMD), the LCD method separates the baseline signal from the original signal through linear transformation, with less calculation and better performance in suppressing endpoint effects and modal aliasing.

[0022] Furthermore, in step (4), the expression of the mutual correlation coefficient is:

[0023]

[0024] In the formula: R(i) is the mutual correlation coefficient; x(i,j) is the ISC component; y(j) is the vibration signal component; M is the signal length; N is the number of ISC components; since the source signal component is a component containing noise, and the decomposed ISC component contains noise and real information. Introducing the mutual correlation coefficient R can more reasonably select the appropriate ISC component for further noise reduction processing.

[0025] Furthermore, in step (5), the threshold function formula of the wavelet new threshold denoising method is:

[0026]

[0027] Where: W j,k is the wavelet coefficient; is the approximate wavelet coefficient; thr is the threshold value, N is the signal length;

[0028] The beneficial effects of the present invention are as follows: the signal noise reduction method based on the combination of LCD and new wavelet threshold is used for preprocessing of mechanical seal acoustic emission signals, which can effectively solve the technical problem that when the mechanical seal is in a low-speed unstable state, the acoustic emission signal has many interference components, which is not conducive to state monitoring and fault diagnosis.

[0029] The LCD method of the present invention separates the ISC components from the original signal in order from high to low frequency, decomposes the signal into components with different frequency scales, and thus fully mines the information. The residual component rn in the decomposition result reflects the change trend of the signal. By eliminating the residual component, a certain degree of noise reduction effect can be achieved. In addition, compared with the signal decomposition method using spline difference, such as empirical mode decomposition (EMD), the LCD method separates the baseline signal from the original signal through linear transformation, has a smaller amount of calculation, and has better performance in suppressing endpoint effects and modal aliasing.

[0030] The wavelet threshold filtering denoising method of the present invention considers that the wavelet coefficients corresponding to all useful signals contain important information of the signal, and the amplitude of the signal is relatively large, but the number is relatively small; while the distribution of the wavelet coefficients corresponding to the noise signal is consistent, and the number is relatively large, but the amplitude is relatively small. In the application of actual signal denoising, the hard threshold function is discontinuous in the entire wavelet domain, there are discontinuities, and only the wavelet coefficients less than the threshold are processed, while those greater than the threshold are not processed, so that the denoised signal has a large variance; the soft threshold function is continuous in the wavelet domain, although there is no problem of discontinuities, but its derivative is indeed discontinuous, and in actual applications, the signal is often processed by first-order derivative or even high-order derivative operations. The threshold function of the wavelet new threshold denoising method of the present invention overcomes the shortcomings of the soft threshold and hard threshold methods in denoising, and constructs a new threshold function to solve this problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Flow chart for noise reduction of mechanical seal acoustic emission signal;

[0032] Figure 2 This is the composition diagram of the mechanical seal acoustic emission signal acquisition system;

[0033] Figure 3 Break down the schematic for the LCD;

[0034] Figure 4 This is the LCD decomposition effect diagram;

[0035] Figure 5 It is the original signal diagram of mechanical seal acoustic emission signal;

[0036] Figure 6 ISC component diagram decomposed for LCD;

[0037] Figure 7 It is the FFT spectrum analysis diagram;

[0038] Figure 8 is the combined noisy ISC component map;

[0039] Fig. 9 It is the noisy ISC component image after wavelet new threshold denoising;

[0040] Fig.10 This is the signal waveform after LCD-wavelet new threshold denoising; DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and this embodiment.

[0042] According to Figure 1 The flowchart shown in Figure 2 The mechanical seal acoustic emission signal acquisition test platform shown in the figure performs a signal noise reduction test on a section of acoustic emission signals collected by the mechanical seal test bench during the startup process. Figure 5 As shown, the pressure selected in this embodiment is 0.1Mpa, and the specific noise reduction steps are as follows:

[0043] 1. First, the original signal is decomposed by local characteristic scale (LCD). The LCD decomposition principle is as follows: Figure 3 As shown;

[0044] Take all extreme points (Tk, Xk) of x(t), k = 1, 2, 3, ..., M, where M is the number of extreme points. Use adjacent extreme points to divide x(t) into intervals. Within the interval divided by the extreme points, linear transformation can be used to transform x(t):

[0045]

[0046] In the formula, is the baseline signal segment; k is the number of linear transformations performed.

[0047] According to the order of extreme point division, the baseline signal segments are combined in sequence to calculate the baseline signal L t , in formula (2-1),

[0048] L k+1 =aA k+1 +(1-a)X k+1 (2-2)

[0049] In the formula, the parameter a is generally taken as 0.5.

[0050] The original signal is composed of the baseline signal L t and the residual signal P1(t), then the two are separated from the original signal x(t). The separated P1(t) is judged by the ISC component criterion. If it meets the criterion conditions, ISC1(t) = P1(t). Otherwise, P1(t) is used as the initial signal, and steps (1) and (2) are recalculated. After k cycles, the ISC component Pk can be obtained, which is recorded as ISC1(t).

[0051] Remove ISC1(t) from the original signal x(t), and repeat steps (1) and (2) for n times, and each time you can get an intrinsic scale component that meets the component criterion. The calculation termination condition is the final residual component r n Monotonic, or less than the threshold set before signal decomposition. So we get

[0052]

[0053] Through LCD decomposition, 10 ISC components can be obtained, such as Figure 6 shown.

[0054] 2. Use the short-time Fourier method to make the spectrum of 10 ISC components and perform spectrum analysis, such as Figure 7 As shown;

[0055] 3. Solve the mutual correlation coefficient between each ISC component and the original signal, as shown in Table 1;

[0056] The expression of the mutual correlation coefficient is:

[0057]

[0058] Where: R is the mutual correlation coefficient; x(i,j) is the ISC component; y(j) is the vibration signal component; M is the signal length; N is the number of ISC components.

[0059] Table 1:

[0060]

[0061] 4. Combine the mutual correlation coefficient and spectrum analysis to screen out the noisy ISC components (ISC1, ISC2, ISC3, ISC7, ISC8, ISC9, ISC10), combine the noisy components, and perform wavelet threshold denoising. Figure 8 , Fig. 9 As shown;

[0062]

[0063] Where: W j,k is the wavelet coefficient; is the approximate wavelet coefficient; thr is the threshold value, N is the signal length;

[0064] 5. Use signal reconstruction technology to reconstruct the pure component and the noise-reduced signal to achieve noise reduction of mechanical seal acoustic emission signals. The noise-reduced signal is as follows: Fig.10 shown.

[0065] 6. Output the signal after noise reduction.

[0066] In order to verify the decomposition effect of LCD, the following simulation signal x(t) is decomposed using the LCD method. The simulation signal consists of an FM signal x1(t), an AM signal x2(t) and a sinusoidal signal x3(t), with a sampling frequency of 1000Hz and a sampling time of 1s.

[0067] x1(t)=[1+0.5sin(5πt)]cos[300πt+2cos(10πt)]

[0068] x2(t)=2e -2t sin(60πt)

[0069] x3(t)=sin(10πt)

[0070] x(t)=x1(t)+x2(t)+x3(t)

[0071] The effect after LCD decomposition is as follows Figure 4 As shown in the figure, it can be seen that the LCD method separates the baseline signal from the original signal through linear transformation, which requires less calculation and has better performance in suppressing endpoint effects and modal aliasing.

[0072] It can be concluded from the above embodiments that the present invention mainly proposes an effective solution to the problems that the acoustic emission signals generated by the mechanical seal during the start-stop stage are chaotic, unstable and contain a lot of noise.

[0073] The above are only preferred embodiments of the invention, and the protection scope of the invention is not limited to the above embodiments. All technical solutions under the concept of the invention belong to the protection scope of the invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the invention should also be regarded as the protection scope of the invention.

Claims

1. A method for reducing noise of mechanical seal acoustic emission signals based on LCD-wavelet new threshold, characterized in that: The following steps are involved: (1) Collect mechanical seal acoustic emission signals through the mechanical seal acoustic emission signal acquisition test platform; (2) Decompose the mechanical seal acoustic emission signal X(t) by local characteristic scale decomposition to obtain several ISC components arranged from high to low frequency; (3) Calculate the mutual correlation coefficient between each ISC component obtained by local feature scale decomposition and the original signal; (4) By performing FFT spectrum analysis on the ISC components, a noisy component screening rule based on the combination of spectrum analysis and mutual correlation coefficient is constructed to screen out pure components and noisy components from the ISC components; (5) Using signal reconstruction technology, the filtered noisy components are combined and reconstructed, and the improved wavelet new threshold denoising method is used to denoise the combined and reconstructed noisy components; (6) Reconstruct the filtered pure signal components and the denoised combined noisy components to obtain the LCD-wavelet new threshold denoised signal; In step (2), the specific method for obtaining the ISC component is: (2-1) Take all extreme points of x(t), use adjacent extreme points to divide x(t) into intervals, and use linear transformation to transform x(t) within the interval divided by the extreme points; (2-2) Assuming that the original signal consists of the baseline signal Lt and the residual signal P1(t), the two are separated from the original signal x(t), and the separated P1(t) is judged using the ISC component criterion; (2-3) Remove ISC1(t) from the original signal x(t) and repeat the calculation n times. Each time, an ISC component that satisfies the component criterion can be obtained. In step (5), the threshold function formula of the wavelet new threshold denoising method is: Where: is the wavelet coefficient; is the approximate wavelet coefficient; thr is the threshold, take thr= , N is the signal length; .

2. The method for reducing noise of mechanical seal acoustic emission signals based on LCD-wavelet new threshold according to claim 1 is characterized in that: In the step (1), the mechanical seal acoustic emission signal collection test platform includes a mechanical seal test bench, a variable frequency starter, an acoustic emission signal collector, a mechanical seal test piece, an air supply system, and an industrial computer.

3. The method for reducing noise of mechanical seal acoustic emission signals based on LCD-wavelet new threshold according to claim 1 is characterized in that: In step (2), the ISC component expression that satisfies the component criterion is: Where: ISC i (t) is the ISC component obtained after each cycle, r n is the final residual component.

4. The method for reducing noise of mechanical seal acoustic emission signals based on LCD-wavelet new threshold according to claim 1, characterized in that: In step (4), the expression of the mutual correlation coefficient is: Where: is the mutual correlation coefficient; is the ISC component; is the vibration signal component; M is the signal length; N is the number of ISC components.

Citation Information

Patent Citations

  • Wind-power gear box fault diagnosis method based on wavelet medium-soft threshold and electronic-magnetic diaphragm (EMD)

    CN102620928A

  • Wavelet threshold and EMD combined noise reduction method based on sparse decomposition

    CN110688964A

  • EEMD- and MSB-based failure feature extraction method for rolling-element bearing

    WO2019179340A1

  • Joint noise reduction method based on variational mode decomposition and permutation entropy

    WO2021056727A1