A method for extracting weak fault features of rolling bearings

Abstract

The present invention proposes a method for extracting weak fault features of rolling bearings, the steps of which are: constructing an adaptive fractal filter to filter the original weak fault signal of the rolling bearing, preliminarily filtering out the noise and enhancing the periodic impact characteristic components when the rolling bearing fails; using the optimal frequency band selection method based on the sparsity index to select the fault information sensitive frequency band of the signal filtered by the adaptive fractal filter; performing envelope spectrum analysis on the obtained fault information sensitive frequency band to extract the rolling bearing fault features. The present invention can effectively enhance the periodic impact characteristics when the rolling bearing fails, while effectively eliminating the interference of other components during the full-band envelope analysis, thereby more effectively extracting the rolling bearing weak fault features. The adaptive fractal filter of the present invention can construct the optimal structural element according to the signal to be analyzed itself, and the optimal information sensitive frequency band selection method based on the sparsity index has stronger noise robustness.

Description

Technical Field

[0001] The invention relates to the technical field of rolling bearing weak fault feature extraction, and in particular to a rolling bearing weak fault feature extraction method. Background Art

[0002] When a rolling bearing has an early weak fault, conventional signal feature extraction methods such as envelope spectrum analysis often have difficulty in achieving good feature extraction results. The main reasons are: (1) The influence of strong background noise: Compared with the interference component, the periodic impact characteristics of a rolling bearing when an early weak fault occurs are often very weak and are thus obscured by the background noise. (2) When the original signal is directly subjected to full-band envelope spectrum analysis, the envelope extraction effect is poor due to the interference of multi-frequency components. For the solution to the first factor, it is often necessary to use signal processing methods such as wavelet analysis and empirical mode decomposition to pre-de-noise the original signal. However, wavelet analysis requires the selection of the optimal wavelet basis in advance, and the selection of the optimal wavelet basis requires prior knowledge of the signal to be analyzed, which is difficult to achieve in most practical projects. Empirical mode decomposition often has the problem of modal aliasing. For the solution to the second factor, the conventional approach is to select the optimal information-sensitive frequency band based on empirical values ​​or frequency domain diagrams, and perform envelope analysis on the selected frequency band to improve the full-band feature extraction effect. However, this often has a large degree of randomness or requires rich fault diagnosis experience.

[0003] The invention patent with application number 201910276672.8 discloses a method for extracting weak fault features of local damage of rolling bearings. The vibration signal of the faulty rolling bearing is collected by an acceleration sensor, and the original signal is decomposed into different nodes by using MODWPT. Then, the SK of the mean square envelope autocorrelation of each node is calculated, and then the node signal of each layer that is not less than half of the maximum SK is selected for MED filtering. Then, the normalized spectrum information of the selected nodes in the same layer is superimposed and averaged according to the number of layers, and finally the spectrum of all layers is merged to extract the local damage fault features of the rolling bearing. The invention provides a method for extracting the local damage fault features of rolling bearings under strong background noise interference, which provides an effective diagnostic tool for realizing the early weak fault feature extraction of PHM of rolling bearings. However, the above patent has a relatively large amount of calculation due to the calculation of MODWPT and subsequent nodes SK, and has the disadvantage of poor timeliness in actual engineering applications. In addition, due to the sensitivity of SK to additional accidental impact components, it cannot truly reflect the periodic impact characteristics of the signal when the rolling bearing has a local damage fault. Summary of the invention

[0004] In view of the technical problems that existing signal feature extraction methods have modal aliasing and the analysis is relatively random, the present invention proposes a method for extracting weak fault features of rolling bearings. An adaptive fractal filter constructs an optimal structural element based on the signal itself. The adaptive fractal filter performs noise reduction on the original signal while effectively enhancing the periodic impact components in the signal. Then, the optimal frequency band is selected for the noise-reduced signal using an optimal information frequency band selection method based on a sparsity index. Finally, envelope analysis is performed on the selected frequency band to extract effective fault features.

[0005] In order to achieve the above object, the technical solution of the present invention is implemented as follows: a method for extracting weak fault features of rolling bearings, the steps of which are as follows:

[0006] Step 1: Construct an adaptive fractal filter to filter the original weak fault signal of the rolling bearing, preliminarily filter out the noise and enhance the periodic impact characteristic components when the rolling bearing fails;

[0007] Step 2: Select the fault information sensitive frequency band of the signal filtered by the adaptive fractal filter using the optimal frequency band selection method based on the sparsity index;

[0008] Step 3: Perform envelope spectrum analysis on the optimal fault information sensitive frequency band obtained in step 2 to extract rolling bearing fault characteristics.

[0009] Preferably, the step of filtering using an adaptive fractal filter is as follows:

[0010] ① Find the maximum and minimum points in the original vibration signal;

[0011] ② Use the cubic spline interpolation method to fit the maximum point and the minimum point to obtain the fitting curve;

[0012] ③ The narrow-domain waveform between all adjacent minimum values ​​of the fitting curve is used as the structural element shape of the adaptive fractal filter;

[0013] ④ Based on the obtained structure element shape, an adaptive fractal filter based on the MHPO operator is constructed to filter the original weak fault signal.

[0014] Preferably, the calculation formula of the MHPO operator is as follows:

[0015] MHPO(n)=AHDE(n)·AHCO(n)

[0016]

[0017]

[0018] In the formula, f represents the original weak fault signal to be filtered, g is the structural element of the adaptive fractal filter, n represents the number of points of the original weak fault signal, Θ、 and represent the dilation operator, erosion operator, opening operator and closing operator in the adaptive fractal filter respectively; MHPO(n) represents the MHPO operator of the original weak fault signal, AHDE(n) and AHCO(n) are the intermediate calculation variables of the MHPO operator.

[0019] Preferably, the method for selecting the fault information sensitive frequency band is:

[0020] ① Segment processing of the signal filtered by the adaptive fractal filter;

[0021] ② Use the Kurtogram algorithm to divide the frequency band of the segmented signal;

[0022] ③ Calculate the sparsity index GI of each frequency band;

[0023] ④ Select the frequency band with the largest sparsity index as the optimal fault information sensitive frequency band.

[0024] Preferably, the sparsity index GI is:

[0025]

[0026] In the formula, ||·||1 represents the l1 norm, N represents the number of segmented signals, SE p is the square wave envelope value of the signal in the pth divided frequency band in analytical form, SE r Indicates the square wave envelope value SE p In descending order, that is: SE r [1]≤SE r [2]≤…≤SE r [N].

[0027] Preferably, the calculation formula of the square wave envelope value SE is:

[0028]

[0029] In the formula, represents the analytical form of the vibration signal to be analyzed, H() is the Hilbert transform operator, j 2 =-1.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows: the proposed adaptive fractal filter and the optimal information sensitive frequency band method are used to perform dual enhancement on the rolling bearing weak fault characteristics, which can quickly and effectively enhance the periodic impact characteristics when the rolling bearing fails, while effectively eliminating the interference of other components during the full-band envelope analysis, thereby more effectively extracting the rolling bearing weak fault characteristics. In addition, the adaptive fractal filter of the present invention can construct the optimal structural element according to the signal to be analyzed, and the invented optimal information sensitive frequency band selection method based on the sparsity index has stronger robustness to accidental impact noise than the kurtosis index. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 It is a schematic diagram of the process of the present invention.

[0033] Figure 2 The time domain diagram and envelope spectrum diagram of the early weak fault signal of the rolling bearing, where (a) is the time domain diagram and (b) is the envelope spectrum diagram.

[0034] Figure 3 It is a time domain diagram of the signal after being filtered by the adaptive fractal filter of the present invention.

[0035] Figure 4 This is the analysis result of the signal after being filtered by the adaptive fractal filter using the optimal information sensitive frequency band analysis method of the present invention.

[0036] Figure 5 The envelope and demodulation results of the optimal information sensitive frequency band signal of the present invention, wherein (a) is the envelope result and (b) is the demodulation result. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] like Figure 1 As shown, a rolling bearing weak fault feature extraction method includes the following steps:

[0039] Step 1: Construct an adaptive fractal filter to filter the original weak fault signal of the rolling bearing, preliminarily filter out the noise and enhance the periodic impact characteristic components when the rolling bearing fails.

[0040] The specific process of filtering by adaptive fractal filter includes:

[0041] ① Find the maximum and minimum points in the original vibration signal.

[0042] ② Use the cubic spline interpolation method with higher computational efficiency to fit the maximum and minimum points to obtain the fitting curve.

[0043] ③ The narrow-domain waveform between all adjacent minimum values ​​of the fitting curve is used as the structural element shape of the adaptive fractal filter. This method can be used to adaptively construct the optimal structural element according to the time domain waveform characteristics of the signal to be filtered.

[0044] ④ Based on the obtained structural element shape, an adaptive fractal filter based on the MHPO operator is constructed to filter the original weak fault signal. The calculation formula of the MHPO operator is as follows:

[0045] MHPO(n)=AHDE(n)·AHCO(n)

[0046]

[0047]

[0048] In the formula, f represents the original weak fault signal to be filtered, g is the structural element of the adaptive fractal filter, n represents the number of points of the original weak fault signal, Θ、 and ·represent the four basic operators in the adaptive fractal filter, namely: dilation operator, erosion operator, opening operator and closing operator. MHPO(n) represents the MHPO operator of the original weak fault signal, and AHDE(n) and AHCO(n) are the intermediate calculation variables of the MHPO operator. AHDE(n) and AHCO(n) both have an enhancement effect on the impact component of the original weak fault signal. MHPO is the time domain multiplication of the two, which can further enhance the impact component.

[0049] Step 2: Use the optimal frequency band selection method based on sparsity index to select the fault information sensitive frequency band of the signal after being filtered by the adaptive fractal filter.

[0050] The method for selecting the fault information sensitive frequency band is:

[0051] ① Segment processing is performed on the signal filtered by the adaptive fractal filter. The specific segmentation algorithm is as follows:

[0052] N=(n seg -1)×(L seg -L overlap )+L seg

[0053] Where N is the number of points of the original signal; L overlap Indicates the overlap rate of two adjacent data segments, and the present invention uses 50%; n seg Represents the number of segments, L seg Indicates the length of each segment of data.

[0054] ② Use the Kurtogram algorithm to divide the frequency band of the segmented signal.

[0055] ③ Calculate the sparsity index GI of each frequency band. The specific calculation process is as follows:

[0056]

[0057] In the formula, ||·||1 represents the l1 norm, N represents the number of segmented signals, SE p is the square wave envelope value of the signal in the pth divided frequency band in analytical form, SE r Indicates the square wave envelope value SE p In descending order, that is: SE r [1]≤SE r [2]≤…≤SE r [N]. The calculation formula of the square wave envelope value SE is as follows:

[0058]

[0059] In the formula, represents the analytical form of the vibration signal to be analyzed, H() is the Hilbert transform operator, j 2 =-1.

[0060] ④ Select the frequency band with the largest sparsity index as the optimal fault information sensitive frequency band. The larger the sparsity index GI is, the richer the periodic shock characteristic components contained in this frequency band are. Envelope spectrum analysis of this frequency band will obtain more effective feature extraction results.

[0061] Step 3: Perform envelope spectrum analysis on the optimal fault information sensitive frequency band obtained in step 2 to extract rolling bearing fault characteristics. Envelope spectrum analysis based on fault information sensitive frequency band can further reduce the interference effect of noise components during full-band spectrum analysis.

[0062] Specific example: Conducting accelerated fatigue full life cycle experiments on rolling bearings to collect early weak fault stage signals of the inner ring of rolling bearings; Figure 2As shown in the figure, it can be seen that neither the time domain nor the envelope spectrum analysis can reflect the fault characteristics. The collected early weak stage vibration signal is filtered by an adaptive fractal filter; Figure 3 The time domain diagram of the filtered vibration signal is shown in Figure 3 It can be seen that relative to Figure 2 (a), the impact characteristics in the time domain are significantly enhanced. The optimal information sensitive frequency band analysis of the filtered signal is performed, and the results are as follows Figure 4 Based on Figure 4 The analyzed results show that the optimal center frequency f c =3200Hz, bandwidth B w = 6400, to construct the optimal band-pass filter, and use the optimal band-pass filter to further filter the vibration signal filtered by the adaptive fractal filter, and envelop the filtered signal (the result is as follows Figure 5 (a)) and spectrum analysis (results as shown in Figure 5 (b) to extract effective fault features: Figure 5 (b) The rolling bearing inner ring fault characteristic frequency 245 Hz and its harmonic frequencies are effectively extracted.

[0063] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for extracting weak fault features of rolling bearings, characterized in that: The steps are as follows: Step 1: Construct an adaptive fractal filter to filter the original weak fault signal of the rolling bearing, preliminarily filter out the noise and enhance the periodic impact characteristic components when the rolling bearing fails; The method of filtering by the adaptive fractal filter in step 1 is: ① Find the maximum and minimum points in the original vibration signal; ② Use the cubic spline interpolation method to fit the maximum point and the minimum point to obtain the fitting curve; ③ The narrow-domain waveform between all adjacent minimum values ​​of the fitting curve is used as the structural element shape of the adaptive fractal filter; ④ Based on the obtained structure element shape, an adaptive fractal filter based on the MHPO operator is constructed to filter the original weak fault signal; Step 2: Select the fault information sensitive frequency band of the signal filtered by the adaptive fractal filter using the optimal frequency band selection method based on the sparsity index; The method for selecting the fault information sensitive frequency band is: ① Segment processing of the signal filtered by the adaptive fractal filter; ② Use the Kurtogram algorithm to divide the frequency band of the segmented signal; ③ Calculate the sparsity index GI of each frequency band; ④ Select the frequency band with the largest sparsity index GI as the optimal fault information sensitive frequency band; Step 3: Perform envelope spectrum analysis on the optimal fault information sensitive frequency band obtained in step 2 to extract rolling bearing fault characteristics.

2. The rolling bearing weak fault feature extraction method according to claim 1 is characterized in that: The calculation formula of the MHPO operator is as follows: MHPO(n)=AHDE(n)·AHCO(n) In the formula, f represents the original weak fault signal to be filtered, g is the structural element of the adaptive fractal filter, n represents the number of points of the original weak fault signal, Θ、 and represent the dilation operator, erosion operator, opening operator and closing operator in the adaptive fractal filter respectively; MHPO(n) represents the MHPO operator of the original weak fault signal, AHDE(n) and AHCO(n) are the intermediate calculation variables of the MHPO operator.

3. The rolling bearing weak fault feature extraction method according to claim 1 or 2, characterized in that: The sparsity index GI is: In the formula, ||·||1 represents the l1 norm, N represents the number of segmented signals, SE p is the square wave envelope value of the signal in the pth divided frequency band in analytical form, SE r Indicates the square wave envelope value SE p In descending order, that is: SE r [1]≤SE r [2]≤…≤SE r [N].

4. The rolling bearing weak fault feature extraction method according to claim 3 is characterized in that: The calculation formula of the square wave envelope value SE is: In the formula, represents the analytical form of the vibration signal to be analyzed, H() is the Hilbert transform operator, j 2 =-1.

Images