Rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA

Abstract

The invention provides a rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA, and relates to the technical field of bearing fault diagnosis. Firstly, a rolling bearing fault vibration signal is collected, secondly, wavelet packet decomposition is used to decompose the vibration signal into eight intrinsic mode components, kurtosis, a correlation coefficient and permutation entropy are weighted and fused into a weighted index according to a weighted index screening criterion, a threshold index is calculated, and then a high signal-to-noise ratio component is screened out and reconstructed. And then, searching an optimal deconvolution period Tbest of the MOMEDA algorithm by using envelope autocorrelation analysis, and inputting the optimal deconvolution period Tbest into the MOMEDA to filter a reconstructed signal component with a high signal-to-noise ratio so as to realize enhancement of a periodic impact component related to a fault. And finally, envelope demodulation is carried out on the output signal, fault feature information is extracted from an envelope spectrum, and fault diagnosis of the rolling bearing is realized.

Description

Technical Field

[0001] The invention relates to the technical field of bearing fault diagnosis, and in particular to a rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA. Background Art

[0002] Rolling bearings are important supporting parts in aircraft engines. They are prone to damage due to harsh working environments. Common failure modes of rolling bearings include fatigue spalling, wear, rust and contamination. When failure occurs, the working performance of the aircraft engine will be greatly reduced, and may even cause major safety accidents. Therefore, the development of accurate and effective bearing early fault feature enhancement algorithms is of great significance to the sustainable development of the aviation industry and flight safety. The early fault types of aircraft engine bearings mainly include cracks, fatigue spalling and rolling element failures. These types of failures will produce periodic impacts during the operation of aircraft engines. However, due to the complex structure of the aircraft engine itself and the large amount of interference information generated during operation, the defect impact signal is difficult to identify, and fault diagnosis is difficult. When a rolling bearing fails, its vibration signal is complex, nonlinear and non-stationary. Conventional signal processing methods are difficult to achieve the enhancement of early weak fault features.

[0003] Bearing early fault signals are often weak and have low signal-to-noise ratios, which makes it challenging to accurately extract fault features. Wavelet packet decomposition combined with effective screening criteria can effectively retain the effective components in bearing fault vibration signals and improve the signal-to-noise ratio, but it is difficult to enhance early weak fault features. The multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) algorithm has been proven to enhance the impact components in vibration signals, but the deconvolution period T of the MOMEDA algorithm has a huge impact on the accuracy of the calculation results. Therefore, it is necessary to develop effective methods to improve the above algorithms in order to more effectively enhance the early fault features of bearings. Summary of the invention

[0004] The technical problem to be solved by the present invention is to address the deficiencies of the above-mentioned prior art and provide a rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA. The fault feature information is enhanced by combining weighted screening with improved multi-point optimal minimum entropy deconvolution (IMOMEDA), and the enhanced signal is used for fault diagnosis through envelope spectrum analysis. The noise components can be effectively filtered out, the fault features can be enhanced, and rolling bearing fault diagnosis can be realized.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA includes the following steps:

[0007] Step 1: using a vibration sensor to collect the vibration signal of the rolling bearing and measuring the relevant parameters of the bearing at the same time; the relevant parameters of the bearing include contact angle β, number of balls Z, ball diameter d and pitch diameter D;

[0008] Step 2: By inputting the basic parameters of the rolling bearing, the fault characteristic frequency values ​​of different fault types are calculated;

[0009] Step 3: Decompose the vibration signal by wavelet packet, and use the weighted screening criterion to fuse the kurtosis, correlation coefficient and permutation entropy of the node components decomposed by the wavelet packet into a weighted index;

[0010] Step 4: Calculate the signal division threshold index C:

[0011]

[0012] Among them, S j is the weighted index of the jth node component; n is the number of node components;

[0013] Step 5: Reconstruct the high signal-to-noise ratio components selected according to the weighted index screening criteria, perform envelope autocorrelation analysis on the reconstructed signal, and set the search interval for the optimal deconvolution period to find the optimal deconvolution period T of MOMEDA within the search interval. best , search for the maximum value of the autocorrelation function within the set search interval, and its corresponding horizontal axis is the optimal deconvolution period T best ;

[0014] Step 6: The optimal deconvolution period T best Input into the MOMEDA algorithm, perform secondary filtering on the reconstructed signal to enhance the fault characteristics;

[0015] Step 7: Perform envelope demodulation on the signal output in step 6, and extract fault feature information from the envelope spectrum.

[0016] Furthermore, in step 2, the specific calculation method of the fault characteristic frequency value of different fault types is as follows:

[0017] In the outer ring fault, the calculation formula of the fault characteristic frequency is:

[0018]

[0019] Among them, D b Represents the rolling element diameter, D c represents the pitch diameter of the raceway, β represents the bearing contact angle, z represents the number of rolling elements, and f zIndicates the rotation frequency of the inner ring of the bearing, f r Indicates the rotation frequency of the outer ring of the bearing, f o Indicates the characteristic frequency of outer race fault;

[0020] In the inner ring fault, the calculation formula of the fault characteristic frequency value is:

[0021]

[0022] Among them, D b Represents the rolling element diameter, D c represents the pitch diameter of the raceway, β represents the bearing contact angle, z represents the number of rolling elements, and f z Indicates the rotation frequency of the inner ring of the bearing, f r Indicates the rotation frequency of the outer ring of the bearing, f i Indicates the characteristic frequency of inner race fault;

[0023] In the case of rolling element failure, when the inner and outer rings rotate simultaneously, the calculation formula for the fault characteristic frequency value is:

[0024]

[0025] Among them, D b Represents the rolling element diameter, D c represents the pitch diameter of the raceway, β represents the bearing contact angle, z represents the number of rolling elements, and f z Indicates the rotation frequency of the inner ring of the bearing, f r Indicates the rotation frequency of the outer ring of the bearing, f b Indicates the characteristic frequency of rolling element failure.

[0026] Furthermore, the specific method of step 3 is:

[0027] Step 3.1: Arrange the kurtosis, correlation coefficient and permutation entropy of each node component into a complete indicator matrix:

[0028] Step 3.2: Standardize the indicators to obtain a standardized indicator matrix; among them, kurtosis and correlation coefficient are positive indicators, and are standardized and calculated using formula (1), while permutation entropy is standardized and calculated using formula (2);

[0029]

[0030] Where i represents the indicator type index; j = 1, 2, ..., n; x i is the i-th index vector, x ij is the i-th index value of the j-th node component; max(x i )、min(x i ) represent x i The maximum and minimum values ​​of

[0031] Step 3.3: Calculate the proportion of the i-th index value of the j-th node component after standardization to the overall index. The mathematical expression is as follows:

[0032]

[0033] Among them, h ij Through formula (1) and formula (2), we can get: i = 1, 2, 3;

[0034] Step 3.4: Calculate the indicator information entropy:

[0035]

[0036] Where, k = 1 / lnn;

[0037] Step 3.5: Calculate indicator weights by information entropy:

[0038]

[0039] Step 3.6: Calculate the weighted index S j :

[0040] S j =d1h 1j +d2h 2j +d3h 3j (6)

[0041] Among them, d i Through formula (5), we can get: ij It is obtained by equation (1) and equation (2).

[0042] Furthermore, the weighted index screening criteria in step 5 are:

[0043] The weighted index S j Compared with the partition threshold index C, if the weighted index S j If it is greater than the partition threshold index C, the j-th node component is a high signal-to-noise ratio component, otherwise it is a low signal-to-noise ratio component.

[0044] The beneficial effects of adopting the above technical solution are: the rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA provided by the present invention proposes a weighted indicator screening criterion for weighting indicators according to the discrete degree of signal indicators on the basis of the existing multi-parameter fusion screening criterion, and fuses the kurtosis value, correlation coefficient and permutation entropy into weighted indicators. The method completely relies on the discrete degree of data to determine the weights, avoiding the interference of subjective factors, so that the obtained results are more objective and reliable. Threshold division is selected, and the weighted indicator S is compared with the threshold indicator C, so as to realize the division of high signal-to-noise ratio and low signal-to-noise ratio signals, and the high signal-to-noise ratio components screened out are reconstructed. The high signal-to-noise ratio components screened out according to the weighted indicator screening criterion retain the effective information in the original signal to the greatest extent. The method of searching for the deconvolution period T of MOMEDA using envelope correlation analysis can achieve the optimal deconvolution period T under strong background noise interference. best The method of the present invention can not only effectively filter out the noise signal in the signal, but also enhance the weak periodic fault-related components to enhance the fault characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flow chart of a rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA provided in an embodiment of the present invention;

[0046] Figure 2 The function image of the autocorrelation function-deconvolution period for finding the best deconvolution period of the simulated bearing provided by the embodiment of the present invention;

[0047] Figure 3 A fault characteristic envelope spectrum for simulating bearing signals provided by an embodiment of the present invention;

[0048] Figure 4 The function image of the autocorrelation function-deconvolution period for finding the best deconvolution period of the experimental inner ring fault provided by the embodiment of the present invention;

[0049] Figure 5 The fault characteristic envelope spectrum of the experimental bearing inner ring fault provided by the embodiment of the present invention;

[0050] Figure 6 This is a fault characteristic envelope spectrum diagram of an experimental bearing outer ring fault provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0052] This embodiment uses real experimental data for analysis, which is taken from a university bearing data center. The faulty bearing selected for analysis is a 6205-2RJEM SKF deep groove ball bearing, and the inner and outer rings of the bearing are processed for single-site damage using electric spark technology. The sampling frequency of the vibration data is 12000Hz.

[0053] like Figure 1 As shown, the method of this embodiment is described as follows.

[0054] Step 1: Use a vibration sensor to collect the vibration signal of the rolling bearing and measure the relevant parameters of the bearing.

[0055] The relevant parameters of the bearing include: contact angle β, number of balls Z, ball diameter d, and pitch diameter D.

[0056] Step 2: By inputting the basic parameters of the rolling bearing, calculate the fault characteristic frequency values ​​of different fault types. The specific calculation theoretical formula is shown in the following table.

[0057] Table 1 Theoretical formula for calculating fault characteristic frequency values ​​of different fault types

[0058]

[0059] Step 3: Decompose the vibration signal using wavelet packets and calculate the kurtosis, correlation coefficient and permutation entropy of each component.

[0060] Step 4: Arrange the kurtosis, correlation coefficient and permutation entropy of the node components into a complete indicator matrix, as shown in Table 2.

[0061] Table 2 Indicator matrix

[0062]

[0063] Step 5: Standardize the indicators to obtain a standardized indicator matrix. Among them, kurtosis and correlation coefficient are positive indicators, which are standardized and calculated using formula (1), while permutation entropy is standardized and calculated using formula (2).

[0064]

[0065] Wherein, i represents the index type index; j=1, 2, ..., 8, and the number of node components n=8 in this embodiment; x i and x ij Through the indicator matrix in Table 2, we can get max(x i )、min(x i ) represent x i The maximum and minimum values ​​of .

[0066] Step 6: Calculate the proportion of the j-th node component of the i-th indicator after standardization to the overall indicator. The mathematical expression is as follows:

[0067]

[0068] Among them, h ij According to formula (1) and formula (2), i=1, 2, 3.

[0069] Step 7: Calculate the indicator information entropy:

[0070]

[0071] Among them, k = 1 / lnn.

[0072] Step 8: Calculate indicator weights by information entropy:

[0073]

[0074] Step 9: Calculate the weighted index S j And the partition threshold index C:

[0075] S j =d1h 1j +d2h 2j +d3h 3j (6)

[0076] Among them, d i Through formula (5), we can get: ij It is obtained by equation (1) and equation (2).

[0077]

[0078] Where n = 8; S j It is obtained by formula (6).

[0079] Step 10: Set the weighted index S j Compare with the threshold index C to achieve the division of high signal-to-noise ratio and low signal-to-noise ratio signals and reconstruct the high signal-to-noise ratio signal component. The division criteria are shown in the following table.

[0080] Table 3 Division criteria

[0081] <![CDATA[S j >C]]> High signal-to-noise ratio component <![CDATA[S j ≤C]]> Low signal-to-noise ratio component

[0082] Step 11: Perform envelope autocorrelation analysis on the reconstructed signal, set a search interval for the optimal deconvolution period, and find the optimal deconvolution period within the set search interval.

[0083] Step 12: Input the searched deconvolution cycle into the MOMEDA algorithm to enhance the fault characteristics of the reconstructed signal.

[0084] Step 13: Perform envelope demodulation on the output signal and extract fault feature information from the envelope spectrum.

[0085] In this embodiment, the fault feature enhancement method combined with weighted screening and IMOMEDA is used to process the simulated bearing fault signal to obtain a function image of the autocorrelation function-deconvolution period as shown in Figure 2 As shown, the envelope spectrum of the new signal is formed as Figure 3 As shown. Figure 3 The optimal deconvolution period T can be clearly extracted best =107.1. The peak factor of the processed signal is 9.7, which is 7.3 compared with the original signal. The impact of the processed signal is improved. The fault characteristic frequency and its multiple frequency appear in the envelope spectrum, indicating that the outer ring of the rolling bearing has a fault. In addition, the outer ring fault characteristic frequency and its multiple frequency are more prominent, which shows that the method has good robustness to background noise.

[0086] The function image of the experimental bearing outer ring fault autocorrelation function-deconvolution period is as follows Figure 4 As shown, the optimal deconvolution period T can be extracted from the figure best =113.2, and input it into MOMEDA to enhance the fault characteristics. The envelope spectrum of the enhanced signal in the frequency band of 0 to 1000 is as follows Figure 5 As shown in the figure, it can be seen that the envelope spectrum after fault feature enhancement combined with weighted screening and IMOMEDA can suppress the interference of noise components and extract very rich fault information. The fault feature frequency f o =106.8Hz, 2f o =213.6Hz, 3f o =320.4Hz, 4f o =427.2Hz, 5f o =534.0Hz, 6f o =640.8Hz, 7f o =747.6Hz、8f o =854.4Hz、9f o =961.2Hz. The overall analysis shows that this method can extract very obvious outer ring fault characteristics. The fault characteristic frequency multiplication peak value repeatedly decreases first and then remains stable. During the first decrease, the peak value reaches the minimum value at the 3rd frequency, and then the fault characteristic frequency peak value fluctuates within a small range.

[0087] After processing, the envelope spectrum of the experimental bearing inner ring fault signal in the range of 0 to 1000 is as follows: Figure 6As shown in the figure, it can be seen that the fault feature enhancement envelope spectrum after combining weighted screening and IMOMEDA can suppress the interference of noise components and extract very rich fault information. The fault feature frequency f i =160.5Hz, 2f i =321.1Hz, 3f i =481.5Hz, 4f i =642.0Hz, 5f i =802.5Hz, 6f i =963.0Hz. The overall analysis shows that the noise is significantly suppressed, the fault characteristic frequency peak is obvious, the fault characteristic frequency multiplication peak shows a continuous fluctuation, and the peak is the smallest at the 5th frequency, and then begins to rise.

[0088] The study found that the method proposed in the present invention can successfully extract the characteristic frequency of the outer ring fault and its multiples. In summary, it can be determined that a fault has occurred in the outer ring of the rolling bearing. At the same time, it can further verify the effectiveness of the fault feature enhancement combined with weighted index screening and IMOMEDA for early weak rolling faults.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA, characterized by: The steps include: Step 1: using a vibration sensor to collect the vibration signal of the rolling bearing and simultaneously measuring the relevant parameters of the bearing; the relevant parameters of the bearing include the contact angle β, the number of balls Z, the ball diameter d and the pitch diameter D; Step 2: By inputting the basic parameters of the rolling bearing, the fault characteristic frequency values ​​of different fault types are calculated; Step 3: Decompose the vibration signal by wavelet packet, and use the weighted screening criterion to fuse the kurtosis, correlation coefficient and permutation entropy of the node components decomposed by the wavelet packet into a weighted index; Step 4: Calculate the signal division threshold index C: Among them, S j is the weighted index of the jth node component; n is the number of node components; Step 5: Reconstruct the high signal-to-noise ratio components selected according to the weighted index screening criteria, perform envelope autocorrelation analysis on the reconstructed signal, and set the search interval for the optimal deconvolution period to find the optimal deconvolution period T of MOMEDA within the search interval. best , search for the maximum value of the autocorrelation function within the set search interval, and its corresponding horizontal axis is the optimal deconvolution period T best ; Step 6: The optimal deconvolution period T best Input into the MOMEDA algorithm, perform secondary filtering on the reconstructed signal to enhance the fault characteristics; Step 7: Perform envelope demodulation on the signal output in step 6, and extract fault feature information from the envelope spectrum.

2. The rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA according to claim 1 is characterized in that: In step 2, the specific calculation method of the fault characteristic frequency values ​​of different fault types is as follows: In the outer ring fault, the calculation formula of the fault characteristic frequency is: Among them, D b Represents the rolling element diameter, D c represents the pitch diameter of the raceway, β represents the bearing contact angle, z represents the number of rolling elements, and f z Indicates the rotation frequency of the inner ring of the bearing, f r Indicates the rotation frequency of the outer ring of the bearing, f o Indicates the characteristic frequency of outer race fault; In the inner ring fault, the calculation formula of the fault characteristic frequency value is: Among them, D b Represents the rolling element diameter, D c represents the pitch diameter of the raceway, β represents the bearing contact angle, z represents the number of rolling elements, and f z Indicates the rotation frequency of the inner ring of the bearing, f r Indicates the rotation frequency of the outer ring of the bearing, f i Indicates the characteristic frequency of inner race fault; In the case of rolling element failure, when the inner and outer rings rotate simultaneously, the calculation formula for the fault characteristic frequency value is: Among them, D b Represents the rolling element diameter, D c represents the pitch diameter of the raceway, β represents the bearing contact angle, z represents the number of rolling elements, and f z Indicates the rotation frequency of the inner ring of the bearing, f r Indicates the rotation frequency of the outer ring of the bearing, f b Indicates the characteristic frequency of rolling element failure.

3. The rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA according to claim 1, characterized in that: The specific method of step 3 is: Step 3.1: Arrange the kurtosis, correlation coefficient and permutation entropy of each node component into a complete indicator matrix: Step 3.2: Standardize the indicators to obtain a standardized indicator matrix; among them, kurtosis and correlation coefficient are positive indicators, and are standardized and calculated using formula (1), while permutation entropy is standardized and calculated using formula (2); Where i represents the indicator type index; j = 1, 2, ..., n; x i is the i-th index vector, x ij is the i-th index value of the j-th node component; max(x i )、min(x i ) represent x i The maximum and minimum values ​​of Step 3.3: Calculate the proportion of the i-th index value of the j-th node component after standardization to the overall index. The mathematical expression is as follows: Among them, h ij Through formula (1) and formula (2), we can get: i = 1, 2, 3; Step 3.4: Calculate the indicator information entropy: Where, k = 1 / lnn; Step 3.5: Calculate indicator weights by information entropy: Step 3.6: Calculate the weighted index S j : S j =d1h 1j +d2h 2j +d3h 3j (6) Among them, d i Through formula (5), we can get: ij It is obtained by equation (1) and equation (2).

4. The rolling bearing fault feature enhancement method based on weighted screening and IMOMEDA according to claim 3 is characterized in that: The weighted index screening criteria in step 5 are: The weighted index S j Compared with the partition threshold index C, if the weighted index S j If it is greater than the partition threshold index C, the j-th node component is a high signal-to-noise ratio component, otherwise it is a low signal-to-noise ratio component.

Images