Rotating machinery rubbing fault recognition method

Abstract

The application provides a rotating machinery rubbing fault recognition method, comprising: obtaining a vibration signal of a rotating machinery; constructing a Hankel matrix of the vibration signal; calculating a covariance matrix of the Hankel matrix; calculating a mean value of each column in the covariance matrix and constructing a reconstructed signal R by using the mean value; taking the local mean value of the signal in a window as a criterion, performing binary quantization on the reconstructed signal R according to a 1D-LBP method to obtain a local texture signal; and performing spectral analysis on the local texture signal and extracting a prominent frequency component in the spectrum to extract a rotating machinery rubbing fault feature and recognize a fault. The rotating machinery rubbing fault recognition method can effectively suppress noise and can effectively extract rubbing fault feature information and accurately recognize a fault.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of fault diagnosis, and particularly relates to a rotating machinery rubbing fault recognition method. BACKGROUND

[0002] Rotating machinery is widely used in aviation, machinery and other fields, and effective monitoring of its running state and accurate judgment of its fault are crucial for reducing economic losses and avoiding major safety accidents. In order to improve the performance of large equipment such as aviation engines, it is usually necessary to reduce the rotating static gap, which greatly increases the possibility of rotating static rubbing. Rubbing fault can cause excessive vibration of the equipment, and even cause catastrophic accidents. Therefore, it is of great practical significance to study the rubbing fault diagnosis method.

[0003] In the prior art, in order to accurately extract rubbing fault feature information, scholars have identified and judged rubbing fault from multiple angles such as signal decomposition (such as variational mode decomposition, empirical mode decomposition), noise reduction algorithm (such as wavelet threshold denoising, singular value denoising), information fusion (such as principal component analysis). These signal analysis methods have their own advantages in processing rubbing fault signals, but also have certain limitations. For example, the determination of the number of decomposition layers and the center frequency of variational mode decomposition has a crucial influence on the accurate extraction of signal fault features. In recent years, the local binary pattern (LBP) and the one-dimensional local binary pattern (1D-LBP) method extended on the basis of LBP provide a new idea for fault feature extraction. The 1D-LBP algorithm can realize the mining of equipment running state information from the perspective of texture feature extraction. However, the current research based on 1D-LBP algorithm mainly focuses on directly classifying the acquired vibration signals according to 1D-LBP algorithm, however, 1D-LBP algorithm itself is sensitive to noise, and small changes in neighborhood elements can cause great differences in local structure. A large amount of noise inevitably exists in engineering practice, which greatly increases the difficulty of texture feature extraction based on 1D-LBP algorithm.

[0004] Therefore, how to improve the noise resistance of 1D-LBP, realize accurate quantization of signals, and accurately extract fault information from multiple angles has become a problem to be solved. SUMMARY

[0005] In view of this, the present application provides a rotating machinery rubbing fault recognition method to solve the problems existing in the prior art.

[0006] The rotating machinery rubbing fault recognition method provided by the present application comprises:

[0007] S1: acquiring a vibration signal U of a rotating machinery, U=(u1, u2, u3, … uN), wherein N represents the length of the vibration signal; N ).

[0008] S2: using the vibration signal U, constructing a Hankel matrix H of the vibration signal U with 1 as the delay step length between each row vector m×n :

[0009]

[0010]

[0011] wherein u (i,j) = u i+j-1 , wherein i = 1, 2, 3…, m; j = 1, 2, 3…, n; N = m + n - 1; n < N, m and n are the number of rows and columns of the Hankel matrix H m×n respectively;

[0012] S3: calculating the covariance matrix C m×n of the Hankel matrix H n×n ;

[0013] S4: calculating the mean of each column of the covariance matrix C n×n and constructing a reconstruction signal R using the mean, wherein R = [R(1), R(2), …R(J), …R(n)], J = 1, 2, 3…, n, R(J) is the mean of the Jth column of C n×n ;

[0014] S5: according to the 1D-LBP method, performing binary quantization on the reconstruction signal R according to the local mean of the signal in the window to obtain a local texture signal LTS, LTS = [LTS(1), LTS(2), … LTS(p), … LTS(n-k+1)], wherein k is the window length of the moving window, k = 2, 3, … 256, p = 1, 2, … n-k+1, n is the length of the reconstruction signal R;

[0015] S6: performing spectral analysis on the local texture signal LTS and extracting the prominent frequency components in the spectrum to extract the rotating machinery rubbing fault feature and identify the fault.

[0016] Preferably, the step of calculating the covariance matrix C m×n of the Hankel matrix H n×n in S3 is as follows:

[0017] S31: calculating the mean of each column of the Hankel matrix H m×n and constructing a vector V using the mean of each column wherein the mean of the jth column of the Hankel matrix H m×n is calculated according to the following formula: ​

[0018]

[0019] The vector The formula is as follows:

[0020]

[0021] In the formula, u (i,j) It is the Hankel matrix H m×n The data in the i-th row and j-th column, where i = 1, 2, 3, ..., m; j = 1, 2, 3, ..., n;

[0022] S32: For the Hankel matrix H m×n Each element is decentralized, as shown in the following formula:

[0023]

[0024] In the formula, u′ (i,j) It is the Hankel matrix H m×n element u in row i and column j (i,j) Decentralized data;

[0025] S33: Construct the covariance matrix C n×n ,in,

[0026]

[0027] The covariance matrix C n×n The data in row I and column J of the middle (I,J) The calculation formula is as follows:

[0028]

[0029] In the formula, u′ (i,I) The Hankel matrix H m×n The decentralized data in the i-th row and i-th column, u′ (i,J) The Hankel matrix H m×n The data in the i-th row and J-th column is the decentralized data, where I = 1, 2, ..., n and J = 1, 2, ..., n.

[0030] Further optimization, in S5, the calculation method for the local texture signal LTS is as follows:

[0031] S51: Calculate the reconstructed signal R within the moving window. (p,q) mean r p and the reconstructed signal R within the moving window, signal r (p,q) With the moving window signal r (p,q) mean r pThe difference b (p,q) Among them, the signal r within the moving window (p,q) mean r p The calculation formula is as follows:

[0032]

[0033] Difference b (p,q) The calculation formula is as follows:

[0034] b (p,q) =r (p,q) -r p ;

[0035] In the formula, r (p,q) Let q be the vibration signal within the p-th moving window, k be the window length of the moving window, k = 2, 3, ... 256, q = 1, 2, ... k, p = 1, 2, ... n-k+1, and n be the length of the reconstructed signal R.

[0036] S52: Apply the 1D-LBP method to the difference b (p,q) Perform binarization to obtain the binarized signal y(b) (p,q) ), where when b (p,q) >0, y(b (p,q) ) equals 1; when b (p,q) ≤0, y(b) (p,q) ) equals 0;

[0037] S53: Convert the binarized signal y(b) (p,q) The data is then converted back to a decimal sequence to obtain the local texture signal LTS, where LTS = [LTS(1), LTS(2), ..., LTS(p), ..., LTS(n-k+1)]. Where p = 1, 2, ..., n-k+1.

[0038] The rotating machinery collision and rubbing fault identification method provided by this invention constructs a Hankel matrix of the signal before extracting texture features from the collision and rubbing fault using the 1D-LBP method. The covariance matrix of the Hankel matrix is ​​used to enhance the collision and rubbing fault information and reduce noise. Simultaneously, when selecting the effective component signals of the covariance matrix, the method no longer relies on signal evaluation metrics in conventional methods, but instead constructs a new signal sequence based on the mean of each column of the covariance matrix. By combining the newly obtained signal sequence with the 1D-LBP algorithm, feature extraction and fault identification of the collision and rubbing fault are achieved from the perspective of texture feature extraction.

[0039] The rotating machinery collision and rubbing fault identification method provided by this invention integrates one-dimensional local binary mode and Hankel matrix covariance matrix. By establishing the covariance matrix of Hankel matrix, the one-dimensional time series is extended to a high-dimensional space, which can highlight periodic features while reducing noise. By constructing a new signal sequence using the average value of each column element in the covariance matrix, the uncertainty problem of selecting effective component signals based on signal evaluation indicators can be solved. By combining 1D-LBP with the covariance matrix of Hankel matrix, the traditional strategy of quantizing "overall vibration signal" is transformed into the method of quantizing the signal Hankel matrix covariance matrix when quantizing the signal. This can solve the problem of 1D-LBP being susceptible to noise and further highlight weak fault information. By introducing 1D-LBP into collision and rubbing fault identification, collision and rubbing fault information can be extracted from the perspective of local texture feature extraction, and collision and rubbing faults can be accurately identified based on the 1D-LBP method. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 A flowchart illustrating the rotating machinery rubbing fault identification method provided by the present invention;

[0042] Figure 2 The corresponding signal diagrams in the process of identifying collision faults using the method provided by the present invention are shown in the figure. In the figure, (a) and (b) are the original vibration signal U and its spectrum, respectively; (c) and (d) are the time domain and spectrum of the reconstructed signal R obtained based on the mean of the covariance matrix, respectively; (e) and (f) are the time domain and spectrum of the signal LTS obtained after binary quantization of the reconstructed signal R according to the 1D-LBP method, respectively; and (g) is a partial enlarged view of (f).

[0043] Figure 3 This figure shows the uncertainty results of the classic method for determining the effective components of the covariance matrix using parametric indices. In the figure, (a) and (b) represent the results for... Figure 2 (a) shows the original vibration signal, which is then filtered for effective components of the covariance matrix based on information entropy and 1D-LBP is performed based on the reconstructed signal. (c) and (d) show the signal time domain and spectrum obtained by filtering for effective components of the covariance matrix based on waveform factors and performing 1D-LBP based on the reconstruction, respectively. (e) is a magnified view of (d). Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Here, it should be noted that to avoid obscuring the present invention due to unnecessary details, only the processing steps closely related to the solution of the present invention are shown in the drawings, while other details less relevant to the present invention are omitted.

[0045] The present invention provides a method for identifying the rubbing fault of a rotating machine, comprising the following steps:

[0046] S1: Obtain the vibration signal U of the rotating machine, U=(u 1, u 2, u 3, …u N ), where N represents the length of the vibration signal;

[0047] S2: Using the vibration signal U, construct the Hankel matrix H of the vibration signal U with a delay step of 1 between each row vector m×n :

[0048]

[0049] where u (i,j) =u i+j-1 , where i = 1, 2, 3…, m; j = 1, 2, 3…, n; N = m + n - 1; n < N, and m and n are the number of rows and columns of the Hankel matrix H m×n respectively;

[0050] S3: Calculate the covariance matrix C m×n of the Hankel matrix H n×n ;

[0051] S4: Calculate the mean of each column in the covariance matrix C n×n and construct the reconstructed signal R, R = [R(1), R(2), …R(J), …R(n)], J = 1, 2, 3…, n, and R(J) is equal to the mean of the Jth column in C n×n . The calculation formula of R(J) is as follows:

[0052]

[0053] [[ID=??]]where c (In,J) represents the data in the Ith row and Jth column of the covariance matrix C n×n ;

[0054] S5: Using the local mean of the signal within the window as the criterion, the reconstructed signal R is binarized according to the 1D-LBP method to obtain the local texture signal LTS, LTS=[LTS(1),LTS(2),…LTS(p),…LTS(n-k+1)], where k is the window length of the moving window, k=2,3,…256, p=1,2,…n-k+1, and n is the length of the reconstructed signal R;

[0055] S6: Perform spectral analysis on the local texture signal LTS and extract prominent frequency components in the spectrum to extract rotating machinery collision fault features and identify faults.

[0056] In S3, the Hankel matrix H is calculated. m×n The covariance matrix C n×n The specific steps are as follows:

[0057] S31: Calculate the Hankel matrix H m×n Find the mean of each column and construct a vector using the mean of each column. Wherein, the Hankel matrix H m×n Mean of column j The calculation formula is as follows:

[0058]

[0059] The vector The formula is as follows:

[0060]

[0061] In the formula, u (i,j) It is the Hankel matrix H m×n The data in the i-th row and j-th column, i = 1, 2, 3, ..., m; j = 1, 2, 3, ..., n;

[0062] S32: For the Hankel matrix H m×n Each element is decentralized, as shown in the following formula:

[0063]

[0064] In the formula, u′ (i,j) It is the Hankel matrix H m×n element u in row i and column j (i,j) Decentralized data;

[0065] S33: Construct the covariance matrix C n×n ,in,

[0066]

[0067] The covariance matrix C n×n The data in row I and column J of the middle (I,J) The calculation formula is as follows:

[0068]

[0069] In the formula, u′ (i,I) The Hankel matrix H m×n The decentralized data in the i-th row and i-th column, u′ (i,J) The Hankel matrix H m×n The data in the i-th row and J-th column is the decentralized data, where I = 1, 2, ..., n and J = 1, 2, ..., n.

[0070] In S5, the calculation method for the local texture signal LTS is as follows:

[0071] S51: Calculate the reconstructed signal R within the moving window. (p,q) mean r p and the reconstructed signal R within the moving window, signal r (p,q) With the moving window signal r (p,q) mean r p The difference b (p,q) Among them, the signal r within the moving window (p,q) mean r p The calculation formula is as follows:

[0072]

[0073] Difference b (p,q) The calculation formula is as follows:

[0074] b (p,q) =r (p,q) -r p ;

[0075] In the formula, r (p,q) Let q be the vibration signal within the p-th moving window, k be the window length of the moving window, k = 2, 3, ... 256, q = 1, 2, ... k, p = 1, 2, ... n-k+1, and n be the length of the reconstructed signal R.

[0076] S52: Apply the 1D-LBP method to the difference b (p,q) Perform binarization to obtain the binarized signal y(b) (p,q) ), where when b (p,q) >0, y(b (p,q) ) equals 1; when b (p,q) ≤0, y(b) (p,q) ) equals 0;

[0077]

[0078] S53: Convert the binarized signal y(b) (p,q) The signal is converted back to a decimal sequence to obtain the local texture signal LTS, which is used to characterize the local features of the vibration signal. LTS = [LTS(1), LTS(2), ... LTS(p), ... LTS(n-k+1)] Where p = 1, 2, ..., n-k+1.

[0079] Example 1:

[0080] To demonstrate the effectiveness of the rotating machinery collision fault identification method proposed in this invention, the collision faults listed in Table 1 were randomly selected for analysis. In this data set, the collision location was vertically above the turbine casing, the accelerometer was installed vertically below the turbine casing, the collision severity was moderate, the rotational speed was 1515.96 r / min, and the corresponding rotational frequency was 25.27 Hz. r =1515.96 / 60=25.27).

[0081] Table 1 Experimental Data Information Table

[0082]

[0083] The vibration signal of the rotating machinery in Case 1 is obtained and processed using the method proposed in this invention, wherein in S5, the length of the moving window is k = 8.

[0084] Among them, such as Figure 2 As shown, (a) and (b) are the original vibration signal U and its spectrum, respectively; (c) and (d) are the time domain and spectrum of the reconstructed signal R obtained based on the mean of the covariance matrix, respectively; (e) and (f) are the time domain and spectrum of the signal LTS obtained after binary quantization of the reconstructed signal R according to the 1D-LBP method, respectively; and (g) is a local magnified view of (f). In this embodiment, the spectrum of the signal has been normalized.

[0085] analyze Figure 2 It can be observed that:

[0086] (1) The spectrum of the original vibration signal Figure 2 (b) High-frequency band features are very prominent, while low-frequency band features are relatively weak. In engineering practice, when monitoring the operating status of equipment such as aero engines, identification and judgment are usually based on the characteristics of the low-frequency band of the signal; therefore, directly monitoring the collision fault based on the signal spectrum is very likely to result in missed diagnosis.

[0087] (2) The spectrum of the signal constructed based on the mean of the covariance matrix.Figure 2 (d) reduces noise interference; at the same time, the relative amplitude of the low-frequency components is significantly increased; however, only 2f can be extracted. r The frequency component (51.27Hz) makes it difficult to accurately determine the fault type;

[0088] (3) The spectrum of the signal obtained after binary quantization of the reconstructed signal R according to the 1D-LBP method ( Figure 2 (g) The low-frequency feature information is further enriched, and the frequency components related to the collision fault are extracted, including 2f r (51.27Hz), 5.5f r (136.7Hz), 10f r (249Hz), 14f r (349.1Hz) and 19f r (481Hz) Based on the integer and fractional harmonics of these frequencies, it can be determined that the equipment has experienced a rubbing fault.

[0089] Combining (1)-(3), it can be found that the method of constructing the signal based on the mean of the covariance matrix can effectively suppress noise and highlight the low-frequency characteristics of the signal. Combining it with 1D-LBP can further enrich the fault characteristic information in the low-frequency band of the spectrum, realize the accurate identification of collision faults, and benefit practical engineering applications.

[0090] The main reason for this is that the signal sequence constructed based on the mean of the column vectors of the covariance matrix reduces the impact of noise and highlights periodic characteristics, while avoiding the randomness problem of selecting component signals based on signal evaluation indicators. At the same time, the 1D-LBP method can better characterize the local texture features of the signal, thus achieving accurate extraction of equipment operating status information.

[0091] Example 2:

[0092] To further demonstrate the effectiveness and superiority of the method proposed in this invention, the rubbing failures in Table 1 are analyzed as an example. Figure 3 The uncertainty results of the classic method for determining the effective components of the covariance matrix using parametric indices are shown. The characteristic parameter indices selected are the representative information entropy and waveform factor. Only when selecting component signals is the signal evaluation index used (in this invention, the mean of the column vectors is taken); the remaining steps are completely consistent with the method proposed in this invention. For comparative verification and analysis, the rubbing fault in Table 1 of Example 1 is still selected as an example for analysis, and the results are as follows... Figure 3 As shown. Figure 3 In the text, (a) and (b) represent the... Figure 2(a) shows the original vibration signal, which is then filtered for effective components of the covariance matrix based on information entropy and 1D-LBP is performed based on the reconstructed signal. (c) and (d) show the signal time domain and spectrum obtained by filtering for effective components of the covariance matrix based on waveform factors and performing 1D-LBP based on the reconstruction, respectively. (e) is a magnified view of (d).

[0093] analyze Figure 3 It can be observed that:

[0094] (1) When information entropy is used as a parameter, the resulting signal spectrum ( Figure 3 (b) It is difficult to extract the harmonic or fractional harmonic components of the frequency conversion in the low frequency band, and it is also difficult to accurately determine the rubbing fault.

[0095] (2) When the waveform factor is used as a parameter index, the resulting signal spectrum ( Figure 3 (d), (e)), in the low-frequency band, some feature information can be extracted, including 2f r (51.27Hz), 3f r (75.68Hz), 5f r (122.1Hz);

[0096] Combining (1) and (2), it can be found that since the manifestation of fault characteristics in the signal is not certain and has a certain degree of randomness, there will be great uncertainty when selecting component signals based on signal evaluation indicators that reflect certain characteristic information of fault signals.

[0097] contrast Figure 2 (g) and Figure 3 (b) and (e) show that the signal construction method based on the mean of the covariance matrix proposed in this invention can reflect the characteristics of the average level of the signal. Compared with the method of selecting effective components for reconstruction based on signal evaluation indicators such as information entropy and waveform factors, it solves the uncertainty problem when selecting sensitive fault component signals based on signal evaluation indicators, and can extract richer low-frequency information.

[0098] The method proposed in this invention can effectively suppress noise and highlight the low-frequency features of the signal; reduce the uncertainty of component selection; and extract collision and rubbing fault information from the perspective of texture feature extraction, thereby achieving accurate identification of collision and rubbing faults.

[0099] It should be noted that the purpose of disclosing the embodiments is to help further understand the present invention; however, those skilled in the art will understand that various substitutions and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the content disclosed in the embodiments, and the scope of protection of the present invention is defined by the scope of the claims.

Claims

1. A rotating machinery rub fault identification method characterized by, Comprising: S1: acquiring a vibration signal of a rotating machine , wherein, denotes the length of the vibration signal; S2: exploiting the vibration signal , with 1 being the delay step size between each row vector, a Hankel matrix of the vibration signal :​ ； In the formula, = wherein, ; n < N, m, n are the row number and column number of the Hankel matrix respectively.​ S3: compute the Hankel matrix covariance matrix of ; S4: computing the covariance matrix the mean of each column and constructing a reconstructed signal using the means wherein, , , is the mean of the first column. S5: according to the 1D-LBP method, the reconstructed signal is binarized with the local mean of the windowed signal as the criterion to obtain a local texture signal , wherein, is the window length of the moving window, , , is the length of the . S6: performing a spectral analysis on the local texture signal Spectrum analysis is performed and the prominent frequency components in the spectrum are extracted for rotating machinery rub-impact fault feature extraction and fault recognition.

2. The rotating machine rub fault recognition method according to claim 1, characterized by: S3 computing the Hankel matrix covariance matrix The steps are as follows: S31: calculating the Hankel matrix Column means are computed and vectors are constructed using the column means where the Hankel matrix The mean of the jth column The formula for the calculation is as follows: ； The vector The formula is as follows: ； wherein is the ith row, jth column data of the Hankel matrix ;​ S32: Decenter each element of the Hankel matrix with the formula: ； In the formula, is the ith row, jth column element of the Hankel matrix decentralized data;​ S33: Constructing the covariance matrix wherein, ； The covariance matrix In the middle of the In the first column of the The data in the first column The calculation formula is as follows: ； wherein is the decentered data of the Hankel matrix is the decentered data of the Hankel matrix is the decentered data of the Hankel matrix is the decentered data of the Hankel matrix is the decentered data of the Hankel matrix is the decentered data of the Hankel matrix , .

3. The rotating machine rub fault recognition method of claim 1, wherein: In S5, the local texture signal is calculated as follows: S51: calculating the reconstructed signal the mean of the signals in the moving window the reconstructed signal the mean of the signals in the moving window the difference between the reconstructed signal the mean of the signals in the moving window wherein the mean of the signals in the moving window is calculated as follows:​​​ ； Difference The calculation formula is as follows: ； wherein is the first vibration signal in the first moving window, is the window length of the moving window, , , , is the length of the reconstructed signal . S52: binarizing the difference values according to the 1D-LBP method to obtain a binarized signal wherein, when , is equal to 1; when , is equal to 0; S53: re-converting the binary signal into a decimal sequence, resulting in a local texture signal , , wherein, .​

Images