Bearing fault diagnosis method based on one-dimensional local binary pattern and Hankel matrix

Abstract

The invention provides a bearing fault diagnosis method based on a one-dimensional local binary pattern and a Hankel matrix. The bearing fault diagnosis method comprises the following steps: acquiring a discrete vibration signal; performing first-order differential operation on the discrete vibration signal to obtain a differential signal; performing inherent time scale decomposition on the differential signal to obtain an inherent rotation component signal; performing quantization and signal reconstruction on each inherent rotation component signal by taking a root mean square as a quantization criterion of a one-dimensional local binary mode method to obtain a decimal feature signal; constructing a Hankel matrix of the decimal characteristic signal and performing signal reconstruction according to a covariance matrix of the Hankel matrix; performing spectral analysis on the reconstructed signal, calculating the fault characteristic frequency of the bearing, and then judging the state and the fault type of the bearing through a frequency component obtained through spectral analysis and the fault characteristic frequency of the bearing obtained through calculation. According to the bearing fault diagnosis method, noise can be effectively suppressed, the bearing fault feature information can be effectively extracted, and the bearing state and the fault type can be accurately identified.

Description

Technical Field

[0001] This invention belongs to the field of fault diagnosis technology, and specifically relates to a bearing fault diagnosis method based on one-dimensional local binary pattern (1D-LBP) and Hankel matrix. Background Technology

[0002] Bearings possess advantages such as high versatility and low power loss, effectively ensuring the reliability and stability of mechanical equipment during operation and playing a crucial role in the equipment's operation. Because large equipment typically operates in harsh environments with large load variations, bearings are prone to damage during operation, which can lead to malfunctions of the entire machine, causing significant economic losses and safety accidents. Therefore, monitoring the operating status of bearings and promptly and accurately detecting faults is of great importance in preventing major safety accidents, minimizing economic losses, and ensuring personnel safety.

[0003] When a bearing fails, its vibration signal exhibits strong nonlinearity and non-stationarity. Signal decomposition algorithms can decompose the signal into multiple component signals from the time domain, frequency domain, and time-frequency domain perspectives, which helps extract fault information and is currently widely used in bearing fault identification. Common signal decomposition algorithms, such as wavelet transform, empirical mode decomposition, intrinsic time-scale decomposition (ITD), and variational mode decomposition, each have their advantages but also limitations. Among them, the ITD method is widely used in bearing fault detection because it avoids some common problems of other algorithms. However, in practical engineering, due to the weakness, complexity, and composite nature of bearing fault characteristics, it is difficult to accurately identify its state by simply relying on signal decomposition. Therefore, morphological filtering and singular value decomposition based on the Hankel matrix are used to extract fault information from the signal. Suppression. Furthermore, researchers have also attempted to characterize bearing fault information using the local binary pattern (LBP) method from the perspective of texture feature extraction. The LBP algorithm has significant advantages in image processing; after its introduction into signal processing, the method based on one-dimensional local binary pattern (1D-LBP) has improved the accuracy of bearing fault identification and classification.

[0004] However, the classic 1D-LBP method is susceptible to noise interference, and it does not fully consider the characteristic patterns of signals exhibited during bearing faults when quantizing the signals; this leads to inaccurate signal quantization and an inability to accurately represent fault information. Therefore, how to improve the noise immunity of 1D-LBP, achieve accurate signal quantization, and accurately extract fault information from multiple perspectives has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, the present invention proposes a bearing fault diagnosis method based on one-dimensional local binary mode and Hankel matrix to solve the problems existing in the prior art.

[0006] The bearing fault diagnosis method proposed in this invention, based on one-dimensional local binary mode and Hankel matrix, includes:

[0007] S1: Acquire discrete vibration signals ;

[0008] S2: For the discrete vibration signal Perform first-order difference operation to obtain the difference signal. ;

[0009] S3: For the differential signal The number of decomposition layers is: The inherent timescale decomposition yields An inherent rotation component signal ,in, It is a positive integer;

[0010] S4: Utilize the improved one-dimensional local binary mode method for each intrinsic rotation component signal Quantization and signal reconstruction are performed to obtain the signal corresponding to the inherent rotation component. One-to-one correspondence decimal feature signal The improved one-dimensional local binary mode method uses the root mean square of the signal within the window as the quantization criterion.

[0011] S5: Construct the decimal feature signal The Hankel matrix is ​​obtained and its covariance matrix is ​​calculated. Then, the signal is reconstructed based on the covariance matrix to obtain the reconstructed signal.

[0012] S6: Perform spectral analysis on the reconstructed signal to obtain frequency components. Calculate the bearing fault characteristic frequency based on the bearing geometric parameters and rotational speed. Then, determine the bearing condition and fault type by comparing the obtained frequency components with the calculated bearing fault characteristic frequency.

[0013] Further optimization, in S4, for the first... An inherent rotation component signal The steps for quantization and signal reconstruction are as follows, where, ;

[0014] S41: Move the sliding window in steps of 1 at the... An inherent rotation component signal Move the signal up and down, and calculate the root mean square value of the signal within each sliding window. , among which, the The root mean square value of the signal within a sliding window Calculated using the following formula:

[0015] ;

[0016] In the formula, The length of the sliding window. , , for Signal length;

[0017] S42: Calculate the difference between each element within each sliding window and the root mean square value of the signal within that sliding window. And based on the difference, obtain a binary encoding sequence that corresponds one-to-one with each element in each sliding window. , among which, the The first sliding window The elements and the root mean square value of the signal within the sliding window The difference Calculated using the following formula:

[0018] ;

[0019] In the formula, For the first The first sliding window One element, ;

[0020] Among them, the The binary encoding sequence corresponding to each sliding window elements in Calculated using the following formula:

[0021] ;

[0022] or ;

[0023] In the formula, Indicates the first The first sliding window The binary code corresponding to each element;

[0024] S43: The binary encoding sequence corresponding to the signal within each sliding window. All are converted into decimal sequences to obtain the characteristic signal corresponding to the signal within each sliding window, where the first... The characteristic signal corresponding to the signal within each sliding window Calculate using the following formula;

[0025] ;

[0026] Corresponding decimal characteristic signal ... , The length of the decimal characteristic signal. .

[0027] Further optimization, in S5, utilizing the first Decimal characteristic signal The method for signal reconstruction is as follows:

[0028] S51: Construct the decimal feature signal Hankel matrix :

[0029]

[0030] ;

[0031] In the formula: ,in, ; , The Hankel matrix is ​​respectively The number of rows and columns;

[0032] S52: Calculate the Hankel matrix Find the mean of each column and construct a vector using the mean of each column. The Hankel matrix Mean of column j The calculation formula is as follows:

[0033] ;

[0034] The vector The formula is as follows:

[0035] ;

[0036] S53: For the Hankel matrix Each element is decentralized, as shown in the following formula:

[0037] ;

[0038] In the formula, It is the Hankel matrix element in row i and column j Decentralized data;

[0039] S54: Constructing the covariance matrix :

[0040] ;

[0041] Wherein, the covariance matrix The Middle Line number Column data The calculation formula is as follows:

[0042] ;

[0043] In the formula, The Hankel matrix The The decentralized data in row I and column I. The Hankel matrix The The decentralized data in row J, where... ;

[0044] S55: Using the covariance matrix Perform signal reconstruction to obtain the reconstructed signal. .

[0045] Further optimization utilizes the covariance matrix Perform signal reconstruction to obtain the reconstructed signal. The method is as follows: The covariance matrix... Add the first row and the last column, or add the covariance matrix. The first and last rows are added together, or the diagonal average method is used.

[0046] This invention provides a bearing fault diagnosis method based on one-dimensional local binary mode and Hankel matrix. It uses a first-order differential signal to replace the original signal for ITD (Integrated Transmission Diagnosis), highlighting the impact characteristics of bearing faults. By replacing local intermediate values ​​with the root mean square of the local signal within the window, the accuracy of quantization of the vibration signal using the 1D-LBP algorithm is improved. Furthermore, by replacing the original signal with the inherent rotational component signal for quantization, the accuracy of quantization is further improved, achieving a fusion of the 1D-LBP and ITD algorithms. The covariance matrix of the Hankel matrix reflects the correlation between different dimensions of the signal. By constructing the covariance matrix of the Hankel matrix of the reconstructed decimal signal after quantization for each component, effective noise suppression and the fusion of phase space reconstruction theory with 1D-LBP are achieved, enabling accurate extraction of bearing fault information. Attached Figure Description

[0047] 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.

[0048] Figure 1 This is a flowchart of the bearing fault diagnosis method based on one-dimensional local binary mode and Hankel matrix proposed in this invention.

[0049] Figure 2 The figure shows the results of diagnosing bearing faults using the method proposed in this invention. In the figure, (a) and (b) are the time-domain plots and spectra of the simulated signal, respectively; (c) and (d) are the first-order differential signal and its spectrum, respectively; (e) is the intrinsic rotational component (PRC) obtained by performing ITD on the first-order differential signal; (f) is the spectrum of each component in (e); (g) is the reconstructed component obtained by quantizing each PRC in (e) according to the proposed 1D-LBP boosting method; and (h) is the spectrum of (g). The covariance matrix of the Hankel matrix of each PRC is constructed and the signal is reconstructed to obtain the autocorrelation function of each component. - The spectrum of the components is shown in (i)-(l). In (i)-(l), the gray spectral lines are consistent with the spectrum corresponding to (f), and the green lines represent the spectrum of each component after processing by the method proposed in this invention.

[0050] Figure 3The figure shows the results of diagnosing a single bearing fault using the method proposed in this invention. In the figure, (a) and (b) are the time-domain plot and spectrum of the acceleration signal when the bearing has a single outer race fault, respectively; (c) is the first-order differential signal of (a); (d) is the spectrum of (c); and (e) is the result of performing ITD on the first-order differential signal. (f) is the spectrum of (e), and (g) is the spectrum obtained by PRC after the proposed 1D-LBP enhancement process. The components, (h) is the spectrum of (g), and the green spectral lines in (i)-(j) are respectively The autocorrelation function components are obtained by constructing the covariance matrix of the Hankel matrix and then reconstructing it. - The spectrum of , the gray spectral line is the spectrum in (f);

[0051] Figure 4 The figure shows the results of diagnosing bearing complex faults using the method proposed in this invention. In the figure, (a) is the time domain diagram of the acceleration signal; (b) is the spectrum diagram of the acceleration signal; (c) is the first-order differential signal of (a); (d) is the spectrum of (c); (e) is the PRC obtained by performing ITD on the first-order differential signal; and (g) is the PRC component obtained by 1D-LBP enhancement. The components are (h) and (g), respectively, and (i)-(k) are the components corresponding to the method of the present invention. - The spectrum of (i)-(k); the green spectral lines in (i)-(k) are the spectra of each component after processing by the method proposed in this invention; the gray spectral lines are the spectrum in (h). Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, to avoid obscuring the invention with unnecessary details, only processing steps closely related to the solution of this invention are shown in the drawings, while other details not closely related to this invention are omitted.

[0053] To address the challenges of accurately extracting bearing fault features under strong noise conditions and the susceptibility to strong noise interference and inaccurate quantization in 1D-LBP, which uses intermediate values ​​as thresholds for signal quantization within a window, this invention proposes a novel bearing fault diagnosis method. By replacing local intermediate values ​​with the root mean square (RMS) of the signal, the quantization criterion of the 1D-LBP algorithm is improved, resulting in an enhanced 1D-LBP method. From the perspective of one-dimensional local feature extraction and phase space reconstruction, the method combines the ITD (Integrated Trace Value), the covariance matrix of the Hankel matrix, and the enhanced 1D-LBP to extract feature information related to bearing faults, thereby achieving accurate judgment of bearing condition and fault.

[0054] like Figure 1 As shown, this invention provides a bearing fault diagnosis method based on one-dimensional local binary mode and Hankel matrix, including the following steps:

[0055] S1: Acquire discrete vibration signals , ,in, Indicates the sequence number of the sampling point. , The length of the vibration signal;

[0056] S2: For the discrete vibration signal Perform first-order difference operation to obtain the difference signal. ;

[0057] Taking the first-order forward difference as an example, , ;

[0058] S3: For the differential signal The number of decomposition layers is: The inherent time-scale decomposition (ITD) yields... A proper rotation component (PRC) signal ,in, For positive integers, the first... An inherent rotation component signal , , for Signal length;

[0059] S4: Use the improved 1D-LBP method (denoted as ILBP) for each intrinsic rotational component signal Quantization and signal reconstruction are performed to obtain the signal corresponding to the inherent rotation component. One-to-one correspondence decimal feature signal The improved 1D-LBP method uses the root mean square of the signal within the window (instead of the signal median) as the quantization criterion.

[0060] S5: Construct the decimal feature signal The Hankel matrix is ​​obtained and its covariance matrix is ​​calculated. Then, the signal is reconstructed based on the covariance matrix to obtain the reconstructed signal.

[0061] S6: Perform spectral analysis on the reconstructed signal to obtain frequency components. Calculate the bearing fault characteristic frequency based on the bearing geometric parameters and rotational speed. Then, determine the bearing condition and fault type by comparing the obtained frequency components with the calculated bearing fault characteristic frequency.

[0062] like Figure 2 As shown, in S4, for the first... An inherent rotation component signal The steps for signal quantization and reconstruction are as follows:

[0063] S41: The sliding window is moved according to the inherent rotation component signal with a step size of 1. Move the signal up and down, and calculate the root mean square value of the signal within each sliding window. , among which, the The root mean square value of the signal within a sliding window Calculated using the following formula:

[0064] ;

[0065] In the formula, The length of the sliding window. , , for Signal length;

[0066] S42: Calculate the difference between each element within each sliding window and the root mean square value of the signal within that sliding window. And based on the difference, obtain a binary encoding sequence that corresponds one-to-one with each element in each sliding window. , among which, the The first sliding window The elements and the root mean square value of the signal within the sliding window The difference Calculated using the following formula:

[0067] ;

[0068] In the formula, For the first The first sliding window One element, ;

[0069] Among them, the The binary encoding sequence corresponding to each sliding window elements in Calculated using the following formula:

[0070] ;

[0071] or ;

[0072] In the formula, Indicates the first The first sliding window The binary code corresponding to each element;

[0073] S43: The binary encoding sequence corresponding to the signal within each sliding window. All are converted into decimal sequences to obtain the characteristic signal corresponding to the signal within each sliding window, where the first... The characteristic signal corresponding to the signal within each sliding window Calculate using the following formula;

[0074] ;

[0075] Corresponding decimal characteristic signal The length of the decimal characteristic signal. .

[0076] In S5, using the first Decimal characteristic signal The method for signal reconstruction is as follows:

[0077] S51: Construct the decimal feature signal Hankel matrix :

[0078]

[0079] ;

[0080] In the formula: ,in, ; , The Hankel matrix is ​​respectively The number of rows and columns;

[0081] S52: Calculate the Hankel matrix Find the mean of each column and construct a vector using the mean of each column. The Hankel matrix Mean of column j The calculation formula is as follows:

[0082] ;

[0083] The vector The formula is as follows:

[0084] ;

[0085] S53: For the Hankel matrix Each element is decentralized, as shown in the following formula:

[0086] ;

[0087] In the formula, It is the Hankel matrix element in row i and column j Decentralized data;

[0088] S54: Constructing the covariance matrix :

[0089] ;

[0090] Wherein, the covariance matrix The Middle Line number Column data The calculation formula is as follows:

[0091] ;

[0092] In the formula, The Hankel matrix The The decentralized data in row I and column I. The Hankel matrix The The decentralized data in row J, where... ;

[0093] S55: Using the covariance matrix Perform signal reconstruction to obtain the reconstructed signal. .

[0094] Among them, the covariance matrix is ​​used Perform signal reconstruction to obtain the reconstructed signal. The method is as follows: The covariance matrix... Add the first row and the last column, or add the covariance matrix. The first and last rows are added together, or the diagonal average method is used.

[0095] Example 1

[0096] To demonstrate the effectiveness of the bearing fault diagnosis method based on one-dimensional local binary mode and Hankel matrix proposed in this invention, the bearing fault is first simulated by establishing a mathematical model of the bearing fault. In this embodiment, all amplitude information of the parameters are typical parameters that are randomly set.

[0097] ;

[0099] in, ; It is the displacement constant. ; It is the damping coefficient. ; The system's inherent frequency, Hz; the impact period T=0.01s, i.e., the fault characteristic frequency. =100 Hz; Indicates frequency conversion. Rotation speed ; For modulation frequency, =85 Hz.

[0100] To simulate a noisy environment, random noise was added. The random noise amplitude is 5. The final simulation signal is obtained. for:

[0101] ;

[0102] The simulation signal from Example 1 is obtained and processed using the method proposed in this invention. Specifically, in S3, the number of ITD decomposition layers P is 4; in S4, the moving window length q = 8; and in S5, the signal reconstruction method involves using the covariance matrix... Add the first row and the last column together.

[0103] Among them, such as Figure 2 As shown, (a) and (b) are the time-domain plots and spectra of the simulated signal, respectively; (c) and (d) are the first-order differential signal and its spectrum, respectively; (e) is the PRC obtained by performing ITD on the first-order differential signal (in this embodiment of the invention, the number of decomposition layers is equal to 4). (f) is the spectrum of each component in (e); (g) is the reconstructed component obtained by quantizing each PRC in (e) according to the proposed 1D-LBP boosting method; (h) is the spectrum of (g). The covariance matrix of the Hankel matrix of each PRC is constructed and the signal is reconstructed to obtain the autocorrelation function of each component. The spectrum of the spectrum is shown in (i)-(l). In (i)-(l), the gray spectral lines are consistent with the spectrum corresponding to (f), and the green lines represent the spectrum of each component after processing by the method proposed in this invention.

[0104] analyze Figure 2 It can be observed that:

[0105] (1) In the spectrum of the simulated signal ( Figure 2 In (b), the set fault characteristic frequencies can be extracted, including 84.91 Hz + , 184.94 Hz (184.94- =99.94 Hz), etc. However, the amplitudes of these frequency components are very low, and the frequencies that are more prominent in the spectrum are at the natural frequencies ( ). The frequency components near Hz are not conducive to accurate fault diagnosis.

[0106] (2) Analysis Figure 2 (f) It can be found that the signal spectrum after first-order differential and ITD processing has a certain noise reduction effect, but the effect is not ideal and the characteristic frequency corresponding to the set fault cannot be extracted.

[0107] (3) Analysis Figure 2 (h) It can be observed that the component signal after 1D-LBP enhancement processing... In the spectrum, relatively obvious fault characteristic frequencies can be found, but there is still some noise interference;

[0108] (4) Analysis Figure 2 (i)-(l), the signal spectrum obtained by the method proposed in this invention (marked with green lines in the figure) is compared with the spectrum of each PRC after ITD of the differential signal (marked with gray lines in the figure). It can be found that after improving the 1D-LBP and covariance matrix processing of the Hankel matrix, the interference of noise and irrelevant components can be reduced very effectively. Moreover, the proposed method can extract typical and prominent fault characteristic frequencies and their harmonics, specifically:

[0109] 1) In Figure 2 In (i), the harmonics of the modulation frequency can be extracted, including: 2 , 169.92 Hz (169.92 / 2=84.96 Hz); 4 ,340.58 Hz (340.58 / 4=85.14 Hz);

[0110] 2) In Figure 2 In (j), the set fault characteristic frequency and its harmonics, as well as the harmonics of the modulated fault characteristic frequency, can be extracted, including: , 99.61 Hz; 2 , 199.95 Hz (199.95 / 2=99.98 Hz); 3 ,300.29 Hz (300.29 / 3=100.1 Hz); 4 , 399.1 Hz (399.1 / 4=99.78 Hz); 5 + 585.2 Hz ((585.2- () / 5 = 100.04 Hz); 10 + , 1085.45 Hz ((1085.45- () / 10=100.05Hz);

[0111] 3) In Figure 2 In (k), the harmonics of the modulated fault characteristic frequency can be extracted, including: 3 + ,385.25 Hz ((385.25- () / 3 = 100.08 Hz); 5 + , 585.2 Hz ((585.2- ) / 5=100.04Hz); 7 + , 785.21 Hz ((785.21- () / 7=100.03 Hz);

[0112] 4) In Figure 2 In (l), the harmonics of the modulated fault characteristic frequency can be extracted, including: 3 + ,385.25 Hz ((385.25- () / 3 = 100.08 Hz); 5 + , 585.2 Hz ((585.2- () / 5=100.04Hz).

[0113] Example 2

[0114] To further demonstrate the effectiveness and superiority of the method proposed in this invention, data 1 in Table 1 is randomly selected as an example to verify the effectiveness of the proposed method. Data 1 corresponds to the data collected by the acceleration sensor when a single outer ring failure occurs in the bearing. For simplicity, in Examples 2 and 3, data 1 is used as an example. Represents the rotation frequency, in Represents rotational speed; with Characteristic frequencies of bearing inner ring failure; Characteristic frequencies of bearing outer ring failure; ;by Bearing cage failure characteristic frequencies.

[0115] Rotational speed in Example 2 =1875 r / min. The characteristic frequencies of the bearing are shown in Table 1, specifically: =31.25Hz, =138.54 Hz, =80.2 Hz, =54.42 Hz, =11.46 Hz.

[0116] The vibration signal of a single outer ring fault in the bearing in Example 2 was obtained and processed using the method proposed in this invention. In Example 2, the number of ITD decomposition layers P is 3; in S4, the moving window length q = 8; and in S5, the signal reconstruction method involves reconstructing the covariance matrix... Add the first and last columns together.

[0117] Figure 3 In the diagram, (a) and (b) are the time-domain plots and spectrum diagrams of the acceleration signal when the bearing has a single outer ring fault, respectively; (c) is the first-order differential signal of (a), (d) is the spectrum of (c); and (e) is the spectrum obtained by performing ITD on the first-order differential signal. (This embodiment takes a decomposition layer of 3 as an example); (f) is the spectrum corresponding to (e); (g) is the spectrum obtained by PRC after the proposed 1D-LBP enhancement process. The components, (h) is the spectrum of (g); the green spectral lines in (i)-(j) are respectively The autocorrelation function components are obtained by constructing the covariance matrix of the Hankel matrix and then reconstructing it. The spectrum of is represented by the gray spectral line in (f).

[0118] Table 1 Typical Data Introduction

[0119]

[0120] analyze Figure 3 It can be observed that:

[0121] (1) Analysis Figure 3 (b) It can be observed that the characteristic frequency and its harmonics corresponding to the bearing outer race fault cannot be extracted from the spectrum of the original vibration signal. The most prominent frequency component in the spectrum is the harmonic of the rotational frequency (5). 11 The spectrum of the original signal contains irrelevant components (705.87 Hz), meaning that the type of bearing failure cannot be determined solely by analyzing the spectrum of the original signal.

[0122] (2) Analysis Figure 3 (h) It can be observed that the component signal after 1D-LBP enhancement processing... In the spectrum, relatively obvious bearing fault characteristic frequencies can be seen, but there is still some noise interference;

[0123] (3) Analysis Figure 3 (i)-(j), after further constructing the covariance matrix of the Hankel matrix, the autocorrelation function components are obtained. - From the spectrum, the characteristic frequencies and their harmonics (represented by green spectral lines) corresponding to the bearing outer ring fault can be extracted, specifically:

[0124] 1) In Figure 3 In (i), it is possible to extract , 79.96 Hz; 2 , 159.9 Hz (159.9 / 2=79.95Hz); 4 , 322.3 Hz (322.3 / 4=80.57 Hz);

[0125] 2) In Figure 3 In (j), 14 can be extracted. , 1124 Hz (1124 / 14=80.28 Hz); 15 , 1205Hz (1205 / 15=80.33 Hz);

[0126] 3) Simultaneously, by comparing the gray spectral lines (obtained by performing ITD on the first-order difference signal)... By comparing the spectrum of the bearings, it can be found that the method proposed in this invention can effectively reduce the interference of noise and irrelevant components, and can extract the prominent bearing fault characteristic frequencies, thus achieving accurate diagnosis of bearing fault types.

[0127] Example 3

[0128] To further demonstrate the effectiveness and superiority of the method proposed in this invention, typical data from Table 1 and data 2 are randomly selected as examples to analyze the effectiveness of the proposed method. Data 2 corresponds to the bearing experiencing a combined outer ring + inner ring + rolling element failure, with the rotational speed... =1875 r / min. The characteristic frequencies are as follows: =31.25Hz, =138.54Hz, =80.2Hz, =54.42Hz, =11.46Hz.

[0129] The vibration signal of the bearing outer ring + inner ring + rolling element composite fault in Example 3 was obtained and processed using the method proposed in this invention. Specifically, in S3, the number of ITD decomposition layers P is 3; in S4, the moving window length q = 8; and in S5, the signal reconstruction method involves reconstructing the covariance matrix... Add the first row and the last column together.

[0130] Figure 4 The time-domain plot of the acceleration signal is shown in (a); (b) is the spectrum of the acceleration signal; (c) is the first-order differential signal of (a); (d) is the spectrum of (c); (e) is the PRC obtained by performing ITD on the first-order differential signal (in this embodiment, the number of decomposition layers is equal to 3); and (g) is the PRC component obtained by performing 1D-LBP enhancement. The component (h) is the spectrum of (g), and the green spectral lines in (i)-(k) are the components corresponding to the method of the present invention. - The spectrum of is represented by the gray spectral line in (f).

[0131] analyze Figure 4 (i)-(k), the signal spectrum (green spectral line) obtained by this invention can extract typical and prominent characteristic frequencies and their harmonics corresponding to bearing composite faults, as well as modulated frequency components, specifically:

[0132] (1) In Figure 4 In (i), 10 can be extracted 802 Hz (802 / 10=80.2 Hz); 7 , 970.46Hz (970.46 / 7=138.64 Hz); 7 + , 1003.42Hz ((1003.42-31.25) / 7=138.88 Hz); 26 , 1437.99 Hz, ((1437.99-2*11.46) / 26=54.43 Hz); 31 , 1686.4 Hz(1686.4 / 31=54.4 Hz); 13 - 1772.46 Hz, ((1772.46+31.25) / 13=138.75 Hz);

[0133] (2) In Figure 4 In (j), 10 can be extracted. 802 Hz (802 / 10=80.2 Hz); 7 , 970.46Hz (970.46 / 7=138.64 Hz); 7 + , 1003.42Hz ((1003.42-31.25) / 7=138.88 Hz);

[0134] (3) In Figure 4 In (k), 3 can be extracted , 284.24 Hz (284.24+3*31.25+3*11.46) / 3=137.46); 11 , 602 Hz (602 / 11=54.77 Hz).

[0135] From the analysis in (1)-(3) above, it can be found that the method proposed in this invention is not only applicable to the diagnosis of single bearing faults, but also to the identification of complex faults. Furthermore, the proposed method has excellent noise reduction and feature enhancement effects, not only increasing the frequency components corresponding to the fault, but more importantly, extracting the frequency components directly corresponding to the bearing fault type; these frequencies include: 10 harmonics of the outer ring fault characteristic frequency. 7th harmonic of the inner ring fault characteristic frequency 11th harmonic of the characteristic frequency of rolling element failure , 26 , 31 These frequencies also contain no modulation frequency components and directly correspond to the characteristic frequencies of bearing faults; this is crucial for the accurate identification of bearing faults in actual engineering projects. This further demonstrates the superiority of the proposed method in identifying complex bearing faults.

[0136] 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 bearing fault diagnosis method based on one-dimensional local binary mode and Hankel matrix, characterized in that, include: S1: Acquire discrete vibration signals ; S2: For the discrete vibration signal Perform first-order difference operation to obtain the difference signal. ; S3: For the differential signal The number of decomposition layers is: The inherent timescale decomposition yields An inherent rotation component signal ,in, It is a positive integer; S4: Utilize the improved one-dimensional local binary mode method for each intrinsic rotation component signal Quantization and signal reconstruction are performed to obtain the signal corresponding to the inherent rotation component. One-to-one correspondence decimal feature signal The improved one-dimensional local binary mode method uses the root mean square of the signal within the window as the quantization criterion. S5: Construct the decimal feature signal The Hankel matrix is ​​obtained and its covariance matrix is ​​calculated. Then, the signal is reconstructed based on the covariance matrix to obtain the reconstructed signal. S6: Perform spectral analysis on the reconstructed signal to obtain frequency components. Calculate the bearing fault characteristic frequency based on the bearing geometric parameters and rotational speed. Then, determine the bearing condition and fault type by comparing the obtained frequency components with the calculated bearing fault characteristic frequency.

2. The bearing fault diagnosis method based on one-dimensional local binary mode and Hankel matrix according to claim 1, characterized in that, In S4, for the first An inherent rotation component signal The steps for quantization and signal reconstruction are as follows, where, ; S41: Move the sliding window in steps of 1 at the... An inherent rotation component signal Move the signal up and down, and calculate the root mean square value of the signal within each sliding window. , among which, the The root mean square value of the signal within a sliding window Calculated using the following formula: ； In the formula, The length of the sliding window. , , for The length of the signal; S42: Calculate the difference between each element within each sliding window and the root mean square value of the signal within that sliding window. And based on the difference, obtain a binary encoding sequence that corresponds one-to-one with each element in each sliding window. , among which, the The first sliding window The elements and the root mean square value of the signal within the sliding window The difference Calculated using the following formula: ； In the formula, For the first The first sliding window One element, ; Among them, the The binary encoding sequence corresponding to each sliding window elements in Calculated using the following formula: ； or ; In the formula, Indicates the first The first sliding window The binary code corresponding to each element; S43: The binary encoding sequence corresponding to the signal within each sliding window. All are converted into decimal sequences to obtain the characteristic signal corresponding to the signal within each sliding window, where the first... The characteristic signal corresponding to the signal within each sliding window Calculate using the following formula; ； Corresponding decimal characteristic signal ... , The length of the decimal characteristic signal. .

3. The bearing fault diagnosis method based on one-dimensional local binary mode and Hankel matrix according to claim 1, characterized in that, In S5, using the first Decimal characteristic signal The method for signal reconstruction is as follows: S51: Construct the decimal feature signal Hankel matrix : ； In the formula: ,in, ; , The Hankel matrix is ​​respectively The number of rows and columns; S52: Calculate the Hankel matrix Find the mean of each column and construct a vector using the mean of each column. The Hankel matrix Mean of column j The calculation formula is as follows: ； The vector The formula is as follows: ； S53: For the Hankel matrix Each element is decentralized, as shown in the following formula: ； In the formula, It is the Hankel matrix element in row i and column j Decentralized data; S54: Constructing the covariance matrix : ； Wherein, the covariance matrix The Middle Line 1 Column data The calculation formula is as follows: ； In the formula, The Hankel matrix The The decentralized data in row I and column I. The Hankel matrix The The decentralized data in row J, where... ; S55: Using the covariance matrix Perform signal reconstruction to obtain the reconstructed signal. .

4. The bearing fault diagnosis method based on one-dimensional local binary mode and Hankel matrix according to claim 3, characterized in that, Using the covariance matrix Perform signal reconstruction to obtain the reconstructed signal. The method is as follows: The covariance matrix... Add the first row and the last column, or add the covariance matrix. The first and last rows are added together, or the diagonal average method is used.

Images