Rolling bearing fault diagnosis method and device based on optimal spectrum weight index

Abstract

The invention discloses a rolling bearing fault diagnosis method and equipment based on an optimal spectrum weight index, belongs to the technical field of bearing fault diagnosis, and is used for solving the problems that fault signals and normal signals are difficult to distinguish due to the fact that various signals collected by an existing rotating machinery fault diagnosis method are mixed and a large number of interference signals exist. And frequent occurrence of a misdiagnosis problem is caused. The method comprises the following steps: carrying out feature extraction on an acquired operation vibration signal of a rolling bearing of a rotating machine, and inputting the extracted signal features into an abnormal working condition discriminator for abnormal diagnosis; when detecting that the signal is abnormal, activating a system diagnosis module; performing symplectic geometric mode decomposition on the abnormal signals to obtain mode components forming the abnormal signals; calculating an optimal spectrum weight index of each modal component of the abnormal signal based on the health signal reference component; and according to the optimal spectrum weight index, screening fault characteristic components in the modal components, performing envelope spectrum fault diagnosis, and identifying a fault type.

Description

Technical Field

[0001] The invention relates to the technical field of bearing fault diagnosis, and in particular to a rolling bearing fault diagnosis method and equipment based on an optimal spectrum weight index. Background Art

[0002] During the operation of rotating machinery, the signals generated by its components have specific characteristic patterns. However, when a component fails, the signal characteristics generated by the component often change significantly and no longer maintain the original characteristic pattern. Especially in the early stages of a fault, the fault signal is usually weak and difficult to effectively extract from the complex mechanical vibration signal. In addition, these weak fault signals are also easily interfered by background noise and signals generated by other equipment excitation sources, further increasing the difficulty of fault signal extraction.

[0003] Traditionally, when performing fault diagnosis on rotating machinery, the collected signals are usually simply extracted or analyzed. However, this method has obvious limitations. Because the actual collected signals not only contain the signals generated by the faulty components, but also the signals of other normal components, the crosstalk signals between components, and the working information during the operation of the machinery. The presence of these interference signals makes it difficult for traditional feature extraction methods to accurately distinguish fault signals from normal signals, resulting in frequent misdiagnosis problems. Summary of the invention

[0004] The embodiment of the present invention provides a rolling bearing fault diagnosis method and device based on the optimal spectral weight index, which is used to solve the following technical problems: the signals collected by the existing rotating machinery fault diagnosis method are mixed with various signals and contain a large number of interference signals, making it difficult to distinguish between fault signals and normal signals, resulting in frequent occurrence of misdiagnosis problems.

[0005] The embodiment of the present invention adopts the following technical solutions: On the one hand, an embodiment of the present invention provides a rolling bearing fault diagnosis method based on an optimal spectral weight index, the method comprising: extracting features from a collected running vibration signal of a rotating machinery rolling bearing, and inputting the extracted signal features into an abnormal condition discriminator for abnormal diagnosis; When the abnormal operating condition discriminator detects an abnormal signal, the system diagnosis module is activated and a fault warning is issued for the abnormal signal; Performing symplectic geometric mode decomposition on the abnormal signal to obtain modal components constituting the abnormal signal; Calculating the optimal spectral weight index of each modal component of the abnormal signal based on the healthy signal reference component; According to the optimal spectrum weight index, the fault characteristic components in the modal components are screened, and envelope spectrum fault diagnosis is performed on the fault characteristic components to identify the corresponding fault type.

[0006] In a feasible implementation, feature extraction is performed on the collected running vibration signal of the rolling bearing of the rotating machinery, and the extracted signal feature is input into the abnormal condition discriminator for abnormal diagnosis, which specifically includes: Collecting the running vibration signal of the rolling bearing of the rotating machinery by means of a vibration sensor; Integrate the running vibration signal once to obtain a velocity signal and an acceleration signal; Extracting time domain statistical features and frequency domain statistical features of the velocity signal and the acceleration signal; wherein the time domain statistical features include at least one or more of the following: mean, root mean square value, waveform factor, pulse factor, skewness, variance, peak value, peak factor, margin factor and kurtosis; the frequency domain statistical features include at least one or more of the following: spectrum mean, spectrum skewness, average frequency, frequency variance, frequency skewness, spectrum root square value, spectrum kurtosis, frequency center of gravity, frequency root mean square and frequency kurtosis; The characteristic values ​​of the time domain statistical features and the frequency domain statistical features are input into the abnormal operating condition discriminator, so that the abnormal operating condition discriminator uses an unsupervised abnormal diagnosis algorithm to perform abnormal diagnosis on the operating vibration signal and discriminate abnormal operating conditions.

[0007] In a feasible implementation manner, performing symplectic geometric mode decomposition on the abnormal signal to obtain the modal components constituting the abnormal signal specifically includes: Based on the abnormal signal, construct a corresponding trajectory matrix and a symplectic geometry matrix; Obtaining an eigenvector matrix of the trajectory matrix in the symplectic geometry matrix, and determining a plurality of groups of initial single-component signals based on the eigenvector matrix; The several groups of initial single-component signals are iteratively decomposed into several modal components constituting the abnormal signals.

[0008] In a feasible implementation, based on the abnormal signal, a corresponding trajectory matrix and a symplectic geometry matrix are constructed, specifically including: Calculating the power spectrum density of the abnormal signal, and determining the target frequency value corresponding to the maximum peak of the power spectrum density; Determining an embedding dimension based on a ratio of the target frequency value to a sampling frequency of the abnormal signal; Based on the embedding dimension, constructing a trajectory matrix of the abnormal signal by using a Takkens embedding method; Performing autocorrelation analysis on the trajectory matrix to obtain a covariance matrix, and constructing a Hamiltonian matrix using the covariance matrix; According to the Hamiltonian matrix, a symplectic geometric matrix is ​​constructed.

[0009] In a feasible implementation manner, obtaining an eigenvector matrix of the trajectory matrix in the symplectic geometric matrix, and determining a plurality of groups of initial single-component signals based on the eigenvector matrix specifically includes: Extracting the eigenvector matrix of the trajectory matrix in the symplectic geometric matrix; Multiplying the feature vector matrix by the trajectory matrix to obtain a coefficient matrix; Multiplying the eigenvector matrix by the coefficient matrix to construct a single-component reconstruction matrix; The single-component reconstruction matrix is ​​transformed into a one-dimensional single-component time series by a diagonal average transformation method to obtain several groups of initial single-component signals.

[0010] In a feasible implementation manner, the several groups of initial single-component signals are iteratively decomposed into several modal components constituting the abnormal signal, specifically including: Among the plurality of groups of initial single-component signals, component signals having a period similarity with the first group of initial single-component signals higher than a first preset threshold are screened out and recombined to obtain a first component signal; The initial single component signal involved in the reorganization is removed to obtain a residual signal, and the residual signal is constructed as a residual term matrix; Calculating a normalized mean square error between the residual signal and the original abnormal signal; If the value of the normalized mean square error is not lower than the second preset threshold, then continue to screen the component signals whose period similarity with the current first group of initial single component signals is higher than the first preset threshold in the residual term matrix and reorganize them to obtain the second component signal, and iterate this process until the value of the normalized mean square error is less than the second preset threshold, and the decomposition process is terminated; The finally obtained plurality of component signals are determined as a plurality of modal components of the abnormal signal.

[0011] In a feasible implementation manner, before calculating the optimal spectral weight index of each modal component of the abnormal signal based on the healthy signal reference component, the method further includes: When the rolling bearing of the rotating machinery is in a healthy state, a healthy vibration signal is collected based on a preset sampling period and a preset sampling duration; The collected healthy vibration signal is subjected to time domain synchronous average denoising, and the denoised signal is stored as a healthy signal reference component.

[0012] In a feasible implementation manner, based on the healthy signal reference component, calculating the optimal spectral weight index of each modal component of the abnormal signal specifically includes: Calculate the normalized Fourier spectrum set of the reference component of the health signal respectively And the normalized Fourier spectrum of the current modal component ; Define the calculation formula of weighted normalized Fourier spectrum: ;in, is the weight vector, b is the bias, and NFS is the normalized Fourier spectrum set; Will and Representing two point clusters in the weighted normalized Fourier spectrum, and optimizing the distance between the two point clusters by the weighted normalized Fourier spectrum; Based on the maximum log-likelihood estimation, the objective function to achieve the maximum distance between two point clusters is defined; According to the objective function, the optimal spectral weight index of each modal component is calculated. In a feasible implementation manner, according to the optimal spectrum weight index, the fault characteristic component in the modal component is screened, and the envelope spectrum fault diagnosis is performed on the fault characteristic component to identify the corresponding fault type, specifically including: Determining the modal component whose optimal spectrum weight index exceeds a third preset threshold as the fault characteristic component; The envelope spectrum of the fault characteristic component is calculated, and based on the frequency characteristics of the envelope spectrum, the fault type of the current rolling bearing of the rotating machinery is identified.

[0013] On the other hand, an embodiment of the present invention further provides a rolling bearing fault diagnosis device based on an optimal spectral weight index, the device comprising: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, so that the at least one processor can execute the rolling bearing fault diagnosis method based on the optimal spectral weight index.

[0014] Compared with the prior art, the rolling bearing fault diagnosis method and device based on the optimal spectral weight index provided by the embodiment of the present invention have the following beneficial effects: When a rotating machine fails, the signal generated by the faulty component often no longer maintains its original symmetry, which makes the vibration signal nonlinear and non-stationary, and is easily affected by background noise and other equipment excitation sources. In the early stage of the fault, the fault signal is weak and difficult to extract. Traditional fault diagnosis methods are difficult to diagnose fault information. In response to this problem, the present invention first uses an abnormal working condition discriminator to preliminarily identify the fault signal, and then uses the symplectic geometric mode decomposition method to perform modal decomposition on the fault signal to obtain a series of symplectic geometric mode components, and calculates the optimal weight index of each component signal. Finally, the sensitive component reflecting the fault characteristics is selected by maximizing the optimal spectral weight index, and the signal is reconstructed with this index. Then, the sensitive component is envelope analyzed to extract the early fault characteristics of the rolling bearing; the optimal spectral weight index can effectively distinguish the interference components in the fault signal using the normal signal, and the symplectic geometric mode decomposition algorithm has the characteristics of unchanged phase space structure, which can effectively ensure the integrity of the internal characteristics of the data.

[0015] It can be seen that the present invention proposes a new method for fault diagnosis of rolling bearings in rotating machinery. This method, by combining advanced signal processing technology, feature extraction algorithm and intelligent diagnosis model, can accurately identify and extract the weak signal characteristics generated by faulty components in a complex signal environment, and effectively eliminate the influence of other interfering signals, thereby significantly improving the accuracy and reliability of fault diagnosis and reducing the risk of misdiagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 A flow chart of a rolling bearing fault diagnosis method based on an optimal spectral weight index provided by an embodiment of the present invention; Figure 2 A schematic structural diagram of a rolling bearing fault diagnosis device based on an optimal spectral weight index provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0018] The embodiment of the present invention provides a rolling bearing fault diagnosis method based on an optimal spectral weight index, such as Figure 1 As shown, the rolling bearing fault diagnosis method based on the optimal spectrum weight index specifically includes steps S101-S105: S101, extracting features from the collected operating vibration signals of the rolling bearings of the rotating machinery, and inputting the extracted signal features into an abnormal operating condition discriminator for abnormal diagnosis.

[0019] Specifically, the present invention mainly uses vibration signals as test quantities. The measurement signals of rotating mechanical equipment during operation usually include target element signals, crosstalk element signals and white noise, and the element signals are divided into working signals and fault signals when a fault occurs. Therefore, the spectrum analysis diagram of the vibration signal can be divided into several areas, namely, the working frequency band, the fault frequency band, the crosstalk frequency band and the noise frequency band. The present invention segments the sampled vibration signal into integer multiples of the characteristic signal period, and each signal segment contains a complete characteristic signal period. Assuming that the initial stage of equipment operation is a healthy state signal, the vibration signal at different times is sampled multiple times by a vibration sensor as a reference component, and this is used as a healthy state reference component.

[0020] As a feasible implementation method, when the rolling bearing of the rotating machinery is in a healthy state, a healthy vibration signal is collected based on a preset sampling period and a preset sampling duration. Then, the collected healthy vibration signal is subjected to time domain synchronous average denoising, and the denoised signal is stored as a healthy signal reference component.

[0021] In one embodiment, when the device is in a healthy state, the vibration signal is sampled once every 10 minutes, each sampling lasts 10 seconds, and the frequency is 12800 Hz. The 6 signals sampled in 1 hour are synchronously averaged in the time domain to improve the signal-to-noise ratio of the signal, and the signal after synchronous average denoising is used as the reference component. .

[0022] Furthermore, during the operation of the equipment, the operation vibration signal of the rolling bearing of the rotating machinery is collected by a vibration sensor; then the collected operation vibration signal is integrated once to obtain a velocity signal and an acceleration signal.

[0023] Furthermore, the time domain statistical features and frequency domain statistical features of the velocity signal and the acceleration signal are extracted; wherein the time domain statistical features include at least one or more of the following: mean, root mean square value, waveform factor, pulse factor, skewness, variance, peak value, peak factor, margin factor and kurtosis; the frequency domain statistical features include at least one or more of the following: spectral mean, spectral skewness, average frequency, frequency variance, frequency skewness, spectral root square value, spectral kurtosis, frequency center of gravity, frequency root mean square and frequency kurtosis.

[0024] Furthermore, the characteristic values ​​of the time domain statistical features and the frequency domain statistical features are input into the abnormal working condition discriminator, so that the abnormal working condition discriminator uses an unsupervised abnormal diagnosis algorithm to perform abnormal diagnosis on the running vibration signal and discriminate the abnormal working condition. The unsupervised abnormal diagnosis algorithm used by the abnormal working condition discriminator is at least one of an isolation forest algorithm, a One-Class SVM algorithm, and an autoencoder algorithm.

[0025] S102: When the abnormal operating condition discriminator detects that the signal is abnormal, the system diagnosis module is activated and a fault warning is issued for the abnormal signal.

[0026] Specifically, when the abnormal operating condition discriminator detects that a certain vibration signal has an abnormal condition, it sends an activation signal to the system diagnosis module, intercepts the abnormal signal segment, sends it to the system diagnosis module, and issues a fault warning to the equipment management terminal.

[0027] S103, performing symplectic geometric mode decomposition on the abnormal signal to obtain mode components constituting the abnormal signal.

[0028] Specifically, based on the abnormal signal, the corresponding trajectory matrix and symplectic geometry matrix are constructed. Then, the eigenvector matrix of the trajectory matrix is ​​obtained in the symplectic geometry matrix, and based on the eigenvector matrix, several groups of initial single-component signals are determined. Finally, several groups of initial single-component signals are iteratively decomposed into several modal components constituting the abnormal signal.

[0029] Since there are multiple components in mechanical equipment, and different parts are affected by the working state differently, they are easily affected by background noise or interference frequency during fault diagnosis, and it is difficult to unify the diagnostic model. In view of this problem, in order to minimize the influence of background noise and interference frequency, the present invention invents a symplectic geometric mode decomposition method based on the optimal spectral weight index to extract the fault component in the abnormal signal. Symplectic Geometry Mode Decomposition (SGMD) is a mathematical method mainly used to decompose a given symplectic manifold into multiple modes. Symplectic manifold is a special manifold with a symplectic structure, that is, its geometric characteristics can be described by symplectic form. The purpose of symplectic geometric mode decomposition is to decompose the geometric characteristics of the symplectic manifold into a combination of different modes in order to better understand and process complex geometric structures. The core idea of ​​symplectic geometric mode decomposition is to find a suitable symplectic transformation in the symplectic space of the signal so that the signal presents the simplest structure under the transformation. By iteratively applying the symplectic transform and the screening process, SGMD is able to decompose the signal into a series of eigenmodal functions (Symplectic Geometric Modal Components, SGCs), each of which represents a single modal component of the signal.

[0030] As a feasible implementation method, the specific implementation process of modal decomposition to obtain modal components is as follows: First, calculate the power spectral density PSD of the abnormal signal and determine the target frequency value f corresponding to the maximum peak of the power spectral density. max ; Based on the target frequency value f max With the abnormal signal x = [x1, x2,...,x n The sampling frequency f s When the wallpaper is smaller than the set threshold, the embedding dimension is set to d=n / 3; otherwise, d=1.2*(f s / f max ).

[0031] Based on the embedding dimension d, the trajectory matrix of the abnormal signal is constructed by the Takkens embedding method ;in, represents the delay time, m=n-(d-1), and n represents the sequence length of the abnormal signal.

[0032] Furthermore, the trajectory matrix X is subjected to autocorrelation analysis to obtain the covariance matrix A=X T X , and use the covariance matrix to construct the Hamiltonian matrix ; Furthermore, according to the Hamiltonian matrix, the symplectic geometry matrix is ​​constructed , where Q is a symplectic orthogonal matrix, B is an upper triangular matrix, and G=M 2 , R is the sub-matrix of matrix transformation.

[0033] The eigenvector matrix Q of the trajectory matrix is ​​extracted from the symplectic geometry matrix, and then the eigenvector matrix is ​​multiplied by the trajectory matrix to obtain a coefficient matrix; the eigenvector matrix is ​​then multiplied by the coefficient matrix to construct a single-component reconstruction matrix.

[0034] Furthermore, the single-component reconstruction matrix is ​​transformed by diagonal average transformation. Z i Transformed into a one-dimensional single-component time series Y i , and obtain several groups of initial single-component signals [ Y 1, Y 2,……, Y d ].

[0035] Furthermore, among several groups of initial single-component signals [ Y 1, Y 2,……, Y d ], calculate each initial single component signal and the first group of initial single component signals respectively Y 1, and screen out the component signals whose period similarity with the first group of initial single component signals is higher than a first preset threshold and recombine them to obtain the first component signal.

[0036] Then remove the initial single component signal involved in the recombination to obtain the residual signal, and construct the residual signal into a residual term matrix. Calculate the normalized mean square error between the residual signal and the original abnormal signal. If the value of the normalized mean square error is not lower than the second preset threshold, continue to screen the component signals whose periodic similarity with the current first group of initial single component signals is higher than the first preset threshold in the residual term matrix and reorganize to obtain the second component signal. According to the above process, iterate and loop until the value of the normalized mean square error is less than the second preset threshold, and the decomposition process terminates. The several component signals finally obtained are determined as several modal components of the abnormal signal.

[0037] S104. Calculate the optimal spectral weight index of each modal component of the abnormal signal based on the healthy signal reference component.

[0038] Specifically, the normalized Fourier spectrum set of the reference component of the healthy signal is calculated by the calculation formula of the normalized Fourier spectrum. And the normalized Fourier spectrum of the current modal component .

[0039] The main difference between fault signals and healthy signals is that there are some narrowband fault components in fault data, while most healthy signals contain interference components and noise related to mechanical equipment. These interference components also exist in the fault signals tested by the same equipment. Therefore, using the useful information in these healthy signals can reduce the noise of the fault signals and screen out the sensitive components that reflect the fault characteristics.

[0040] The present invention defines the calculation formula of the weighted normalized Fourier spectrum as: ;in, is the weight vector, b is the bias, and NFS is the normalized Fourier spectrum set.

[0041] Will and Represented as two point clusters in the weighted normalized Fourier spectrum, the weighted normalized Fourier spectrum can be defined as a hyperplane in high-dimensional space. The main purpose of the hyperplane is to optimize the distance between the two point clusters. By optimizing the distance between the two point clusters through the weighted normalized Fourier spectrum, based on the maximum log-likelihood estimation, the objective function to achieve the maximum distance between the two point clusters is defined: ;in, ; , NFS i is the normalized Fourier spectrum set of the i-th sample set, λ is the regularization term, P , Q are the number of healthy sample sets and fault sample sets respectively, y i is the i-th sample set.

[0042] The objective function is solved by the gradient descent method to obtain an optimal weight vector, and then the optimal weight vector is substituted into the optimal spectral weight index calculation formula to calculate the optimal spectral weight index of each modal component.

[0043] S105 , screening fault characteristic components in the modal components according to the optimal spectrum weight index, and performing envelope spectrum fault diagnosis on the fault characteristic components to identify corresponding fault types.

[0044] Specifically, the modal component whose optimal spectrum weight index exceeds the third preset threshold is determined as the fault characteristic component. Then, the envelope spectrum of the fault characteristic component is calculated, and the fault type of the current rotating machinery rolling bearing is identified based on the frequency characteristics of the envelope spectrum.

[0045] It should be noted that the various preset thresholds mentioned in the present invention are values ​​set through experiments or based on experience in actual processes and can be set flexibly, so no numerical limitations are made in the specification.

[0046] In addition, the embodiment of the present invention also provides a rolling bearing fault diagnosis device based on the optimal spectrum weight index, such as Figure 2 As shown, the rolling bearing fault diagnosis device based on the optimal spectrum weight index specifically includes: at least one processor; and a memory in communication with the at least one processor; wherein, The memory stores instructions executable by at least one processor to enable the at least one processor to perform: Extracting features from the collected running vibration signals of the rolling bearings of the rotating machinery, and inputting the extracted signal features into an abnormal condition discriminator for abnormal diagnosis; When the abnormal operating condition discriminator detects an abnormal signal, the system diagnosis module is activated and a fault warning is issued for the abnormal signal; Performing symplectic geometric mode decomposition on the abnormal signal to obtain modal components constituting the abnormal signal; Calculating the optimal spectral weight index of each modal component of the abnormal signal based on the healthy signal reference component; According to the optimal spectrum weight index, the fault characteristic components in the modal components are screened, and envelope spectrum fault diagnosis is performed on the fault characteristic components to identify the corresponding fault type.

[0047] Each embodiment of the present invention is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0048] The above describes specific embodiments of the present invention. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0049] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present invention should be included in the protection scope of the present invention.

Claims

1. A rolling bearing fault diagnosis method based on optimal spectral weight index, characterized in that: The method comprises: Extracting features from the collected running vibration signals of the rolling bearings of the rotating machinery, and inputting the extracted signal features into an abnormal condition discriminator for abnormal diagnosis; When the abnormal operating condition discriminator detects an abnormal signal, the system diagnosis module is activated and a fault warning is issued for the abnormal signal; Performing symplectic geometric mode decomposition on the abnormal signal to obtain modal components constituting the abnormal signal; Calculating the optimal spectral weight index of each modal component of the abnormal signal based on the healthy signal reference component; According to the optimal spectrum weight index, the fault characteristic components in the modal components are screened, and envelope spectrum fault diagnosis is performed on the fault characteristic components to identify the corresponding fault type.

2. A rolling bearing fault diagnosis method based on optimal spectral weight index according to claim 1, characterized in that: The collected vibration signals of the rolling bearings of the rotating machinery are subjected to feature extraction, and the extracted signal features are input into the abnormal condition discriminator for abnormal diagnosis, which specifically includes: Collecting the running vibration signal of the rolling bearing of the rotating machinery by means of a vibration sensor; Integrate the running vibration signal once to obtain a velocity signal and an acceleration signal; Extracting time domain statistical features and frequency domain statistical features of the velocity signal and the acceleration signal; wherein the time domain statistical features include at least one or more of the following: mean, root mean square value, waveform factor, pulse factor, skewness, variance, peak value, peak factor, margin factor and kurtosis; the frequency domain statistical features include at least one or more of the following: spectrum mean, spectrum skewness, average frequency, frequency variance, frequency skewness, spectrum root square value, spectrum kurtosis, frequency center of gravity, frequency root mean square and frequency kurtosis; The characteristic values ​​of the time domain statistical features and the frequency domain statistical features are input into the abnormal operating condition discriminator, so that the abnormal operating condition discriminator uses an unsupervised abnormal diagnosis algorithm to perform abnormal diagnosis on the operating vibration signal and discriminate abnormal operating conditions.

3. A rolling bearing fault diagnosis method based on optimal spectral weight index according to claim 1, characterized in that: Performing symplectic geometric mode decomposition on the abnormal signal to obtain the modal components constituting the abnormal signal specifically includes: Based on the abnormal signal, construct a corresponding trajectory matrix and a symplectic geometry matrix; Obtaining an eigenvector matrix of the trajectory matrix in the symplectic geometry matrix, and determining a plurality of groups of initial single-component signals based on the eigenvector matrix; The several groups of initial single-component signals are iteratively decomposed into several modal components constituting the abnormal signals.

4. A rolling bearing fault diagnosis method based on optimal spectral weight index according to claim 3, characterized in that: Based on the abnormal signal, a corresponding trajectory matrix and a symplectic geometry matrix are constructed, specifically including: Calculating the power spectrum density of the abnormal signal, and determining the target frequency value corresponding to the maximum peak of the power spectrum density; Determining an embedding dimension based on a ratio of the target frequency value to a sampling frequency of the abnormal signal; Based on the embedding dimension, constructing a trajectory matrix of the abnormal signal by using a Takkens embedding method; Performing autocorrelation analysis on the trajectory matrix to obtain a covariance matrix, and constructing a Hamiltonian matrix using the covariance matrix; According to the Hamiltonian matrix, a symplectic geometric matrix is ​​constructed.

5. A rolling bearing fault diagnosis method based on optimal spectral weight index according to claim 3, characterized in that: Obtaining an eigenvector matrix of the trajectory matrix in the symplectic geometric matrix, and determining a plurality of groups of initial single-component signals based on the eigenvector matrix, specifically includes: Extracting the eigenvector matrix of the trajectory matrix in the symplectic geometric matrix; Multiplying the feature vector matrix by the trajectory matrix to obtain a coefficient matrix; Multiplying the eigenvector matrix by the coefficient matrix to construct a single-component reconstruction matrix; The single-component reconstruction matrix is ​​transformed into a one-dimensional single-component time series by a diagonal average transformation method to obtain a plurality of groups of initial single-component signals.

6. A rolling bearing fault diagnosis method based on optimal spectral weight index according to claim 3, characterized in that: Iteratively decomposing the several groups of initial single-component signals into several modal components constituting the abnormal signal, specifically including: Among the plurality of groups of initial single-component signals, component signals having a period similarity with the first group of initial single-component signals higher than a first preset threshold are screened out and recombined to obtain a first component signal; The initial single component signal involved in the reorganization is removed to obtain a residual signal, and the residual signal is constructed as a residual term matrix; Calculating a normalized mean square error between the residual signal and the original abnormal signal; If the value of the normalized mean square error is not lower than the second preset threshold, then continue to screen the component signals whose period similarity with the current first group of initial single component signals is higher than the first preset threshold in the residual term matrix and reorganize them to obtain the second component signal, and iterate this process until the value of the normalized mean square error is less than the second preset threshold, and the decomposition process is terminated; The finally obtained plurality of component signals are determined as a plurality of modal components of the abnormal signal.

7. A rolling bearing fault diagnosis method based on optimal spectral weight index according to claim 1, characterized in that: Before calculating the optimal spectral weight index of each modal component of the abnormal signal based on the healthy signal reference component, the method further includes: When the rolling bearing of the rotating machinery is in a healthy state, a healthy vibration signal is collected based on a preset sampling period and a preset sampling duration; The collected healthy vibration signal is subjected to time domain synchronous average denoising, and the denoised signal is stored as a healthy signal reference component.

8. A rolling bearing fault diagnosis method based on optimal spectral weight index according to claim 1, characterized in that: Based on the healthy signal reference component, the optimal spectral weight index of each modal component of the abnormal signal is calculated, specifically including: Calculate the normalized Fourier spectrum set of the reference component of the health signal respectively And the normalized Fourier spectrum of the current modal component ; Define the calculation formula of weighted normalized Fourier spectrum: ;in, is the weight vector, b is the bias, and NFS is the normalized Fourier spectrum set; Will and Representing two point clusters in the weighted normalized Fourier spectrum, and optimizing the distance between the two point clusters by the weighted normalized Fourier spectrum; Based on the maximum log-likelihood estimation, the objective function to achieve the maximum distance between two point clusters is defined; According to the objective function, the optimal spectral weight index of each modal component is calculated.

9. A rolling bearing fault diagnosis method based on optimal spectral weight index according to claim 1, characterized in that: According to the optimal spectrum weight index, the fault characteristic components in the modal components are screened, and envelope spectrum fault diagnosis is performed on the fault characteristic components to identify the corresponding fault type, specifically including: Determining the modal component whose optimal spectrum weight index exceeds a third preset threshold as the fault characteristic component; The envelope spectrum of the fault characteristic component is calculated, and based on the frequency characteristics of the envelope spectrum, the fault type of the current rolling bearing of the rotating machinery is identified.

10. A rolling bearing fault diagnosis device based on optimal spectral weight index, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, so that the at least one processor can execute the rolling bearing fault diagnosis method based on the optimal spectral weight index according to any one of claims 1-9.

Images