A rotating machinery composite fault diagnosis method and system

Abstract

The present invention provides a method and system for diagnosing composite faults in rotating machinery, relating to the technical fields of mechanical signal processing and equipment fault diagnosis. This method collects fault signals from the inner and outer races of rolling bearings, uses a sliding window filter to locate the resonant frequency band, and calculates modal optimization indicators to select modes containing fault characteristics. The method then estimates the fault characteristic frequencies using the envelope harmonic product spectrum and removes non-convergent sub-signals. Composite fault signals are separated using an improved sparse Bayesian learning (SBL) classification model combined with variational Bayesian inference (VBI). Finally, envelope demodulation is used to determine the fault type. This method addresses the difficulty of separating composite fault characteristics under strong noise interference and the reliance of traditional methods on prior knowledge, achieving efficient and accurate composite fault diagnosis for rotating machinery with strong noise immunity.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical signal processing and equipment fault diagnosis, and in particular to a rotating machinery composite fault diagnosis method and system. Background Art

[0002] In modern industrial systems, rotating machinery, as a critical piece of equipment, is widely used in numerous important fields, including energy and power, rail transportation, and aerospace. Rolling bearings, as the core supporting components of rotating machinery, have a direct impact on the reliability and stability of the entire equipment. Failure of a rolling bearing can cause equipment downtime, production interruptions, and even serious safety incidents, resulting in significant economic losses.

[0003] In real-world operating environments, rotating machinery operates in extremely complex environments, and its composite fault vibration signals exhibit nonstationary and nonlinear characteristics. The cross-coupling of signals generated by different faults poses significant challenges in separating and diagnosing fault signatures. Furthermore, the prevalence of strong noise further obscures fault signals, making it difficult to effectively extract and identify fault signatures, significantly increasing the difficulty of composite fault diagnosis.

[0004] Existing decomposition methods have significant drawbacks when diagnosing complex faults in rotating machinery. For one thing, these methods rely heavily on accurate theoretical fault signatures and prior knowledge of the decomposition modulus. In practical applications, accurately acquiring this prior information is challenging due to the diversity and uncertainty of equipment operating conditions. Any deviation from this prior information significantly reduces the accuracy of diagnostic results. Furthermore, traditional signal processing methods typically employ frequency band decomposition techniques. However, when signals of similar frequencies exist within the same frequency band, they overlap, making effective separation difficult. Furthermore, as noise intensity increases, traditional methods lose their ability to resist interference, making it impossible to guarantee the accuracy and reliability of diagnostic results.

[0005] In summary, developing a composite fault diagnosis method and system for rotating machinery that can operate efficiently under strong noise interference is of great practical significance. This will not only improve the safety and reliability of rotating machinery operation, but also provide a strong guarantee for stable production in related industrial fields. Summary of the Invention

[0006] To this end, the present invention provides a method and system for diagnosing composite faults of rotating machinery, which is used to solve the problem in the existing technology that composite fault diagnosis of rotating machinery relies too much on accurate theoretical fault characteristics and decomposition modulus priors, has poor robustness under strong noise interference, and traditional frequency band decomposition methods are difficult to separate signals due to the overlap of similar frequency signals in the same frequency band, and the accuracy cannot be guaranteed when the noise increases, resulting in the difficulty in effectively separating and identifying the composite fault characteristics of rolling bearings.

[0007] In order to solve the above problems, an embodiment of the present invention provides a rotating machinery compound fault diagnosis method, which includes:

[0008] S1: Collect the fault signals of the inner and outer rings of the rolling bearing, locate the resonant frequency band of the fault signal based on the sliding window filter, separate the sub-signal set, calculate the modal optimization index value of each sub-signal one by one, select the first four sub-signals with the largest modal optimization index value, and eliminate redundant components by calculating the correlation coefficient between the signals, retaining the mode containing the fault characteristics;

[0009] S2: Estimate the fault characteristic frequency of the resonant frequency band based on envelope demodulation, use the envelope harmonic product spectrum to estimate the frequency of the remaining sub-signals, remove the non-convergent sub-signals, and use the envelope harmonic product spectrum value of the remaining signal as the fault characteristic frequency;

[0010] S3: Classify the optimal sub-signals based on the improved sparse Bayesian learning classification model, and solve the maximum posterior probability through variational Bayesian inference to separate the composite fault signals;

[0011] S4: Perform envelope demodulation on the separated signal and determine the specific fault type based on whether the bearing fault characteristic frequency exists.

[0012] Preferably, the modal optimization index (MOI) value is composed of kurtosis (Kurt), envelope spectrum kurtosis (ESK) and mutual information (MI), which is calculated as follows:

[0013] ;

[0014] in

[0015] ;

[0016] ;

[0017] ;

[0018] Where, Optimize the index value for the mode; is the kurtosis; ESK is the envelope spectrum kurtosis; MI is the mutual information; is the time domain sequence of the signal; is the mean value of the time domain signal; is the signal length; is the time series index; Indicates the envelope spectrum at frequency The amplitude at is the mean value of the envelope spectrum; For signal The entropy value of For signal When the signal is known The conditional entropy of .

[0019] Preferably, the initialization of the sliding window filter includes:

[0020] Divide the original signal into segments, each segment is , and according to the sampling frequency Determine the upper and lower cutoff frequencies for each segment and , build A finite-length unit impulse response filter with upper and lower cutoff frequencies and satisfy:

[0021] ;

[0022] Where, represents the filter index, is the total number of filters.

[0023] Preferably, the updating method of the sliding window filter includes:

[0024] The correlation kurtosis is used as the objective function to update the filter bank, and the generalized eigenvalue problem is solved iteratively to approximate the filtered signal corresponding to the maximum correlation kurtosis. The update formula is:

[0025] ;

[0026] ;

[0027] Where, is the relevant kurtosis value, is the number of shifts; For the The signal components of the filter output; is the sampling period; Signal length; For the The coefficient vector of the filter; is the filter length; 、 、 、 Both represent indexes; Indicates constraints.

[0028] Preferably, the method for estimating the frequency of the remaining sub-signal using the envelope harmonic product spectrum is:

[0029] ;

[0030] Where, is the envelope harmonic product spectrum value, 、 、 、 、 is the Fourier transform amplitude of the signal envelope; is the number of harmonics; Represents an index; is the angular frequency.

[0031] Preferably, the improved sparse Bayesian learning classification model adopts a block sparse structure, and its mathematical expression is:

[0032] ;

[0033] Joint probability density The formula is:

[0034] ;

[0035] Where, is the conditional probability of block sparse structure; is a block structure matrix; is the fault characteristic component, For control Hyperparameters of sparsity, subscript Indicates the position within the block; is the block structure weight coefficient, defined The degree of dependence on adjacent sparsity parameters; The mean is 0 and the variance is Gaussian distribution; It is a fault signal; To observe the signal; is a set of latent variables; is the likelihood function; Noise accuracy The prior distribution of is the sparse parameter The prior distribution of is the conditional probability of block sparse structure; is the prior distribution of the block structure matrix; is the signal length; 、 Both represent indexes.

[0036] Preferably, the method of classifying the optimal sub-signal based on the improved sparse Bayesian learning classification model and solving the maximum posterior probability by variational Bayesian inference update specifically includes:

[0037] A relaxed evidence lower bound formula is designed, and the latent variables are iteratively updated through variational Bayesian inference to solve the maximum posterior probability approximate solution. The relaxed evidence lower bound calculation formula is:

[0038] ;

[0039] in

[0040]

[0041] Where, is the relaxed evidence lower bound, is a variational distribution, which is used to approximate the true posterior distribution which is difficult to calculate ; is the generalized likelihood function; is the observation likelihood function; is the conditional probability; is a set of latent variables; is the variational parameter.

[0042] Preferably, the relaxed evidence lower bound is obtained by introducing the KL divergence to minimize the variational distribution and the true posterior distribution The specific optimization formula is:

[0043] ;

[0044] And by decomposing the latent variable update rules, the computational complexity is reduced.

[0045] Preferably, the method of performing envelope demodulation on the separated signal is:

[0046] First, the separated signal is Hilbert transformed to construct the analytical signal; then, the amplitude of the analytical signal is extracted; finally, the amplitude is Fourier transformed to obtain the envelope spectrum, and the characteristic frequency is observed through the envelope spectrum.

[0047] An embodiment of the present invention further provides a rotating machinery compound fault diagnosis system, which is used to implement the above-mentioned rotating machinery compound fault diagnosis method, specifically comprising:

[0048] The frequency band adaptive positioning module is used to collect the fault signals of the inner and outer rings of the rolling bearing, locate the resonant frequency band of the fault signal based on the sliding window filter, separate the sub-signal set, calculate the modal optimization index value of each sub-signal one by one, select the first four sub-signals with the largest modal optimization index value, and eliminate redundant components by calculating the correlation coefficient between the signals, retaining the mode containing the fault characteristics;

[0049] The characteristic frequency estimation module is used to estimate the fault characteristic frequency of the resonant frequency band based on envelope demodulation, use the envelope harmonic product spectrum to estimate the frequency of the remaining sub-signals, eliminate the non-converged sub-signals, and use the envelope harmonic product spectrum value of the remaining signal as the fault characteristic frequency;

[0050] The fault separation module is used to classify the optimal sub-signals based on the improved sparse Bayesian learning classification model, and to solve the maximum posterior probability through variational Bayesian inference to achieve separation of composite fault signals;

[0051] The envelope demodulation module is used to perform envelope demodulation on the separated signal and determine the specific fault type based on whether there is a bearing fault characteristic frequency.

[0052] It can be seen from the above technical solutions that the present invention has the following beneficial effects:

[0053] The embodiments of the present invention provide a method and system for diagnosing composite faults of rotating machinery, innovatively proposing adaptive modal decomposition and an improved sparse Bayesian method. Adaptive modal decomposition can achieve adaptive selection of decomposition patterns with the help of a filter bank, while estimating the fault characteristic frequency, which can solve the problem that existing methods are overly dependent on accurate theoretical fault characteristic frequencies and prior knowledge of decomposition moduli. The improved sparse Bayesian learning method has strong resistance to noise interference, and the proposed relaxed evidence lower bound can effectively reduce the computational complexity of the model, ultimately achieving more accurate and rapid composite fault diagnosis of rotating machinery. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the implementation cases of the present invention or the technical solutions in the prior art, the following is a brief description of the drawings required for use in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be understood as limiting the present invention in any way. Those skilled in the art can derive other drawings based on these drawings without inventive effort. Among them:

[0055] Figure 1 A flowchart of a rotating machinery composite fault diagnosis method provided by the present invention;

[0056] Figure 2 It is a waveform diagram of the fault signal of the inner and outer rings of the original rolling bearing in the present invention;

[0057] Figure 3 It is an envelope diagram of the rolling bearing separation fault signal in the present invention;

[0058] Figure 4 This is a block diagram of a rotating machinery composite fault diagnosis system provided by the present invention. DETAILED DESCRIPTION

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0060] Example 1:

[0061] In order to solve the problem that the diagnosis of rotating machinery composite faults in the existing technology is too dependent on accurate theoretical fault features and decomposition modulus priors, and has poor robustness under strong noise interference, the traditional frequency band decomposition method is difficult to separate signals due to the overlap of similar frequency signals in the same frequency band, and the accuracy cannot be guaranteed when the noise increases, resulting in the difficulty in effectively separating and identifying the composite fault features of rolling bearings. Figure 1 As shown, the present invention proposes a rotating machinery composite fault diagnosis method, which includes:

[0062] S1: Collect the fault signals of the inner and outer rings of the rolling bearing, locate the resonant frequency band of the fault signal based on the sliding window filter, separate the sub-signal set, calculate the modal optimization index value of each sub-signal one by one, select the first four sub-signals with the largest modal optimization index value, and eliminate redundant components by calculating the correlation coefficient between the signals, retaining the mode containing the fault characteristics;

[0063] S2: Estimate the fault characteristic frequency of the resonant frequency band based on envelope demodulation, use the envelope harmonic product spectrum to estimate the frequency of the remaining sub-signals, remove the non-convergent sub-signals, and use the envelope harmonic product spectrum value of the remaining signal as the fault characteristic frequency;

[0064] S3: Classify the optimal sub-signals based on the improved sparse Bayesian learning classification model, and solve the maximum posterior probability through variational Bayesian inference to separate the composite fault signals;

[0065] S4: Perform envelope demodulation on the separated signal and determine the specific fault type based on whether the bearing fault characteristic frequency exists.

[0066] From the above technical solution, it can be seen that the present invention proposes a method for composite fault diagnosis of rotating machinery, which uses a filter group to accurately locate the fault frequency band, and at the same time estimates the fault characteristic frequency through the envelope harmonic product spectrum, thereby realizing adaptive selection of the fault characteristic frequency and decomposition modulus; in the process of separating faults, a sparse Bayesian learning classification model is established, a relaxed evidence lower bound is established, and the maximum posterior probability is solved by variational Bayes to accurately separate faults and enhance the noise resistance of the method; the frequency clear characterization and noise resistance of different fault bearing signals are improved, and ultimately more accurate rolling bearing composite fault diagnosis is achieved.

[0067] In step S1, the present invention collects the fault signals of the inner and outer rings of the rolling bearing, such as Figure 2 As shown, based on the sliding window filter, the fault signal resonance frequency band is located and the sub-signal set is separated. , calculate the modal optimization index value of each sub-signal one by one, select the first four sub-signals with the largest modal optimization index value, and eliminate redundant components by calculating the correlation coefficient between signals, retaining the mode containing the fault characteristics.

[0068] Specifically, the modal optimization index (MOI) value of the present invention is composed of kurtosis (Kurt), envelope spectral kurtosis (ESK) and mutual information (MI).

[0069] The calculation formula of the signal kurtosis (Kurt) is:

[0070] ;

[0071] Where, is the time domain sequence of the signal; is the mean value of the time domain signal; is the signal length; Represents an index.

[0072] The calculation formula of envelope spectral kurtosis (ESK) is:

[0073] ;

[0074] Where, Indicates the envelope spectrum at frequency The amplitude at is the mean value of the envelope spectrum.

[0075] In order to remove the interference component which also has high kurtosis and envelope spectrum kurtosis but has low correlation with the original signal and does not contain fault information, the present invention introduces mutual information, and the calculation formula is:

[0076] ;

[0077] Where, For signal The entropy value of For signal When the signal is known The conditional entropy of .

[0078] Finally, the Modal Optimization Index (MOI) value is defined as follows:

[0079] .

[0080] Furthermore, in order to initialize the sliding window filter, the present invention divides the original signal into segments, each segment is , and according to the sampling frequency Determine the upper and lower cutoff frequencies for each segment and , build A finite-length unit impulse response (FIR) filter (for example, a Hanning window is selected as the filter with a length of ), where the upper and lower cutoff frequencies and satisfy:

[0081] ;

[0082] Where, Represents an index, is the total number of filters.

[0083] Furthermore, the present invention uses the correlation kurtosis as the objective function to update the filter bank and approximates the filtered signal corresponding to the maximum correlation kurtosis value by iteratively solving the generalized eigenvalue problem, wherein the update formula is:

[0084] ;

[0085] ;

[0086] Where, is the relevant kurtosis value, is the number of shifts; For the The signal components of the filter output; is the sampling period; is the signal length; For the The coefficient vector of the filter; is the filter length; 、 、 、 Both represent indexes; Indicates constraints.

[0087] The present invention proposes an iterative eigenvalue decomposition method to solve the above equation, as follows:

[0088] ;

[0089] in, , represents the matrix form of the signal component output by the kth filter; , represents the matrix form of the signal; , which represents the matrix form of the k-th FIR filter.

[0090] Correlation kurtosis Defined as:

[0091] ;

[0092] Where, is the conjugate transpose operation; is the weighted correlation matrix. Further deduction:

[0093] ;

[0094] Where, is the weighted correlation matrix; is the correlation matrix.

[0095] Furthermore, the present invention will The maximization problem is transformed into a generalized eigenvalue problem, and the maximum eigenvalue is calculated. The associated eigenvector is expressed as follows:

[0096] .

[0097] Through continuous iteration, the filter coefficients, thereby continuously approaching the filtered signal corresponding to the maximum value of the relevant kurtosis.

[0098] In step S2, the present invention estimates the fault characteristic frequency of the resonant frequency band based on envelope demodulation, uses the envelope harmonic product spectrum to estimate the frequency of the remaining sub-signals, eliminates the unconverged sub-signals, and uses the envelope harmonic product spectrum value of the remaining signal as the fault characteristic frequency.

[0099] Specifically, the present invention uses the Envelope Harmonic Product Spectrum (EHPS) to estimate the frequency of the remaining sub-signal, and its mathematical expression is:

[0100] ;

[0101] Where, is the envelope harmonic product spectrum value, 、 、 、 、 is the Fourier transform amplitude of the signal envelope; is the number of harmonics; Represents an index; is the angular frequency.

[0102] In step S3, the present invention classifies the optimal sub-signals based on the improved sparse Bayesian learning classification model, and solves the maximum posterior probability through variational Bayesian inference update to achieve separation of composite fault signals.

[0103] Specifically, according to the traditional sparse Bayesian model, the fault signal It is expressed as obeying the mean of 0 and the mean square error of , the process is as follows:

[0104] ;

[0105] in Influence The sparsity of Modeled as an independent and identically distributed Gaussian distribution, the process is as follows:

[0106] .

[0107] Since the main elements of the signal appear mainly in the form of blocks, that is, block sparse structure. Not only with Related to and Therefore, the improved sparse Bayesian learning (SBL) classification model designed by the present invention adopts a block sparse structure, and its mathematical expression is:

[0108] ;

[0109] Joint probability density The formula is:

[0110] ;

[0111] Where, is the conditional probability of block sparse structure; is a block structure matrix, , ; is the fault characteristic component, For control Hyperparameters of sparsity, subscript Indicates the position within the block; is the block structure weight coefficient, defined The degree of dependence on adjacent sparsity parameters; The mean is 0 and the variance is Gaussian distribution; It is a fault signal; To observe the signal; is a set of latent variables; is the likelihood function; Noise accuracy The prior distribution of is the sparse parameter The prior distribution of is the conditional probability of block sparse structure; is the prior distribution of the block structure matrix; is the signal length; 、 Both represent indexes.

[0112] Since in order to extract the fault pulse, the true posterior distribution must be estimated However, the traditional SBL algorithm has high computational complexity when updating latent variables. This invention reduces the computational complexity by designing a relaxed lower bound of evidence and introduces variational Bayesian inference (VBI) to solve the maximum a posteriori probability approximation:

[0113] .

[0114] Due to the variational distribution is an approximation. This paper introduces KL divergence to minimize variational distribution and the true posterior distribution The specific optimization formula is:

[0115] .

[0116] Expand the above formula to get:

[0117] ;

[0118] make: ;

[0119] This is the lower bound of evidence. The relaxed lower bound of evidence proposed in this invention is:

[0120] ;

[0121] in

[0122] ;

[0123] Where, is the relaxed evidence lower bound, is a variational distribution, which is used to approximate the true posterior distribution which is difficult to calculate ; is the generalized likelihood function; is the observation likelihood function; is the conditional probability; is a set of latent variables; is the variational parameter.

[0124] The improved sparse Bayesian learning (SBL) model of the present invention can reduce the computational complexity from 𝒪(N³) to 𝒪(FN) by introducing a relaxed evidence lower bound.

[0125] According to the alternating update algorithm, the optimal approximate solution can be updated as follows:

[0126] ;

[0127] ;

[0128] ;

[0129] ;

[0130] Where, For the iterations; is an additive scalar.

[0131] Will Substitution We can get:

[0132] ;

[0133] because For each can be separated, so:

[0134] ;

[0135] From this we can see that Obey Gaussian distribution .

[0136] Similarly, Substituting into other formulas we get:

[0137] ;

[0138] Obviously, Obeys the gamma distribution.

[0139] ;

[0140] therefore, For each are all separable:

[0141] ;

[0142] Gamma distribution .

[0143] ;

[0144] because Discrete, only one element is 1, and For each is also separable, so we only need to calculate it by exhaustive method , , which is expressed as follows:

[0145] ;

[0146] Finally, the optimal approximate solution is updated as follows:

[0147] ;

[0148] ;

[0149] ;

[0150] ;

[0151] Where, For the signal Probability density after iterations; Noise accuracy Probability density after iteration; is the sparsity Probability density after iteration; is the nth element of the signal The mean after iterations; is a block structure matrix Probability density after iterations; is the nth element of the signal covariance after iterations; is a gamma distribution; and Usually set to a small constant; is a constant; and is a smaller constant; .

[0152] In step S4, the present invention performs envelope demodulation on the separated signal and determines the specific fault type according to whether there is a bearing fault characteristic frequency.

[0153] Specifically, first, the separated signal is subjected to Hilbert transform to construct the analytical signal :

[0154] ;

[0155] Where, is the signal after Hilbert transform; Is an imaginary unit.

[0156] Then, extract the amplitude of the analytical signal:

[0157] .

[0158] Finally, the amplitude is Fourier transformed to obtain the envelope spectrum , and observe the characteristic frequency through the envelope spectrum.

[0159] Figure 3 In order to separate the fault signal envelope diagram, the fault frequency and its multiple of the inner and outer rings of the bearing are observed, and it can be determined whether the inner and outer rings of the rolling bearing are faulty.

[0160] Example 2:

[0161] like Figure 4 As shown, the present invention provides a rotating machinery compound fault diagnosis system, which is used to implement the rotating machinery compound fault diagnosis method of the above embodiment 1, specifically comprising:

[0162] The frequency band adaptive positioning module 100 is used to collect the fault signals of the inner and outer rings of the rolling bearing, locate the resonant frequency band of the fault signal based on the sliding window filter, separate the sub-signal set, calculate the modal optimization index value of each sub-signal one by one, select the first four sub-signals with the largest modal optimization index value, and eliminate redundant components by calculating the correlation coefficient between the signals, retaining the mode containing the fault characteristics;

[0163] The characteristic frequency estimation module 200 is used to estimate the fault characteristic frequency of the resonant frequency band based on envelope demodulation, perform frequency estimation on the remaining sub-signals using the envelope harmonic product spectrum, eliminate the non-converged sub-signals, and use the envelope harmonic product spectrum value of the remaining signal as the fault characteristic frequency;

[0164] The fault separation module 300 is used to classify the optimal sub-signals based on the improved sparse Bayesian learning classification model, and to solve the maximum posterior probability through variational Bayesian inference to achieve separation of composite fault signals;

[0165] The envelope demodulation module 400 is used to perform envelope demodulation on the separated signal and determine the specific fault type according to whether there is a bearing fault characteristic frequency.

[0166] A rotating machinery composite fault diagnosis system of the present embodiment is used to implement the aforementioned rotating machinery composite fault diagnosis method. Therefore, the specific implementation method of the rotating machinery composite fault diagnosis system can be found in the embodiment part of the rotating machinery composite fault diagnosis method mentioned above. For example, the frequency band adaptive positioning module 100, the characteristic frequency estimation module 200, the fault separation module 300, and the envelope demodulation module 400 are respectively used to implement steps S1, S2, S3, and S4 in the aforementioned rotating machinery composite fault diagnosis method. Therefore, its specific implementation method can refer to the description of the corresponding embodiments of each part. In order to avoid redundancy, it will not be repeated here.

[0167] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0168] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0169] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1A step that specifies a function in one or more boxes.

[0170] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications derived therefrom remain within the scope of protection of the present invention.

Claims

1. A rotating machinery composite fault diagnosis method, characterized in that: include: S1: Collect the fault signals of the inner and outer rings of the rolling bearing, locate the resonant frequency band of the fault signal based on the sliding window filter, separate the sub-signal set, calculate the modal optimization index value of each sub-signal one by one, select the first four sub-signals with the largest modal optimization index value, and eliminate the redundant components by calculating the correlation coefficient between the signals, retaining the mode containing the fault characteristics; wherein the modal optimization index value is composed of kurtosis, envelope spectrum kurtosis and mutual information, and is calculated as follows: in MI(x1,x2)=H(x2)-H(x2|x1); Where MOI is the modal optimization index value; Kurt is the kurtosis; ESK is the envelope spectrum kurtosis; MI is the mutual information; x(n) is the time domain sequence of the signal; is the mean of the time domain signal; N is the signal length; n represents the index; A(f n ) represents the envelope spectrum at frequency f n The amplitude at is the mean of the envelope spectrum; H(x2) is the entropy of signal x2; H(x2|x1) is the conditional entropy of signal x2 when signal x1 is known; S2: Estimate the fault characteristic frequency of the resonant frequency band based on envelope demodulation, use the envelope harmonic product spectrum to estimate the frequency of the remaining sub-signals, remove the non-convergent sub-signals, and use the envelope harmonic product spectrum value of the remaining signal as the fault characteristic frequency; S3: Classify the optimal sub-signals based on the improved sparse Bayesian learning classification model, and solve the maximum posterior probability through variational Bayesian inference to separate the composite fault signals; S4: Perform envelope demodulation on the separated signal and determine the specific fault type based on whether the bearing fault characteristic frequency exists.

2. The rotating machinery composite fault diagnosis method according to claim 1, characterized in that: The initialization of the sliding window filter includes: The original signal is divided into K segments, each segment is of length N, and the sampling frequency f s Determine the upper and lower cutoff frequencies f of each segment l and f u , construct K finite length unit impulse response filters, where the upper and lower cutoff frequencies f l and f u satisfy: Where k represents the index and K is the total number of filters.

3. The rotating machinery composite fault diagnosis method according to claim 1 or 2, characterized in that: The updating method of the sliding window filter comprises: The correlation kurtosis is used as the objective function to update the filter bank, and the generalized eigenvalue problem is solved iteratively to approximate the filtered signal corresponding to the maximum correlation kurtosis. The update formula is: Where, CK M is the relevant kurtosis value, M is the number of shifts; u k (n) is the signal component output by the kth filter; T s is the sampling period; N is the signal length; f k (l) is the coefficient vector of the kth filter; L is the filter length; l, m, n, k all represent indexes; st represents the constraint condition.

4. The rotating machinery composite fault diagnosis method according to claim 1, characterized in that: The method for estimating the frequency of the remaining sub-signal using the envelope harmonic product spectrum is: Where EHPS(w) is the envelope harmonic product spectrum value, F(w), F(2w), …, F(Hw), and F(hw) are the Fourier transform amplitudes of the signal envelope; H is the number of harmonics; h represents the index; and w is the angular frequency.

5. The rotating machinery composite fault diagnosis method according to claim 1, characterized in that: The improved sparse Bayesian learning classification model adopts a block sparse structure, and its mathematical expression is: The formula for the joint probability density p(y,Ω) is: p(y,Ω)=p(y|x,α)p(α)p(γ)p(x|γ,Z)p(Z); Where p(x|γ,Z) is the conditional probability of block sparse structure; Z is the block structure matrix; x n is the fault characteristic component, γ n+m-2 To control x n The sparsity hyperparameter, the subscript n+m-2 indicates the position within the block; z n,m is the block structure weight coefficient, and defines x n The degree of dependence on adjacent sparsity parameters; The mean is 0 and the variance is Gaussian distribution; x is the fault signal; y is the observation signal; Ω is the set of latent variables; p(y|x,α) is the likelihood function; p(α) is the prior distribution of noise precision α, p(γ) is the prior distribution of sparsity parameter γ; p(x|γ,Z) is the conditional probability of block sparse structure; p(Z) is the prior distribution of block structure matrix; N is the signal length; m and n both represent indexes.

6. The rotating machinery composite fault diagnosis method according to claim 5, characterized in that: The improved sparse Bayesian learning classification model is used to classify the optimal sub-signal and to update and solve the maximum posterior probability through variational Bayesian inference, specifically including: A relaxed evidence lower bound formula is designed, and the latent variables are iteratively updated through variational Bayesian inference to solve the maximum posterior probability approximate solution. The relaxed evidence lower bound calculation formula is: in G(y,Ω,ω)=F(y,x,ω,γ)p(x|α)p(α)p(γ); where L(q,ω) is the relaxed evidence lower bound, q(Ω) is the variational distribution used to approximate the difficult-to-calculate true posterior distribution p(Ω|y), G(y,Ω,ω) is the generalized likelihood function, F(y,x,ω,γ) is the observation likelihood function, p(x|α) is the conditional probability, Ω is the set of latent variables, and ω is the variational parameter.

7. The rotating machinery composite fault diagnosis method according to claim 6, characterized in that: The relaxed evidence lower bound minimizes the error between the variational distribution q(Ω) and the true posterior distribution p(Ω|y) by introducing the KL divergence. The specific optimization formula is: And by decomposing the latent variable update rules, the computational complexity is reduced.

8. The rotating machinery composite fault diagnosis method according to claim 1, characterized in that: The method for performing envelope demodulation on the separated signal is: First, the separated signal is Hilbert transformed to construct the analytical signal; then, the amplitude of the analytical signal is extracted; finally, the amplitude is Fourier transformed to obtain the envelope spectrum, and the characteristic frequency is observed through the envelope spectrum.

9. A rotating machinery composite fault diagnosis system, characterized in that: The system is used to implement the rotating machinery composite fault diagnosis method according to any one of claims 1 to 8, specifically comprising: The frequency band adaptive positioning module is used to collect the fault signals of the inner and outer rings of the rolling bearing, locate the resonant frequency band of the fault signal based on the sliding window filter, separate the sub-signal set, calculate the modal optimization index value of each sub-signal one by one, select the first four sub-signals with the largest modal optimization index value, and eliminate redundant components by calculating the correlation coefficient between the signals, retaining the mode containing the fault characteristics; The characteristic frequency estimation module is used to estimate the fault characteristic frequency of the resonant frequency band based on envelope demodulation, use the envelope harmonic product spectrum to estimate the frequency of the remaining sub-signals, eliminate the non-converged sub-signals, and use the envelope harmonic product spectrum value of the remaining signal as the fault characteristic frequency; The fault separation module is used to classify the optimal sub-signals based on the improved sparse Bayesian learning classification model, and to solve the maximum posterior probability through variational Bayesian inference to achieve separation of composite fault signals; The envelope demodulation module is used to perform envelope demodulation on the separated signal and determine the specific fault type based on whether there is a bearing fault characteristic frequency.

Images