Rotating machinery system fault identification method and device based on harmonic Fourier decomposition

Abstract

The embodiment of the invention provides a rotating mechanical system fault identification method and device based on harmonic Fourier decomposition, and the method comprises the steps: collecting an equipment signal, converting the equipment signal into a digital signal, carrying out the calculation based on the digital signal, obtaining a power spectrum density key function, and carrying out the fault identification of a rotating mechanical system. Performing inverse discrete Fourier transform on the power spectral density key function according to the interception length to obtain a power spectral density trend spectrum; searching a minimum value point in the power spectral density trend spectrum and setting the minimum value point as a boundary, segmenting a Fourier spectrum into different frequency bands based on the boundary, setting a zero-phase Fourier low-pass or band-pass filter in each frequency band, and reconstructing a signal in each filter into a component; the harmonic spectrum kurtosis of each component is calculated, and the component with the maximum harmonic spectrum kurtosis is extracted for fault diagnosis of the rotating mechanical system.

Description

Technical Field

[0001] This document relates to the field of computer technology, and in particular to a method, device, electronic device and medium for fault identification of a rotating mechanical system based on harmonic Fourier decomposition. Background Art

[0002] As rotating machinery becomes increasingly important in the industrial field, its fault identification methods have attracted the attention of scholars. There are various types of rotating equipment. Some large equipment faces harsh working environments, and some small and precise equipment has difficulty in collecting signals. The collected signals may be interfered and have little fault information, so highly adaptable and targeted algorithms are needed to decompose signals or extract useful information. Rotating machinery is in a periodic working environment, and bearings or gears are the key research targets. As the most vulnerable bearings or gears, they have also become the key research targets. The traditional empirical wavelet transform can adaptively decompose the signal into several intrinsic mode functions, which are assumed to be independent components with tight support characteristics that are unrelated to each other. Of course, many previous studies have proven the shortcomings of this method and pointed out some new research directions.

[0003] Empirical Wavelet Transform (EWT): The proposed Empirical Wavelet Transform (EWT) uses the extreme points of the Fourier spectrum to distinguish different empirical modes. In view of the uncertainty of the extreme points, the idea of ​​calculating the histogram based on the Fourier spectrum and segmenting the spectrum in the scale space representation is proposed. This idea is very consistent with the characteristics of bearing faults. EWT can construct an exclusive filter for each frequency band, and most of the information in the corresponding frequency band will be reconstructed into a time domain component. Although this segmentation technology is adaptive, it is easily interfered by noise and always brings a large number of invalid boundaries.

[0004] Analytical Mode Decomposition (AMD): An analytical mode decomposition (AMD) is proposed, which requires manual selection of useful frequencies and calculation of the bisection frequency. It decomposes the signal into high-frequency and low-frequency parts from the Fourier spectrum. Although the application of AMD in structural inspection has been verified, it has been rarely extended to bearing fault diagnosis.

[0005] Frequency Domain Decomposition (FDD): combines power spectral density (PSD) and singular value decomposition and constructs reference matrices that are pre-scaled and re-scaled.

[0006] Adaptive Fourier Decomposition Method (FDM): A multivariate nonlinear and nonstationary time series analysis method based on zero-phase filter banks is proposed. The adaptive Fourier decomposition method (FDM) draws on the basic decomposition mode of EMD and the spectrum segmentation idea of ​​EWT to decompose the signal into several Fourier eigenband functions (FIBFs). Dou et al. explored the feasibility of FDM in processing near-frequency signals and proved that FDM can suppress modal aliasing and end effects in EMD. Subsequently, FDM was applied to the fault diagnosis of turbine gearbox vibration signals, and the results were better than EEMD and VMD. Although FDM reduces modal aliasing outside the boundary, it is not suitable for analyzing multi-component signals with intersecting instantaneous frequencies. A zero-phase non-rectangular bandpass filter is used to decompose the signal to ensure that the sum of each component is the same as the original signal. It can be seen that FDM eliminates the stationarity and linearity requirements of Fourier-based methods, and also has a filter bank method of wavelet transform, which is very suitable for extracting fault information from bearing signals in the frequency domain. However, FDM requires the preset number of modes and optimized parameters, and there are some problems in the spectrum segmentation in current research that need to be solved.

[0007] Empirical Fourier Decomposition (EFD): The scale-space representation method in EWT is introduced into FDM and the empirical Fourier decomposition (EFD) is proposed. Although the adaptability of the spectrum segmentation method is improved, the model based on many extreme points is still strongly disturbed by noise and the number of extreme points.

[0008] In summary, the fault signal of the bearing, which is the most vulnerable to damage in rotating equipment, has the characteristic of concentrated energy in the spectrum, which brings great opportunities for the application of one-dimensional signal decomposition methods in engineering. The traditional empirical wavelet transform (EWT) can calculate the histogram based on the Fourier spectrum and segment the spectrum in the scale space representation. This idea fits the characteristics of bearing faults. Although this segmentation technology is adaptive, it is easily interfered by noise and always brings a large number of invalid boundaries. The adaptive Fourier decomposition method (FDM) draws on the basic decomposition mode of EMD and the spectrum segmentation idea of ​​EWT to decompose the signal into several Fourier eigenband functions. Although FDM reduces the modal aliasing outside the boundary, it is not suitable for analyzing multi-component signals with intersecting instantaneous frequencies. The empirical Fourier decomposition method (EFD) introduces the scale-space representation method in EWT into FDM. Although the adaptability of the spectrum segmentation method is improved, the mode based on many extreme points is still strongly interfered by noise and the number of extreme points.

[0009] Therefore, as the accuracy of the sensor is updated, there may be a lot of complex information in the collected signal. Weak fault information or more high-frequency noise will make fault diagnosis and signal decomposition difficult. In the process of signal decomposition, unreasonable modal segmentation is the main factor that brings many invalid components. Incomplete filtering methods may also add unnecessary noise to the reconstructed components. Summary of the invention

[0010] The object of the present invention is to provide a method, device, electronic device and medium for fault identification of a rotating mechanical system based on harmonic Fourier decomposition, aiming to solve the above-mentioned problems in the prior art.

[0011] The present invention provides a method for fault identification of a rotating mechanical system based on harmonic Fourier decomposition, comprising:

[0012] Collecting device signals and converting the device signals into digital signals, calculating a power spectrum density key function based on the digital signals, performing an inverse discrete Fourier transform on the power spectrum density key function according to a cut length, and obtaining a power spectrum density trend spectrum;

[0013] Searching for a minimum point in the power spectrum density trend spectrum and setting it as a boundary, dividing the Fourier spectrum into different frequency bands based on the boundary, setting a corresponding filter in each frequency band, and reconstructing the signal in each filter into a component;

[0014] The harmonic spectral kurtosis of each component is calculated and the component with the maximum harmonic spectral kurtosis is extracted for fault diagnosis of rotating machinery systems.

[0015] The present invention provides a rotating machinery system fault identification device based on harmonic Fourier decomposition, comprising:

[0016] A calculation module is used to collect device signals and convert the device signals into digital signals, calculate a power spectrum density key function based on the digital signals, perform an inverse discrete Fourier transform on the power spectrum density key function according to a cut length, and obtain a power spectrum density trend spectrum;

[0017] A segmentation and reconstruction module, used for searching for a minimum point in the power spectrum density trend spectrum and setting it as a boundary, segmenting the Fourier spectrum into different frequency bands based on the boundary, setting a corresponding filter in each frequency band, and reconstructing the signal in each filter into a component;

[0018] The fault diagnosis module is used to calculate the harmonic spectrum kurtosis of each component and extract the component with the maximum harmonic spectrum kurtosis for fault diagnosis of the rotating mechanical system.

[0019] An embodiment of the present invention further provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the above-mentioned rotating machinery system fault identification method when executed by the processor.

[0020] An embodiment of the present invention further provides a computer-readable storage medium, on which a program for implementing information transmission is stored, and when the program is executed by a processor, the steps of the above-mentioned rotating machinery system fault identification method are implemented.

[0021] By adopting the embodiment of the present invention, based on the harmonic Fourier decomposition method (HFD), the fluctuation trend of the spectrum is obtained by calculating the Fourier trend spectrum of the power spectrum density, which reduces the complexity of the spectrum segmentation object on the one hand, and suppresses the noise and reduces the number of extreme points on the other hand. The simulation signal proves that this boundary segmentation and mode recognition method is superior to the adaptive scale space representation method. In order to completely extract the information in each frequency band, HFD adopts a zero-phase Fourier filter group. The construction method of this filter can reduce energy leakage and improve filtering efficiency. In order to improve the adaptability of the algorithm, quantize the periodic pulse information in the component and suppress the interference of random pulses and noise, the embodiment of the present invention also adopts the harmonic spectrum kurtosis (HSK). The simulation signal and the experimental signal verify that the proposed method can successfully extract the fault characteristics of the inner and outer rings of the bearing. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0023] Figure 1 is a flow chart of a method for fault identification of a rotating machinery system based on harmonic Fourier decomposition according to an embodiment of the present invention;

[0024] Figure 2 is a flowchart of a detailed process of a method for identifying a fault of a rotating machinery system according to an embodiment of the present invention;

[0025] Figure 3 Schematic diagram of the process of calculating the PSD trend spectrum and boundary according to an embodiment of the present invention;

[0026] Figure 4 is a schematic diagram of a Fourier filter bank of an HFD and a Meyer filter bank according to an embodiment of the present invention;

[0027] Figure 5is a schematic diagram of a simulated signal and its spectrum for verifying a filtering effect according to an embodiment of the present invention;

[0028] Figure 6 is a schematic diagram of filtering results of three algorithms when the signal-to-noise ratio is set to -3dB in an embodiment of the present invention;

[0029] Figure 7 is a schematic diagram of energy loss and filtering efficiency in different frequency bands decomposed according to an embodiment of the present invention;

[0030] Figure 8 is a schematic diagram of HSK sensitivity of five signal components under different SNR conditions in an embodiment of the present invention;

[0031] Fig. 9 is a schematic diagram of a rotating machinery system fault identification device based on harmonic Fourier decomposition according to an embodiment of the present invention;

[0032] Fig.10 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more 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 this document.

[0034] Method Embodiment

[0035] According to an embodiment of the present invention, a method for fault identification of a rotating machinery system based on harmonic Fourier decomposition is provided. Figure 1 FIG. 1 is a flow chart of a method for identifying a rotating mechanical system fault based on harmonic Fourier decomposition according to an embodiment of the present invention. Figure 1 As shown, the method for identifying faults in a rotating machinery system according to an embodiment of the present invention specifically includes:

[0036] Step S101, collecting device signals and converting the device signals into digital signals, calculating the power spectrum density key function based on the digital signals, performing inverse discrete Fourier transform on the power spectrum density key function according to the interception length, and obtaining the power spectrum density trend spectrum; specifically comprising:

[0037] The vibration data or sound data of the equipment is collected and converted into a digital signal, the power spectrum density and its Fourier transform of the digital signal are calculated to obtain the power spectrum density key function, and the power spectrum density key function is subjected to inverse discrete Fourier transform according to the cut length to obtain a power spectrum density trend spectrum representing the power spectrum density fluctuation trend.

[0038] Specifically: the power spectrum density of the digital signal is determined according to formula 1:

[0039]

[0040] Where T represents the length of the time interval, P represents the average power of the signal, y(t) represents the original signal, and t represents the time variable;

[0041] The Fourier transform of the signal in the interval [0, T] of the power spectrum density is calculated according to Formula 2, and the key function of the power spectrum density is determined according to Formula 3:

[0042]

[0043] in, represents the Fourier transform of the signal, and f represents the frequency variable;

[0044]

[0045] Among them, s y (f) represents the power spectrum density of the signal, and E represents the mathematical expectation;

[0046] Will s y (f) Discretized into s y (n), n=1,2,3,…,N, N is s y The length of the power spectrum density key function is subjected to inverse discrete Fourier transform according to formula 4 according to the intercept length:

[0047]

[0048] in, is the key function of power spectral density, u represents the discrete frequency variable, n=1,2,…,N;

[0049] According to Formula 5, the power spectrum density key function is subjected to inverse discrete Fourier transform according to the intercept length to obtain a power spectrum density trend spectrum representing the power spectrum density fluctuation trend:

[0050]

[0051] Among them, T k (ω) is the power spectral density trend spectrum, is the k-order spectrum estimator of the signal.

[0052] Step S102, searching for the minimum point in the power spectral density trend spectrum and setting it as a boundary, dividing the Fourier spectrum into different frequency bands based on the boundary, setting a corresponding filter in each frequency band, and reconstructing the signal in each filter into a component; wherein the corresponding filter is a zero-phase Fourier low-pass or band-pass filter, specifically comprising:

[0053] The minimum point in the power spectral density trend spectrum is searched and set as the boundary, and the Fourier spectrum is divided into a number of frequency bands with different center frequencies based on the boundary; an exclusive zero-phase Fourier low-pass filter or band-pass filter is set in each frequency band, and the signal in each zero-phase Fourier low-pass filter or band-pass filter is reconstructed into a component.

[0054] Specifically:

[0055] The essence of harmonic Fourier decomposition is to accurately reconstruct the information in each frequency band through a set of zero-phase filters. The zero-phase filter is defined by using non-negative real values. The Fourier spectrum is calculated and normalized in the range of [0,π]. A set of boundaries is obtained based on the modal identification method based on the power spectrum. The Fourier spectrum is divided into several parts based on the boundaries. Each part is regarded as a useful mode. The number of empirical modes is assumed to be N, and ω is defined as the obtained boundary, then ω0=0,ω N =π. Each frequency band is described as: Λ1=[ω0,ω1], Λ2=[ω1,ω2], Λ3=[ω2,ω3], Λ4=[ω3,ω4]… Let n=1,2,…,N,Λ n =[ω n-1 ,ω n ], each frequency band is represented as:

[0056] The analysis signal is defined according to Equation 6:

[0057]

[0058] Where m = 1, 2, ..., M, H[] represents the Hilbert transform of the signal, It is estimated by discrete Fourier transform, h m (t) is used to represent the system’s response to different inputs, z m (t) corresponds to h m (t) analysis signal;

[0059] Extract each analysis signal, construct a filter bank through the obtained boundaries, and define the scaling function according to Formula 7 and Formula 8 and empirical function

[0060]

[0061] Among them, ω n Indicates the frequency division point;

[0062] Let the Fourier transform be F(·) and the inverse Fourier transform be F -1 (·), determine the approximate coefficient according to Formula 9

[0063]

[0064] Among them, y() represents the input signal, represents the Fourier transform of the signal, φ1(t) represents a basis function, τ, t represents the time variable, represents the Fourier transform of φ1(t);

[0065] Determine the detail coefficient according to formula 10:

[0066]

[0067] in, represents the detail coefficient;

[0068] The signal is reconstructed according to formula 11:

[0069]

[0070] in, express Fourier transform of

[0071] Determine the empirical mode according to formula 12:

[0072]

[0073] Step S103, calculate the harmonic spectrum kurtosis of each component and extract the component with the maximum harmonic spectrum kurtosis to perform fault diagnosis of the rotating mechanical system. Specifically, since the bearing fault signal may not only have pulses with periodic characteristics, but also single pulses may appear in uncertain noise. The operation of the equipment will generate modulation information to interfere with the diagnosis method. In order to accurately filter the inner ring or outer ring fault information from the obtained frequency band, the harmonic spectrum kurtosis (HSK) is used to filter or locate the fault information in each frequency band. The most important feature of HSK is to identify the fault characteristic frequency and its harmonics from the envelope spectrum. There are periodic pulses with similar amplitudes in the waveform of the outer ring fault signal of the bearing, and there are amplitude-modulated periodic pulses in the inner ring fault signal. This is the biggest difference between the two fault characteristics in the time domain waveform. The spectrum of the two types of fault signals includes the center frequency and bandwidth, and the fault information covers the entire spectrum. It is time-consuming and unreasonable to judge each envelope spectrum by observation. If a certain frequency and its possible harmonics are superimposed or multiplied, the total energy of the characteristic frequency and its harmonics will appear in the new spectrum. HSK can not only suppress the interference of single pulse, random noise and modulation information, but also is sensitive to the fault information of the inner and outer rings of the bearing.

[0074] Specifically include:

[0075] The Wold-Cramer decomposition of the stationary process in the frequency domain is described as Equation 13:

[0076]

[0077] Where G(f) represents the Hilbert transform of g(t), d X(f) Represents the spectrum process of signal x(t);

[0078] After adding the time-varying information, the non-stationary signal is summarized as Equation 14:

[0079]

[0080] Where g(t,t-τ) represents the impulse response function of the system, x(τ) represents an input signal that varies with time, and τ represents time;

[0081] Using P(t,f) to represent the complex envelope of y(t) at f, the frequency counterpart describes Equation 15:

[0082]

[0083] Wherein, P(t,f) represents the complex envelope used to represent y(t) at f;

[0084] The random cyclostationary process is defined by formula 16:

[0085] g(t,s)=g(t+T,s)=∑k g k (s)e j2πkt / T ;Formula 16;

[0086] Among them, g(t,s) represents the periodic function of time;

[0087] The energy intensity of the complex envelope at t and f is measured by the 2n-order instantaneous moment according to Equation 17:

[0088] S 2nY (t,f)=E{|P(t,f)d x(f) | 2n} / d f =|P(t,f)| 2n ·S 2nX ;Formula 17;

[0089] Among them, the 2n-order instantaneous moment S 2nY (t,f) represents the instantaneous spectrum or time-frequency energy density of y(t);

[0090] Assume n = 1, and determine the spectral moment according to formula 18:

[0091] S 2Y (f) = E{S 2Y (t,f)}=E{|P(t,f)d x(f) | 2} / d f =E{|P(t,f)| 2}·S 2X ;

[0092] Formula 18;

[0093] Among them, S 2Y (f) represents the second-order spectral density of the signal, S 2Y (t,f) represents the second-order time-frequency representation of the signal;

[0094] When characterizing a conditionally non-stationary process, the 2n-order mean moment is defined according to Formula 19:

[0095]

[0096] The instantaneous moment of the characteristic time is defined according to formula 20:

[0097] g0(t,Δt)=∑ n g n (nt)e j2πnt / T ;Formula 20;

[0098] Among them, g0(t,Δt) represents a function that may intercept periodic instantaneous pulses in the region (t,Δt) close to g(t,s);

[0099] The local energy of the complex envelope at t and f is measured by the 2nth harmonic instantaneous moment based on Equation 21:

[0100] S 2HY (t,f)=E{|P n (t,f)d X(f) | 2} / d f =|P n (t,f)| 2 ·S 2x ;Formula 21;

[0101] Among them, S 2HY (t,f) represents the local energy of the complex envelope at t and f;

[0102] Determine S according to formula 22 2HY The spectral moment of (t,f) is:

[0103] S 2HY (f) = E{S 2HY (t,f)}=E{|P n (t,f)d X(f) | 2} / d f =E{|P n (t,f)| 2}·S 2X ;

[0104] Formula 22;

[0105] Among them, S 2HY (f) indicates S 2HY spectral moment of (t,f);

[0106] Determine the 2n-order harmonic average moment according to formula 23:

[0107]

[0108] Among them, S 2HY (f) represents the average moment of 2n-order harmonics;

[0109] The normalized fourth-order harmonic spectral moment of y(t) is determined according to formula 24:

[0110] C 4HY (f) = S 4HY (f); f≠0; Formula 24;

[0111] Among them, C 4HY (f) represents the normalized fourth-order harmonic spectrum moment of y(t); S 4HY (f) represents the average moment of 4n-order harmonics;

[0112] The harmonic spectrum kurtosis is defined as the energy-normalized fourth-order harmonic spectrum cumulant according to formula 25:

[0113]

[0114] Among them, HSK Y (f) represents the kurtosis value of the harmonic spectrum, C 4HY (f) represents the fourth-order harmonic spectral moment, Represents the square of the second-order spectral density of the signal.

[0115] The above technical solution of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings.

[0116] The method for identifying faults in a rotating mechanical system based on harmonic Fourier decomposition according to an embodiment of the present invention is shown in the flowchart of the specific method. Figure 2 As shown, the method comprises the following steps:

[0117] Step (1): Collect the vibration data or sound of the equipment and convert it into a digital signal, calculate the power spectral density and Fourier transform of the input signal, obtain the PSD key function, perform inverse discrete Fourier transform on the PSD key function according to the interception length, and obtain the PSD trend spectrum representing the PSD fluctuation trend. Figure 3 This is the step of calculating the PSD trend spectrum and boundary in the present invention.

[0118] Step (2): Search for the minimum points in the PSD trend spectrum and set them as boundaries. These boundaries will divide the Fourier spectrum into several frequency bands with different center frequencies. Set a dedicated zero-phase Fourier low-pass filter or band-pass filter for each frequency band. The signal in each filter can be reconstructed into a component. Figure 4 To show the details of the Fourier filter (FFB) used in this invention and its difference from the Meyer filter (MFB) used in EWT, it can be found that since the Fourier filter used in the present invention removes the transition section, the energy leakage can be greatly reduced.

[0119] Step (3): Calculate the harmonic spectral kurtosis (HSK) of each component and extract the component with the largest HSK for fault diagnosis.

[0120] In order to verify that HFD still has efficient filtering performance for complex signals, the present invention simulates the outer race fault of the bearing. This type of signal has several characteristics. From the waveform in the time domain, the signal has several pulses with similar amplitudes, which has obvious periodicity; from the spectrum, the signal has a center frequency (2500Hz) and a sideband (100Hz). Obviously, the width of the sideband is related to the period of the waveform.

[0121]

[0122] The damping coefficient g = 0.07, the natural frequency f n = 2500Hz. The repetition period of the periodic pulse is T = 0.01s. The analog signal and its spectrum are as follows Figure 5 shown.

[0123] Figure 5 The displayed signal is an ideal model without noise, which does not exist in actual conditions. Therefore, the present invention adds noise of different intensities to the simulated signal. When the signal contains strong noise, the pulses in the time domain waveform may be buried, which means that the periodicity is difficult to observe. The center frequency of the spectrum may also be buried. The part far away from the center frequency contains far more noise than fault information. Two boundaries are set: [2200Hz, 2800Hz], and the spectrum is divided into three parts. The frequency bands between the boundaries contain almost most of the fault information, which helps signal decomposition and noise reduction. The three algorithms HFD, EWT and FIR are used to extract the frequency band with a center frequency of 2500Hz. Band A with a frequency below 2200Hz and band C with a frequency above 2800Hz are treated as noise, and band B is treated as useful information. When SNR = -3dB, Figure 6 The results of three methods for extracting Band B are shown. It is easy to see that HFD has less energy leakage in Band A / C and has the most efficient filtering.

[0124] In order to verify the reliability of the above conclusions, 21 groups of strong noise [-20dB~0dB] were added to the signal. The energy loss of bands A and C and the filtering efficiency of band B were calculated. Figure 7 The processing results of HFD, EWT and FIR are shown. The energy loss of HFD in frequency band A is within 4%, the energy loss of EWT is between 1% and 8%, and the error fluctuation of FIR is very strong. The energy loss of HFD and EWT in frequency band C is similar, with an overall loss of less than 8%, and the energy loss of FIR reaches 15%. In frequency band B, there is a significant difference in the filtering efficiency of the three methods. HFD has the highest filtering efficiency, followed by EWT, and the filtering efficiency of FIR is less than 80%.

[0125] In order to verify the anti-noise capability of HSK, the present invention adds noise with a signal-to-noise ratio of [-20dB, 20dB] to five analog signals. Figure 8HSK with five components under different SNR conditions is shown. When the SNR is higher than -10dB, HSK is more sensitive to periodic pulse signals than other interference signals. When the SNR is lower than -14dB, the sensitivity of HSK decreases, and it may even be impossible to extract fault information. Of course, it is difficult for the environmental noise to reach -14 decibels, and this extreme environment needs to be explored in subsequent research. It can be seen that HSK has strong anti-noise ability, and in theory, it can realize fault feature extraction in practical applications.

[0126] Device Example 1

[0127] According to an embodiment of the present invention, a rotating machinery system fault identification device based on harmonic Fourier decomposition is provided. Fig. 9 Schematic diagram of a rotating machinery system fault identification device according to an embodiment of the present invention. Fig. 9 As shown, the rotating machinery system fault identification device according to an embodiment of the present invention specifically includes:

[0128] The calculation module 90 is used to collect device signals and convert the device signals into digital signals, calculate the power spectrum density key function based on the digital signals, perform inverse discrete Fourier transform on the power spectrum density key function according to the interception length, and obtain the power spectrum density trend spectrum; specifically used for:

[0129] Collect vibration data or sound data of the equipment and convert it into a digital signal, calculate the power spectrum density and Fourier transform of the digital signal, obtain a power spectrum density key function, perform an inverse discrete Fourier transform on the power spectrum density key function according to the interception length, and obtain a power spectrum density trend spectrum representing the fluctuation trend of the power spectrum density;

[0130] The segmentation and reconstruction module 92 is used to search for the minimum point in the power spectrum density trend spectrum and set it as a boundary, divide the Fourier spectrum into different frequency bands based on the boundary, set a corresponding filter in each frequency band, and reconstruct the signal in each filter into a component; specifically used for:

[0131] The minimum point in the power spectral density trend spectrum is searched and set as the boundary, and the Fourier spectrum is divided into a number of frequency bands with different center frequencies based on the boundary; an exclusive zero-phase Fourier low-pass filter or band-pass filter is set in each frequency band, and the signal in each zero-phase Fourier low-pass filter or band-pass filter is reconstructed into a component.

[0132] The fault diagnosis module 94 is used to calculate the harmonic spectrum kurtosis of each component and extract the component with the maximum harmonic spectrum kurtosis to perform fault diagnosis of the rotating mechanical system. Specifically, since the bearing fault signal may not only have pulses with periodic characteristics, but also single pulses may appear in uncertain noise. The operation of the equipment will generate modulation information to interfere with the diagnosis method. In order to accurately filter the inner ring or outer ring fault information from the obtained frequency band, the harmonic spectrum kurtosis (HSK) is used to filter or locate the fault information in each frequency band. The most important feature of HSK is to identify the fault characteristic frequency and its harmonics from the envelope spectrum. There are periodic pulses with similar amplitudes in the waveform of the outer ring fault signal of the bearing, and there are amplitude-modulated periodic pulses in the inner ring fault signal. This is the biggest difference between the two fault characteristics in the time domain waveform. The spectrum of the two types of fault signals includes the center frequency and bandwidth, and the fault information covers the entire spectrum. It is time-consuming and unreasonable to judge each envelope spectrum by observation. If a certain frequency and its possible harmonics are superimposed or multiplied, the total energy of the characteristic frequency and its harmonics will appear in the new spectrum. HSK can not only suppress the interference of single pulse, random noise and modulation information, but also is sensitive to the fault information of the inner and outer rings of the bearing.

[0133] The embodiment of the present invention is a device embodiment corresponding to the above method embodiment. The specific operations of each module can be understood by referring to the description of the method embodiment, which will not be repeated here.

[0134] Device Example 2

[0135] An embodiment of the present invention provides an electronic device, such as Fig.10 As shown, it includes: a memory 100, a processor 102, and a computer program stored in the memory 100 and executable on the processor 102, and when the computer program is executed by the processor 102, the steps described in the method embodiment are implemented.

[0136] Device Example 3

[0137] An embodiment of the present invention provides a computer-readable storage medium, on which a program for implementing information transmission is stored. When the program is executed by the processor 102, the steps described in the method embodiment are implemented.

[0138] The computer-readable storage medium in this embodiment includes, but is not limited to, ROM, RAM, magnetic disk or optical disk, etc.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for fault identification of rotating machinery system based on harmonic Fourier decomposition, characterized in that: include: Collecting device signals and converting the device signals into digital signals, calculating a power spectrum density key function based on the digital signals, performing an inverse discrete Fourier transform on the power spectrum density key function according to a cut length, and obtaining a power spectrum density trend spectrum; Searching for a minimum point in the power spectrum density trend spectrum and setting it as a boundary, dividing the Fourier spectrum into different frequency bands based on the boundary, setting a corresponding filter in each frequency band, and reconstructing the signal in each filter into a component; The harmonic spectral kurtosis of each component is calculated and the component with the maximum harmonic spectral kurtosis is extracted for fault diagnosis of rotating machinery systems.

2. The method according to claim 1, characterized in that Collecting device signals and converting the device signals into digital signals, calculating the key function of power spectrum density based on the digital signals, performing inverse discrete Fourier transform on the key function of power spectrum density according to the interception length, and obtaining the trend spectrum of power spectrum density specifically includes: The vibration data or sound data of the equipment is collected and converted into a digital signal, the power spectrum density and its Fourier transform of the digital signal are calculated to obtain the power spectrum density key function, and the power spectrum density key function is subjected to inverse discrete Fourier transform according to the cut length to obtain a power spectrum density trend spectrum representing the power spectrum density fluctuation trend.

3. The method according to claim 2, characterized in that The vibration data or sound data of the equipment is collected and converted into a digital signal, the power spectrum density and its Fourier transform of the digital signal are calculated, the power spectrum density key function is obtained, and the power spectrum density key function is subjected to inverse discrete Fourier transform according to the interception length to obtain a power spectrum density trend spectrum representing the fluctuation trend of the power spectrum density: The power spectral density of the digital signal is determined according to Formula 1: Where T represents the length of the time interval, P represents the average power of the signal, y(t) represents the original signal, and t represents the time variable; The Fourier transform of the signal in the interval [0, T] of the power spectrum density is calculated according to Formula 2, and the key function of the power spectrum density is determined according to Formula 3: in, represents the Fourier transform of the signal, and f represents the frequency variable; Among them, s y (f) represents the power spectrum density of the signal, and E represents the mathematical expectation; Will s y (f) Discretized into s y (n), n=1,2,3,…,N, N is s y The length of the power spectrum density key function is subjected to inverse discrete Fourier transform according to formula 4 according to the intercept length: in, is the key function of power spectral density, u represents the discrete frequency variable, n=1,2,…,N; According to Formula 5, the power spectrum density key function is subjected to inverse discrete Fourier transform according to the intercept length to obtain a power spectrum density trend spectrum representing the power spectrum density fluctuation trend: Among them, T k (ω) is the power spectral density trend spectrum, is the k-order spectrum estimator of the signal.

4. The method according to claim 1, characterized in that: Searching for a minimum point in the power spectrum density trend spectrum and setting it as a boundary, dividing the Fourier spectrum into different frequency bands based on the boundary, setting a corresponding filter in each frequency band, and reconstructing the signal in each filter into a component specifically includes: The minimum point in the power spectral density trend spectrum is searched and set as the boundary, and the Fourier spectrum is divided into a number of frequency bands with different center frequencies based on the boundary; an exclusive zero-phase Fourier low-pass filter or band-pass filter is set in each frequency band, and the signal in each zero-phase Fourier low-pass filter or band-pass filter is reconstructed into a component.

5. The method according to claim 4, characterized in that Searching for a minimum point in the power spectral density trend spectrum and setting it as a boundary, dividing the Fourier spectrum into a number of frequency bands with different center frequencies based on the boundary; setting an exclusive zero-phase Fourier low-pass filter or band-pass filter in each frequency band, and reconstructing the signal in each zero-phase Fourier low-pass filter or band-pass filter into a component specifically includes: The Fourier spectrum is calculated and normalized in the range of [0,π]. A set of boundaries is obtained based on the modal identification method based on the power spectrum. The Fourier spectrum is divided into several parts based on the boundaries. Each part is regarded as a useful mode. The number of empirical modes is assumed to be N. ω is defined as the obtained boundary, then ω0=0,ω N =π. Each frequency band is described as: Λ1=[ω0,ω1], Λ2=[ω1,ω2], Λ3=[ω2,ω3], Λ4=[ω3,ω4]… Let n=1,2,…,N,Λ n =[ω n-1 ,ω n ], each frequency band is represented as: The analysis signal is defined according to Equation 6: Where m = 1, 2, ..., M, H[] represents the Hilbert transform of the signal, It is estimated by discrete Fourier transform, h m (t) is used to represent the system’s response to different inputs, z m (t) corresponds to h m (t) analysis signal; Extract each analysis signal, construct a filter bank through the obtained boundaries, and define the scaling function according to Formula 7 and Formula 8 and empirical function Among them, ω n Indicates the frequency division point; Let the Fourier transform be F(·) and the inverse Fourier transform be F -1 (·), determine the approximate coefficient according to Formula 9 Among them, y() represents the input signal, represents the Fourier transform of the signal, φ1(t) represents a basis function, τ, t represents the time variable, represents the Fourier transform of φ1(t); Determine the detail coefficient according to formula 10: in, represents the detail coefficient; The signal is reconstructed according to formula 11: in, express Fourier transform of Determine the empirical mode according to formula 12: 。 6. The method according to claim 1, characterized in that Calculating the harmonic spectrum kurtosis of each component and extracting the component with the maximum harmonic spectrum kurtosis for fault diagnosis of rotating machinery system specifically includes: The Wold-Cramer decomposition of the stationary process in the frequency domain is described as Equation 13: Where G(f) represents the Hilbert transform of g(t), d X(f) Represents the spectrum process of signal x(t); After adding the time-varying information, the non-stationary signal is summarized as Equation 14: Where g(t,t-τ) represents the impulse response function of the system, x(τ) represents an input signal that varies with time, and τ represents time; Using P(t,f) to represent the complex envelope of y(t) at f, the frequency counterpart describes Equation 15: Wherein, P(t,f) represents the complex envelope used to represent y(t) at f; The random cyclostationary process is defined by formula 16: g(t,s) = g(t + T,s) = ∑ k g k (s)e j2πkt / T ; Formula 16; Among them, g(t,s) represents the periodic function of time; The energy intensity of the complex envelope at t and f is measured by the 2n-order instantaneous moment according to Equation 17: S 2nY (t,f)E{|P(t,f)d x(f) | 2n } / d f Z|P(t,f)| 2n ·S 2nX 10 / 17 / 1 Among them, the 2n-order instantaneous moment S 2nY (t,f) represents the instantaneous spectrum or time-frequency energy density of y(t); Assume n = 1, and determine the spectral moment according to formula 18: S 2Y (f)=E{S 2Y (t,f)}=E{|P(t,f)d X(f) | 2 } / d f =E{|P(t,f)| 2 }·S 2X ; Formula 18; Among them, S 2Y (f) represents the second-order spectral density of the signal, S 2Y (t,f) represents the second-order time-frequency representation of the signal; When characterizing a conditionally non-stationary process, the 2n-order mean moment is defined according to Formula 19: The instantaneous moment of the characteristic time is defined according to formula 20: g0(t,Δt)=∑ n g n (nt)e j2πnt / T ;Formula 20; Among them, g0(t,Δt) represents a function that may intercept periodic instantaneous pulses in the region (t,Δt) close to g(t,s); The local energy of the complex envelope at t and f is measured by the 2nth harmonic instantaneous moment based on Equation 21: S 2HY (t,f) = E{|P n (t,f)d X(f) | 2} / d f = |P n (t,f)| 2 ·S 2X ; Equation 21; Among them, S 2HY (t,f) represents the local energy of the complex envelope at t and f; Determine S according to formula 22 2HY The spectral moment of (t,f) is: S 2HY (f)=E{S 2HY (t,f)}=E{|P n (t,f)d X(f) | 2 } / d f =E{|P n (t,f)| 2 }·S 2X ; Formula 22; Among them, S 2HY (f) indicates S 2HY The spectral moment of (t,f); Determine the 2n-order harmonic average moment according to formula 23: Among them, S 2HY (f) represents the average moment of 2n-order harmonics; The normalized fourth-order harmonic spectral moment of y(t) is determined according to formula 24: C 4HY (f) = S 4HY (f); f≠0; Formula 24; Among them, C 4HY (f) represents the normalized fourth-order harmonic spectrum moment of y(t); S 4HY (f) represents the average moment of 4n-order harmonics; The harmonic spectrum kurtosis is defined as the energy-normalized fourth-order harmonic spectrum accumulation according to formula 25: Among them, HSK Y (f) represents the kurtosis value of the harmonic spectrum, C 4HY (f) represents the fourth-order harmonic spectral moment, Represents the square of the second-order spectral density of the signal.

7. A rotating machinery system fault identification device based on harmonic Fourier decomposition, characterized in that: include: A calculation module is used to collect device signals and convert the device signals into digital signals, calculate a power spectrum density key function based on the digital signals, perform an inverse discrete Fourier transform on the power spectrum density key function according to a cut length, and obtain a power spectrum density trend spectrum; A segmentation and reconstruction module, used for searching for a minimum point in the power spectrum density trend spectrum and setting it as a boundary, segmenting the Fourier spectrum into different frequency bands based on the boundary, setting a corresponding filter in each frequency band, and reconstructing the signal in each filter into a component; The fault diagnosis module is used to calculate the harmonic spectrum kurtosis of each component and extract the component with the maximum harmonic spectrum kurtosis for fault diagnosis of the rotating mechanical system.

8. The device according to claim 7, characterized in that The calculation module is specifically used for: Collect vibration data or sound data of the equipment and convert it into a digital signal, calculate the power spectrum density and Fourier transform of the digital signal, obtain a power spectrum density key function, perform an inverse discrete Fourier transform on the power spectrum density key function according to the interception length, and obtain a power spectrum density trend spectrum representing the fluctuation trend of the power spectrum density; The segmentation and reconstruction module is specifically used for: The minimum point in the power spectral density trend spectrum is searched and set as the boundary, and the Fourier spectrum is divided into a number of frequency bands with different center frequencies based on the boundary; an exclusive zero-phase Fourier low-pass filter or band-pass filter is set in each frequency band, and the signal in each zero-phase Fourier low-pass filter or band-pass filter is reconstructed into a component.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the rotating machinery system fault identification method as claimed in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an implementation program for information transmission, and when the program is executed by a processor, the steps of the rotating machinery system fault identification method according to any one of claims 1 to 6 are implemented.

Images