Abnormal vibration signal diagnosis method based on unknown frequency band morlet wavelet

Abstract

The invention discloses an abnormal vibration signal diagnosis method based on an unknown frequency band morlet wavelet, and relates to the technical field of bearing fault detection, and the method comprises the steps: signal collection, LMD decomposition, PF component information entropy calculation, high-frequency resonant frequency selection, center frequency and bandwidth determination, band-pass filtering, envelope demodulation and spectrum analysis. The method is technically characterized in that an LMD nonlinear and non-stationary signal analysis method is adopted, self-adaptive decomposition is carried out according to information of signals, and generated PF components have real physical significance; according to the LMD method, frequency bands can be adaptively selected according to the characteristics of the signals, the resolutions of the signals in different frequency bands can be determined, the accuracy of extracting effective information is improved, and the LMD method is very suitable for multi-component nonlinear and non-stationary signals.

Description

Technical Field

[0001] The invention relates to the technical field of bearing fault detection, and in particular to an abnormal vibration signal diagnosis method based on unknown frequency band morlet wavelet. Background Art

[0002] Bearing fault detection is to monitor and diagnose the operating status of bearings to determine whether there are any abnormalities or potential faults. Bearings are key components in mechanical equipment, supporting rotating parts and reducing friction. Due to various factors (such as wear, overload, poor lubrication, etc.), bearings may fail. If these faults are not discovered and handled in time, they may cause equipment damage, production interruption or even safety accidents.

[0003] Therefore, it is very important to carry out bearing fault detection, which helps to detect abnormal conditions of bearings in advance and take timely measures to repair or replace them, thus avoiding more serious consequences. There are various methods for bearing fault detection, including but not limited to vibration analysis, acoustic analysis, temperature monitoring, oil analysis, etc. Each method has its unique advantages and applicable scenarios. For example, vibration analysis can detect bearing wear, cracks and other faults by measuring bearing vibration signals; acoustic analysis determines the state of bearings by capturing sound signals during operation; temperature monitoring detects abnormal conditions by measuring temperature changes of bearings; and oil analysis evaluates the degree of bearing wear by analyzing impurities such as particles and moisture in the bearing lubricant.

[0004] Then there are many problems with the existing detection methods and approaches, specifically:

[0005] First, rolling bearings are key components of rotating mechanical transmission devices. When mechanical equipment is running, wear, fatigue, corrosion, overload and other reasons may cause wear of the outer ring of the rolling bearing, rolling element failure, pitting and indentation and other damage. The collected rolling bearing vibration acceleration signal is a typical nonlinear, time-varying coupled non-stationary signal. Its energy characteristics reflecting state information are relatively weak, the extracted feature value is not obvious, and it is difficult to obtain a large number of typical fault samples in reality, which brings difficulties to the fault diagnosis of rolling bearings.

[0006] Secondly, in the fault analysis of rolling bearings, the bearing fault feature information is often submerged in the background noise and difficult to be discovered and extracted. It is necessary to use effective signal processing technology to highlight the fault features.

[0007] Finally, the traditional impact test is used to determine the high-frequency resonance frequency, and then determine the center frequency of the bandpass filter, which requires skilled experimental personnel. In addition, the center frequency and bandwidth of the bandpass filter are designed for a specific bearing system under test, and are no longer applicable to the resonance demodulation system required for other bearing systems, limiting the promotion of resonance demodulation technology in engineering applications. Summary of the invention

[0008] 1. Technical issues to be solved

[0009] In view of the shortcomings of the prior art, the present invention provides a method for diagnosing abnormal vibration signals based on morlet wavelet of unknown frequency band. The method performs adaptive decomposition according to the information of the signal itself, and the generated PF components have real physical meanings. The time-frequency distribution obtained thereby can clearly and accurately reflect the distribution law of the signal energy at various spatial scales. The LMD method can adaptively select the frequency band according to the characteristics of the signal itself, determine the resolution of the signal in different frequency bands, and improve the accuracy of extracting effective information. The method is very suitable for multi-component nonlinear and non-stationary signals, and solves the problems raised in the background technology.

[0010] (II) Technical solution

[0011] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0012] A method for diagnosing abnormal vibration signals based on unknown frequency band morlet wavelet comprises the following steps:

[0013] The method comprises the steps of:

[0014] Select the target rolling bearing and collect the vibration signal under the given fault state as the original signal;

[0015] The original signal is decomposed using the LMD method to obtain k PF components and a monotonic function u k ;

[0016] Calculate the information entropy of the PF component and perform spectrum analysis on the PF component corresponding to each type of fault state;

[0017] Based on the vibration spectrum obtained by spectrum analysis, the high-frequency components are selected through the vibration spectrum, and the corresponding type of fault and bandwidth with the largest characteristic frequency are selected to perform bandpass filtering and envelope spectrum analysis;

[0018] The morlet wavelet is used to construct a bandpass filter to extract the transient components of the frequency band;

[0019] Hilbert transform is used to obtain the envelope spectrum of the selected high-frequency components. By shifting the phase of the test signal by 90°, the signal after the phase shift and the original signal are combined into an analytical signal. Based on the analytical signal, the adaptive resonance demodulation technology is used to obtain the diagnostic result.

[0020] Further, the types of fault conditions identified include outer ring wear, rolling element failure, and inner ring pitting and indentation.

[0021] Furthermore, for any signal x(t), it is decomposed into k PF components and a monotonic function u k The sum is:

[0022]

[0023] In formula (1), PF p (t) is a single component signal, p = 1, 2, ..., k, k is a positive integer, u k (t) is a monotonic function.

[0024] Furthermore, the steps for calculating the information entropy of the PF component are as follows:

[0025] Let c i (t) is a PF component of the vibration signal c(t), c i The maximum value of (t) is assigned to c max , the minimum value is assigned to c min ;

[0026] Set the interval [cmax min ] i ,[c1 min ],…,(A i , A i =1],…,(A N-1 , c max ], these N intervals are used as N discrete value domains B∈{B1,B2,…,B N}, when the value of a sample attribute falls within (A i , A i＝1 ] interval, it is considered that the sample has the corresponding discrete attribute value B on this attribute. I ;

[0027] The total number of points for collecting vibration signals is n, and the number of sample points where c(t) falls in the i-th interval is m i , then calculate the probability P(B i )=m i / n;

[0028] The PF information entropy is:

[0029]

[0030] In formula (2), p i is the probability of the i-th interval; i = 1, 2, 3, ... N, N is a positive integer.

[0031] Furthermore, the standard for selecting high-frequency components through the vibration spectrum is:

[0032] Select the highest frequency or select the resonant frequency corresponding to the largest resonance peak.

[0033] Furthermore, the Morlet wavelet is defined as:

[0034]

[0035] In formula (3), t is time; j is an imaginary unit; e is a natural constant; f b is the bandwidth; f c is the center frequency;

[0036] For real signal analysis, the real part of formula (3) is used as the basis wavelet, then:

[0037]

[0038] The Fourier transform of the morlet wavelet is:

[0039] ψ(f)=exp[-π 2 (ff c ) 2 / f b ] (5)

[0040] Through the convolution characteristics and Fourier transform, the filtering process of the vibration signal of the target rolling bearing to be analyzed by Morlet wavelet in the frequency domain is:

[0041]

[0042] In the formula, x(f) represents the Fourier transform of x(t), and by constructing ψ(f) to achieve bandpass filtering, the high-frequency transient components in the fault signal are obtained;

[0043] According to the construction method of the expanded wavelet function, the expanded wavelet with ψ(t) as the wavelet mother function is ψ s (t), its expression is:

[0044]

[0045] Derived ψ s The frequency domain expression of (t)π s (f) is:

[0046] ψ s (f) = exp[-π 2 (ff c ) 2 / sf b ] (8)

[0047] Where s is a positive real number;

[0048] In fault feature extraction, Shannon entropy is used as the evaluation criterion;

[0049] The coefficients of the wavelet transform are processed into a probability distribution sequence P, and the entropy value calculated is used as the Shannon wavelet entropy; it is defined as:

[0050]

[0051] In the formula, p i is a random probability distribution, processed by wavelet coefficients;

[0052] When the Shannon wavelet entropy is the smallest, the corresponding mother wavelet is the wavelet that best matches the characteristic component. When s is changed within a preset range, the s that minimizes the wavelet entropy is the optimized Morlet wavelet shape parameter.

[0053] Furthermore, the process of using Hilbert transform to obtain the envelope spectrum of the selected high-frequency component is as follows:

[0054] If there is a real signal x(t), perform Hilbert transform on x(t) to get:

[0055]

[0056] The resulting analytical signal is:

[0057] xa(t)=x(t)+jh(t) (11)

[0058] Among them, the amplitude of the analytical signal A(t) is:

[0059]

[0060] The phase φ(t) is:

[0061]

[0062] A(t) is the envelope of the real signal x(t).

[0063] (III) Beneficial effects

[0064] The present invention provides a method for diagnosing abnormal vibration signals based on unknown frequency band morlet wavelet, which has the following beneficial effects:

[0065] (1) This scheme uses a feature extraction method suitable for non-stationary vibration signals to diagnose the vibration signals of rolling bearings, which facilitates the fault diagnosis of rolling bearings;

[0066] (2) This scheme uses resonance demodulation technology in rolling bearing fault diagnosis. As a signal processing method, it greatly improves the signal-to-noise ratio and is particularly suitable for early diagnosis of bearing faults;

[0067] (3) This scheme designs a rolling bearing vibration signal resonance demodulation feature extraction method that combines local mean decomposition (LMD) and information entropy calculation to adaptively select frequency bands, determine the resolution of signals in different frequency bands, and improve the accuracy of extracting effective information;

[0068] (4) In summary, this scheme studies the resonant demodulation feature extraction method of rolling bearing vibration signal based on the morlet wavelet of the unknown frequency band LMD information entropy feature. As a new nonlinear and non-stationary signal analysis method, LMD is adaptively decomposed according to the information of the signal itself. The generated PF component has real physical meaning. The time-frequency distribution obtained can clearly and accurately reflect the distribution law of signal energy at various spatial scales. The LMD method can adaptively select the frequency band according to the characteristics of the signal itself, determine the resolution of the signal in different frequency bands, and improve the accuracy of extracting effective information. It is very suitable for multi-component nonlinear and non-stationary signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 Schematic diagram of various fault types;

[0070] Figure 2 This is a schematic diagram of the original vibration signal of a rolling bearing fault;

[0071] Figure 3 It is a schematic diagram of the LMD decomposition results of the vibration signal;

[0072] Figure 4 The spectrum diagram of PF2 of the outer ring wear vibration signal after LMD decomposition;

[0073] Figure 5 It is a schematic diagram of the Morlet wavelet waveform in time domain and frequency domain;

[0074] Figure 6 is a schematic diagram of a bandpass filtered signal;

[0075] Figure 7 is the spectrum of the bandpass filtered signal;

[0076] Figure 8 is the envelope diagram of the bandpass filtered signal;

[0077] Fig. 9 is a schematic diagram of fault diagnosis results;

[0078] Fig.10 The figure is an operation flow chart of an abnormal vibration signal diagnosis method based on morlet wavelet of unknown frequency band. DETAILED DESCRIPTION

[0079] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] See also Figures 1 to 10 :Provide a method for diagnosing abnormal vibration signals based on morlet wavelet of unknown frequency band;

[0081] The overall steps of this method are as follows:

[0082] S1, vehicle gearbox rolling bearing signal acquisition (i.e. original signal);

[0083] The N218 rolling bearings in the gearbox were tested in three typical conditions: outer ring wear, rolling element failure, and inner ring pitting and indentation (see Figure 1 ), the vibration signals under three states are sampled respectively, and the characteristic frequency of the rolling bearing fault is set as shown in Table 1 below, and the corresponding vibration signal waveform is shown in Figure 2 As shown;

[0084] Table 1 (showing characteristic frequencies of rolling bearing faults):

[0085]

[0086] S2, LMD decomposition;

[0087] LMD adaptively decomposes a complex coupled signal into the sum of several PF components with physically meaningful instantaneous frequencies, where each PF component is obtained by multiplying an envelope signal and a pure FM signal. The instantaneous amplitude of the PF component is represented by the envelope signal, and the instantaneous frequency can be obtained from the pure FM signal. The instantaneous amplitude represents the amplitude modulation information of the PF component, and the instantaneous frequency represents the frequency modulation information of the PF component.

[0088] Therefore, LMD is essentially a demodulation process that decomposes a multi-component signal into a series of single-component signals;

[0089] For any signal x(t), it can be decomposed into k PF components and a monotonic function u k The sum is:

[0090]

[0091] In formula (1), PF p(t) is a single component signal, p = 1, 2, 3, ..., u k (t) is a monotonic function; it should be noted that the signal x(t) represents any original signal;

[0092] The LMD method is used to decompose the outer ring wear vibration signal of the rolling bearing. The results are as follows: Figure 3 As shown;

[0093] S3, calculate the information entropy of PE component;

[0094] According to the information entropy theory, the more obvious the fault characteristics of the PF component are, the more regular the periodic vibration is, and the smaller its information entropy is. The steps to calculate the information entropy of the PF component are as follows:

[0095] S301, let c i (t) is a PF component of the vibration signal c(t), c i The maximum value in (t) is assigned to c max , the minimum value is assigned to c min ;

[0096] S302, set interval [cmax min ] i ,[c1 min ],…,(A i , A i＝1 ],....(A N-1 , c max ] These N intervals are used as N discrete value domains B∈{B1,B2,…,B N}, when the value of a sample attribute falls within (A i , A i =1] interval, it is considered that the sample has the corresponding discrete attribute value B on this attribute. I ;

[0097] S303: The total number of points of vibration signal collected in this embodiment is n, and the number of sample points where c(t) falls in the i-th interval is m i , then calculate the probability P(B i )=m i / n;

[0098] Therefore, the PF information entropy can be obtained as:

[0099]

[0100] In formula (2), p i is the probability of the i-th interval; i = 1, 2, 3, ... N, N is a positive integer;

[0101] The information entropy is calculated for all decomposed PF components, and the results are shown in Table 2:

[0102] Table 2 (Information entropy of each type of PF component):

[0103]

[0104] The information entropy values ​​of the PF components of each state are calculated from Table 2. It can be seen that the information entropy of different PF components is different in the three different states. The PF component with the smallest information entropy value contains the most obvious fault characteristics. Therefore, the corresponding PF component of LMD decomposition of each type of fault is selected for spectrum analysis;

[0105] S4, select the high frequency resonance frequency and determine the center frequency and bandwidth;

[0106] The spectrum of PF2 of the outer ring wear vibration signal after LMD decomposition is as follows: Figure 4 As shown;

[0107] The high-frequency components are selected according to the vibration spectrum of PF2. The definition criteria for selecting high-frequency resonance are: first, high frequencies are selected first; second, the resonance frequency corresponding to the maximum resonance peak is selected as much as possible, so that the signal-to-noise ratio of the envelope signal obtained after envelope detection processing will be higher;

[0108] Figure 4 As can be seen in (a), the resonance frequency corresponding to the second group is higher than that of the first group and the resonance peak is also larger, so the center frequency of the bandpass filter is determined to be 2.6KHz, and the bandwidth should be at least twice the highest fault characteristic frequency of the bearing to ensure that at least the second-order harmonic of the highest fault characteristic frequency can be seen in the resonance demodulation envelope spectrum; among the fault characteristic frequencies of the rolling bearing, the inner ring fault characteristic frequency is the largest, so the bandwidth is selected as 0.5kHz for bandpass filtering and envelope spectrum analysis;

[0109] S5, morlet wavelet bandpass filtering;

[0110] In order to better extract the characteristic information of rolling bearing vibration signal, Morlet wavelet is used to construct a suitable bandpass filter, so as to realize the extraction of transient components of frequency band.

[0111] The morlet wavelet is defined as:

[0112]

[0113] In formula (3), t is time; j is an imaginary unit; e is a natural constant; f b is the bandwidth; f c is the center frequency;

[0114] For real signal analysis, the real part of formula (3) is used as the basis wavelet, then:

[0115]

[0116] The Fourier transform of the morlet wavelet is:

[0117] ψ(f)=exp[-π 2 (ff c ) 2 / f b ] (5)

[0118] Since the morlet wavelet function is a cosine signal with square exponential decay, its waveform is very similar to the vibration signal, so it is more appropriate to use the morlet wavelet as the wavelet basis function. Figure 5 (a) is the time domain waveform of Morlet wavelet, f c The change of directly affects the oscillation frequency of the Morlet wavelet, f b The change of affects the decay time of the morlet wavelet oscillation; Figure 5 (b) is the spectrum of the Morlet wavelet. It can be seen from the figure that changing f c and f b That is, the filtering characteristics can be changed; through the convolution characteristics and Fourier transform, the filtering process of the morlet wavelet on the analyzed rolling bearing vibration signal in the frequency domain can be written as:

[0119]

[0120] In the formula, x(f) represents the Fourier transform of x(t). By constructing a suitable ψ(f), bandpass filtering can be achieved to obtain the high-frequency transient components in the fault signal; s(t) represents the scaled wavelet function in the wavelet transform; specifically, s(t) is usually a wavelet function that is scaled and widened based on the mother wavelet.

[0121] According to the construction method of the expanded wavelet function, the expanded wavelet with ψ(t) as the wavelet mother function is ψ s (t), its expression is:

[0122]

[0123] It can be deduced that ψ s The frequency domain expression of (t)ψ s (f) is:

[0124] ψ s (f) = exp[-π 2 (ff c ) 2 / sf b ] (8)

[0125] In the formula, s is a positive real number, which is used to determine the width of the wavelet in the time domain and frequency domain and to adapt to the transformation of the attenuation frequency;

[0126] Outside a very small interval, the function value is zero, which ensures the rapid drop characteristics of the function and can obtain time domain localization. When s increases, the wavelet function is compressed; the corresponding Fourier transform increases, the center of the frequency domain window increases, and vice versa; the spectrum of the wavelet function has a bandpass feature, and the waveform of the Morlet wavelet function is very similar to the waveform of the attenuated oscillation of the rolling bearing vibration signal. The wavelet transform based on this wavelet function can well extract the fault characteristics in the rolling bearing vibration signal;

[0127] In fault feature extraction, it is often hoped that the characteristic components can be highlighted while the irrelevant components can be suppressed. This requires the selection of three s that can make the Morlet wavelet and the characteristic components have the greatest similarity. Among them, Shannon entropy is a good evaluation criterion. The coefficients of the wavelet transform are processed into a probability distribution sequence P. The entropy value calculated from it reflects the sparsity of the wavelet coefficient matrix. This entropy is called Shannon wavelet entropy. It is defined as:

[0128]

[0129] Where: p i is an uncertain probability distribution, processed by wavelet coefficients;

[0130] When the Shannon wavelet entropy is the smallest, the corresponding mother wavelet is the wavelet that best matches the characteristic component. According to this principle, by changing s within a certain range, the s that minimizes the wavelet entropy is the optimized Morlet wavelet shape parameter.

[0131] The corresponding center frequency and bandwidth are selected for rolling element failure and inner ring failure in the same way;

[0132] The corresponding bandpass filtered signal is Figure 6 As shown, the spectrum diagram corresponding to the filtered signal is as follows Figure 7 As shown in (a), (b) and (c);

[0133] S6, envelope demodulation and spectrum analysis;

[0134] The envelope spectrum of the selected high-frequency natural vibration signal is obtained by using Hilbert transform. By shifting the phase of the test signal by 90°, the signal after the phase shift and the original signal are combined into an analytical signal.

[0135] If there is a real signal x(t), perform Hilbert transform on x(t) to get:

[0136]

[0137] Then the analytical signal can be obtained as:

[0138] x a (t) = x(t) + jh(t) (11)

[0139] The amplitude of the analytical signal A(t) is:

[0140]

[0141] The phase φ(t) is:

[0142]

[0143] A(t) is the envelope of the real signal x(t);

[0144] The envelope diagram of the bandpass filtered signal is as follows: Figure 8 As shown in (a), (b) and (c);

[0145] from Figure 8 (a), (b), and (c) show that the demodulated signals of the three rolling bearing fault signals after Morlet wavelet filtering also have the largest amplitude of the rolling bearing rolling element fault filtering demodulated signal. The fault demodulation diagnosis results are as follows: Fig. 9 As shown;

[0146] from Fig. 9 It can be seen from (a) that the spectrum peak is 64Hz, which is close to the characteristic frequency of outer ring wear 62.63Hz. It can be considered here as the characteristic frequency of inner ring fault. In actual analysis, due to factors such as non-pure rolling of rolling elements and bearing installation errors, the characteristic frequency of fault is not always exactly equal to the theoretical calculated value. Therefore, when looking for each characteristic frequency on the spectrum graph, it is necessary to find its approximate value above and below the calculated frequency to make a diagnosis. There is a very obvious frequency doubling feature (125Hz≈2fi, 189≈3fi) in the envelope spectrum.

[0147] This result shows that the adaptive resonance demodulation method can diagnose the outer ring wear fault of the rolling bearing very well; Fig. 9 In (b), it can be clearly seen that the spectrum peak is 28Hz, which is slightly different from the characteristic frequency of rolling element failure, 28.71Hz, and has a very obvious frequency doubling in the envelope spectrum; Fig. 9 (c) The mid-spectrum peak of 63 Hz is close to the characteristic frequency of pitting and indentation of the inner ring of 62.63 Hz, and there is a very obvious frequency doubling in the envelope spectrum;

[0148] In summary, the adaptive resonance demodulation technology based on LMD information entropy morlet wavelet proposed in the present invention is applied to the analysis of outer ring wear, rolling element failure, and inner ring pitting and indentation. The results obtained are consistent with the laws of their respective characteristic failures, and the failure characteristics are relatively obvious. The failure status of each working condition can be diagnosed. The example verifies the effectiveness and feasibility of the method proposed in the present invention.

[0149] It should be noted that:

[0150] The present invention realizes adaptive resonance demodulation diagnosis of rolling bearing fault vibration signals based on Morlet wavelet and LMD information entropy characteristics of unknown frequency band. Compared with the traditional resonance demodulation method, the present invention has the following advantages:

[0151] Strong adaptability: The present invention uses LMD decomposition technology to perform adaptive frequency band selection, and can determine the resolution according to the characteristics of the signal itself, overcoming the problem of fixed frequency setting and reliance on professional experience in traditional methods, making the diagnostic system more universal and applicable to a variety of rolling bearing systems;

[0152] The signal-to-noise ratio is significantly improved: The Morlet wavelet is used to construct a bandpass filter, which significantly improves the signal-to-noise ratio in the process of extracting high-frequency transient components, making the fault characteristics more prominent in complex background noise, especially suitable for accurate identification of early faults;

[0153] Efficient and accurate feature extraction: The optimal PF component of fault characteristics is identified through information entropy calculation, making the diagnosis process more targeted and significantly improving the accuracy of fault information extraction, especially in the diagnosis of different fault types such as outer ring wear, rolling element failure and inner ring pitting.

[0154] Feasibility of engineering application: This method can determine the bandpass filter parameters without relying on impact tests, which greatly simplifies the operation process and facilitates the promotion of engineering applications;

[0155] In summary, the present invention effectively makes up for the shortcomings of traditional methods, can quickly and accurately diagnose the fault status of rolling bearings, and has broad application prospects and significant practical value in industrial sites.

[0156] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0157] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0158] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for diagnosing abnormal vibration signals based on unknown frequency band morlet wavelet, characterized in that: The method comprises the steps of: Select the target rolling bearing and collect the vibration signal under the given fault state as the original signal; The original signal is decomposed using the LMD method to obtain k PF components and a monotonic function u k ; Calculate the information entropy of the PF component and perform spectrum analysis on the PF component corresponding to each type of fault state; Based on the vibration spectrum obtained by spectrum analysis, the high-frequency components are selected through the vibration spectrum, and the corresponding type of fault and bandwidth with the largest characteristic frequency are selected to perform bandpass filtering and envelope spectrum analysis; The morlet wavelet is used to construct a bandpass filter to extract the transient components of the frequency band; Hilbert transform is used to obtain the envelope spectrum of the selected high-frequency components. By shifting the phase of the test signal by 90°, the signal after the phase shift and the original signal are combined into an analytical signal. Based on the analytical signal, the adaptive resonance demodulation technology is used to obtain the diagnostic result.

2. The abnormal vibration signal diagnosis method based on unknown frequency band morlet wavelet according to claim 1 is characterized in that: The types of failure conditions identified include outer race wear, rolling element failure, and inner race pitting and scoring.

3. The abnormal vibration signal diagnosis method based on unknown frequency band morlet wavelet according to claim 1 is characterized in that: For any signal x(t), it is decomposed into k PF components and a monotonic function u k The sum is: In formula (1), PF p (t) is a single component signal, p = 1, 2, ..., k, k is a positive integer, u k (t) is a monotonic function.

4. The abnormal vibration signal diagnosis method based on unknown frequency band morlet wavelet according to claim 1 is characterized in that: The steps to calculate the information entropy of the PF component are as follows: Let c i (t) is a PF component of the vibration signal c(t), c i The maximum value of (t) is assigned to c max , the minimum value is assigned to c min ; Set the interval [cmax min ] i ,[c1 min ],…,(A i , A i＝1 ],…,(A N-1 , c max ], these N intervals are used as N discrete value domains B∈{B1,B2,…,B N }, when the value of a sample attribute falls within (A i , A i＝1 ] interval, it is considered that the sample has the corresponding discrete attribute value B on this attribute. I ; The total number of points for collecting vibration signals is n, and the number of sample points where c(t) falls in the i-th interval is m i , then calculate the probability P(B i )=m i / n; The PF information entropy is: In formula (2), p i is the probability of the i-th interval; i = 1, 2, 3, ... N, N is a positive integer.

5. The abnormal vibration signal diagnosis method based on unknown frequency band morlet wavelet according to claim 1 is characterized in that: The criteria for selecting high-frequency components through the vibration spectrum are: Select the highest frequency or select the resonant frequency corresponding to the largest resonance peak.

6. The abnormal vibration signal diagnosis method based on unknown frequency band morlet wavelet according to claim 5 is characterized by: The morlet wavelet is defined as: In formula (3), t is time; j is an imaginary unit; e is a natural constant; f b is the bandwidth; f c is the center frequency; For real signal analysis, the real part of formula (3) is used as the basis wavelet, then: The Fourier transform of the morlet wavelet is: ψ(f)=exp[-π 2 (ff c ) 2 / f b ] (5) Through the convolution characteristics and Fourier transform, the filtering process of the vibration signal of the target rolling bearing to be analyzed by Morlet wavelet in the frequency domain is: In the formula, x(f) represents the Fourier transform of x(t), and by constructing ψ(f) to realize bandpass filtering, the high-frequency transient components in the fault signal are obtained; s(t) represents the expanded wavelet function in the wavelet transform; According to the construction method of the expanded wavelet function, the expanded wavelet with ψ(t) as the wavelet mother function is ψ s (t), its expression is: Derived ψ s The frequency domain expression of (t)ψ s (f) is: ψ s (f)=exp[-π 2 (f-f c ) 2 / sf b ] (8) Where s is a positive real number; In fault feature extraction, Shannon entropy is used as the evaluation criterion; The coefficients of the wavelet transform are processed into a probability distribution sequence P, and the entropy value calculated is used as the Shannon wavelet entropy; it is defined as: In the formula, p i is a random probability distribution, processed by wavelet coefficients; When the Shannon wavelet entropy is the smallest, the corresponding mother wavelet is the wavelet that best matches the characteristic component. When s is changed within a preset range, the s that minimizes the wavelet entropy is the optimized Morlet wavelet shape parameter.

7. The abnormal vibration signal diagnosis method based on unknown frequency band morlet wavelet according to claim 1 is characterized by: The process of using Hilbert transform to obtain the envelope spectrum of the selected high-frequency component is as follows: If there is a real signal x(t), perform Hilbert transform on x(t) to get: The resulting analytical signal is: x a (t)=x(t)+jh(t) (11) Among them, the amplitude of the analytical signal A(t) is: The phase φ(t) is: A(t) is the envelope of the real signal x(t).

Images