Bearing fault identification method based on time sequence parameter statistics

Abstract

The invention discloses a bearing fault identification method based on time sequence parameter statistics, and relates to the field of rotating equipment state monitoring and intelligent operation and maintenance, and the method comprises the steps: S1, collecting a vibration signal sample of a target bearing; s2, short-time Fourier transform: performing short-time Fourier transform on the vibration signal to obtain observation of the model; s3, calculating a forward probability and a backward probability; s4, counting times: calculating a bearing vibration signal of the duration ts of each impact cycle by using a dynamic time sequence # imgabs0 #, including a noise state and an impact state, namely # imgabs1 #, and counting the times Nc of the signal falling in a narrow frequency interval # imgabs4 # with the # imgabs2 # as the center and the width of # imgabs3 # within the duration ts; s5, calculating the posterior probability of the fault; and S6, automatic fault identification: automatically identifying the fault type of the bearing by selecting the highest fault posterior probability. And estimating the signal impact frequency through parameters in the dominant time hidden Markov model so as to perform fault identification on the bearing.

Description

Technical Field

[0001] The present invention relates to the field of rotating equipment condition monitoring and intelligent operation and maintenance, and in particular to a bearing fault identification method based on time series parameter statistics. Background Art

[0002] As core components of rotating systems, bearings play a vital role. They not only support rotating mechanical parts but also significantly reduce friction and wear during operation, thereby extending the life of the equipment. Bearing performance is directly related to the reliability and efficiency of rotating systems. Bearing failure can lead to equipment downtime, production interruptions, and significant increases in maintenance costs. Therefore, strict control of bearing quality and real-time monitoring of bearing condition are crucial to ensuring efficient and stable equipment operation.

[0003] Bearing fault pattern recognition based on vibration signals has become a research hotspot in academia and engineering practice in recent years. With the rapid development of science and technology, bearing diagnosis and intelligent operation and maintenance technologies have continued to advance, gradually forming a multi-dimensional, multi-domain research framework. Key research areas include the following: 1) Signal decomposition: Using techniques such as empirical mode decomposition (EMD), local mean decomposition (LMD), blind deconvolution, and matrix factorization, physically meaningful characteristic components are extracted from complex signals. 2) Signal modeling: Using methods such as cyclostationary analysis and hidden Markov models (HMMs), statistical characteristic models of the signal are established to capture the dynamic characteristics of fault evolution. 3) Signal transformation: Using transformation methods such as short-time Fourier transform (STFT), Wigner-Ville distribution, wavelet transform, and cepstrum analysis, signals are converted from the time domain to the time-frequency domain or feature domain to reveal the underlying mechanisms of bearing faults. 4) Artificial intelligence: Combining intelligent algorithms such as deep learning, support vector machines (SVMs), and neural networks, automatic fault pattern recognition and prediction are achieved. While the above methods significantly improve the accuracy and efficiency of vibration signal processing, their core remains revolving around demodulation and spectrum analysis of vibration signals. By carefully observing the spectrum, the frequency characteristics of bearing faults can be identified, allowing accurate determination of the fault type. This process requires a high level of professional knowledge and technical experience, especially in complex operating conditions or when multiple faults overlap, posing even greater challenges to both the algorithm and the engineer's capabilities.

[0004] In view of the above situation, the present invention proposes a bearing fault identification method based on time series parameter statistics. Summary of the Invention

[0005] The present invention provides a bearing fault identification method based on time series parameter statistics, which overcomes the shortcomings of existing bearing fault identification methods and provides a method for automatically identifying bearing faults using the time interval of the time domain waveform of a vibration signal. The signal impact frequency is estimated by parameters in an explicit time hidden Markov model, thereby identifying bearing faults.

[0006] According to one aspect of the present disclosure, a bearing fault identification method based on time series parameter statistics is provided, the method comprising: S1: Collect vibration signal samples of the target bearing; S2: Short-time Fourier transform: Perform short-time Fourier transform on the vibration signal to obtain the observation of the model; S3: Calculate forward probability and backward probability: convert the sequence As an observation sequence of the explicit time hidden Markov model EDHMM random model, based on the sequence , ..., Calculate forward probability and backward probability and dynamic time series ; S4: Frequency Statistics: Using Dynamic Time Series , calculate the duration of each cycle in the bearing vibration signal t s , t s It contains the duration of a noise state and the duration of an impact state, that is, , statistics duration t s Falling inside Centered at Narrow frequency range The number of signals in N c ,as follows, ; Among them, card represents the number of elements in the set, σ For frequency error, the characteristic frequency is expanded to a characteristic interval; S5: Calculate the posterior probability of failure , calculated as follows: ; C represents the maximum value of the index, and c represents different failure modes; S6: Automatic fault identification: Automatically identify the bearing fault type by selecting the one with the highest posterior probability of fault.

[0007] In a possible implementation, the vibration signal sample y(t), t = 0, 1, 2, ...., L-1, where t is the time index of the signal sample and the signal length is L.

[0008] In a possible implementation, in step S2, the time domain signal of the vibration signal sample is y ( t ), whose short-time Fourier transform coefficients Y ( n , k ) is written as follows: ; Among them, the window w ( m ) is the length of N w ; Y ( n , k )'s time index n =1, 2, 3,…, N ; Frequency index k =1, 2,3,…, N f ; then the Fourier transform coefficients Y ( n , k ) is considered as N indivual N f dimensional observation sequence; According to the central limit theorem, the STFT coefficient The distribution converges to the complex-valued Gaussian distribution. After decentralization, it is further simplified to a circularly symmetric Gaussian distribution, as shown below: ;in, represents a diagonal covariance matrix whose diagonal elements are expressed as follows: Indicates the time index n and frequency index k The expected value of the signal energy at represents the conjugate transpose.

[0009] In one possible implementation, in step S3, the hidden state sequence variable in the random model z n , n =1,2,3,…, N , there are two states in the bearing vibration signal ; The noise state i =1, indicating that only noise exists; transient impact state i =2, indicating that there is an impact phenomenon; Sequence-based , ..., Calculate forward probability and backward probability ,as follows, ; ; Among them, dynamic time series The update rule is calculated as follows: .

[0010] In a possible implementation, the bearing fault types include: outer ring fault, inner ring fault, rolling element fault, and cage fault.

[0011] Compared with the prior art, the present invention has the following beneficial effects: One embodiment of this disclosure presents a bearing fault identification method based on time series parameter statistics, aiming to achieve automated bearing fault identification. By combining the bearing impact time and impact interval, a statistical method is proposed to estimate the signal impact frequency. This method automatically identifies the bearing fault type without relying on professional spectrum observation, significantly lowering the diagnostic technical threshold and improving identification efficiency and practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flow chart of a bearing fault identification method based on time series parameter statistics according to an embodiment of the present disclosure is shown; Figure 2 A schematic diagram of vibration signal modeling according to an embodiment of the present disclosure is shown; Figure 3 A sample of a vibration signal of a data set according to an embodiment of the present disclosure is shown; Figure 4 shows the square envelope spectrum of the vibration signal according to an embodiment of the present disclosure; Figure 5 A histogram showing the transient pulse period frequency of a vibration signal according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0013] Various exemplary embodiments, features, and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0014] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0015] In addition, numerous specific details are provided in the following detailed description to better illustrate the present disclosure. Those skilled in the art will appreciate that the present disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art are not described in detail in order to highlight the main points of the present disclosure.

[0016] Figure 1 A flow chart of a bearing fault identification method based on time series parameter statistics according to an embodiment of the present disclosure is shown; Figure 2 A schematic diagram of vibration signal modeling according to an embodiment of the present disclosure is shown, which respectively shows the vibration signal, the posterior probability, the noise state signal τ(1) and the impact state signal τ(2); Figure 3 A sample of a vibration signal of a data set according to an embodiment of the present disclosure is shown; Figure 4 shows the square envelope spectrum of the vibration signal according to an embodiment of the present disclosure; Figure 5 A histogram showing the transient pulse period frequency of a vibration signal according to an embodiment of the present disclosure.

[0017] The collected signal (such as Figure 3 The short-time Fourier transform (STFT) is used for illustration in this embodiment, and other time-frequency transforms are also applicable.

[0018] For the time domain signal of the vibration signal sample y ( t ), whose short-time Fourier transform coefficients Y ( n , k ) is written as follows: Among them, the window w ( m ) is the length of N w ; Y ( n , k )'s time index n =1,2, 3,…, N ; Frequency index k =1, 2, 3,…, N f ; then the Fourier transform coefficients Y ( n , k ) is considered as N indivual N f dimensional observation sequence; According to the central limit theorem, the STFT coefficient The distribution converges to the complex-valued Gaussian distribution. After decentralization, it is further simplified to a circularly symmetric Gaussian distribution, as shown below: in, represents a diagonal covariance matrix whose diagonal elements are expressed as follows: Indicates the time index n and frequency index k The expected value of the signal energy at represents the conjugate transpose.

[0019] For example, the STFT window width in the method is The window shift R needs to be specifically selected according to the vibration signal speed sampling frequency. In this embodiment , R=8. After short-time Fourier transform, the coefficients are obtained , . Used in subsequent random models.

[0020] By iterating the expected maximum method, the impact time in the signal can be obtained and impact interval sequence, and calculate the adjacent and The sum of the shock cycles t s , s =1,2,3,..., S . S represents the number of cycles.

[0021] For example, the bearing model of the signal is LDK UER204, the bearing rotation frequency is 2400RPM (revolutions per minute), and its characteristic frequency is The calculations show that the characteristic frequency of the outer ring is 123.32Hz; the characteristic frequency of the inner ring is 196.67Hz; the characteristic frequency of the rolling element is 82.66Hz; and the characteristic frequency of the cage is 15.41Hz.

[0022] Count all cycles t s The center of the bearing characteristic frequency is Narrow frequency range The number of times N c , the error σ Set it to 5% of the fault characteristic frequency. Taking the outer ring as an example, its frequency range is [117.15 129.48] Hz.

[0023] The statistical probabilities of the four intervals are shown in the following table. It can be found that the posterior probability of the outer ring fault is The highest, so the fault mode can be given by data analysis, and the bearing fault type can be automatically identified without spectrum observation.

[0024] Table 1. Posterior probabilities of four types of faults ; The correctness of this conclusion can be verified from two perspectives. First, Figure 3 The outer characteristic frequency can be found by making a square envelope spectrum of the vibration signal shown, such as Figure 4 As shown. t s Make a histogram of the distribution statistics Figure 5 As shown, it can be found that the characteristic frequency of the bearing outer ring (as shown by the red dotted line) falls into the highest group, which indicates that the impact in the signal is caused by the outer ring fault to a certain extent, proving the correctness of the analysis results.

[0025] Through the verification of actual cases, it can be clearly proved that the invention proposes to use the signal waveform to shock the cycle t s This statistical analysis method can automatically and accurately identify bearing failure modes. Compared to traditional methods that rely on manual spectrum observation, this method offers significant advantages in ease of use and accuracy. This invention reduces reliance on operator expertise and is suitable for large-scale industrial applications.

[0026] While various embodiments of the present disclosure have been described above, the foregoing description is intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technological improvements in the marketplace, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A bearing fault identification method based on time series parameter statistics, characterized in that: The method comprises: S1: Collect vibration signal samples of the target bearing; S2: Short-time Fourier transform: Perform short-time Fourier transform on the vibration signal to obtain the observation of the model; S3: Calculate forward probability and backward probability: convert the sequence As an observation sequence of the explicit time hidden Markov model EDHMM random model, based on the sequence , ..., Calculate forward probability and backward probability and dynamic time series ; S4: Frequency Statistics: Using Dynamic Time Series , calculate the duration of each cycle in the bearing vibration signal t s , t s It contains the duration of a noise state and the duration of an impact state, that is, , statistics duration t s Falling inside Centered at Narrow frequency range The number of signals in N c ,as follows, ； Among them, card represents the number of elements in the set, σ For frequency error, the characteristic frequency is expanded to a characteristic interval; S5: Calculate the posterior probability of failure , calculated as follows: ; C represents the maximum value of the index, and c represents different failure modes; S6: Automatic fault identification: Automatically identify the bearing fault type by selecting the one with the highest posterior probability of fault.

2. The method for identifying bearing faults based on time series parameter statistics according to claim 1, characterized in that: Vibration signal samples y(t), t = 0, 1, 2, ...., L-1, where t is the time index of the signal sample and the signal length is L.

3. The method for identifying bearing faults based on time series parameter statistics according to claim 1, characterized in that: In step S2, the time domain signal of the vibration signal sample y ( t ), whose short-time Fourier transform coefficients Y ( n , k ) is written as follows: ； Among them, the window w ( m ) is the length of N w ; Y ( n , k )'s time index n =1, 2, 3,…, N ; Frequency index k =1, 2,3,…, N f ; then the Fourier transform coefficients Y ( n , k ) is considered as N indivual N f dimensional observation sequence; According to the central limit theorem, the STFT coefficient The distribution converges to the complex-valued Gaussian distribution. After decentralization, it is further simplified to a circularly symmetric Gaussian distribution, as shown below: ;in, represents a diagonal covariance matrix whose diagonal elements are expressed as follows: Indicates the time index n and frequency index k The expected value of the signal energy at represents the conjugate transpose.

4. The method for identifying bearing faults based on time series parameter statistics according to claim 3, characterized in that: In step S3, the hidden state sequence variable in the random model z n , n =1,2,3,…, N , there are two states in the bearing vibration signal ; The noise state i =1, indicating that only noise exists; transient impact state i =2, indicating that there is an impact phenomenon; Sequence-based , ..., Calculate forward probability and backward probability ,as follows, ； ； Among them, dynamic time series The update rule is calculated as follows: 。 5. The method for identifying bearing faults based on time series parameter statistics according to claim 1, characterized in that: Bearing failure types include: outer ring failure, inner ring failure, rolling element failure, and cage failure.

Images