Variable working condition vibration signal stationary section interception method based on time frequency

Abstract

The invention discloses a time-frequency-based variable working condition vibration signal stationary segment interception method, which belongs to the field of intelligent operation and maintenance of rotating equipment, and comprises the following steps: S1, carrying out the preprocessing of a vibration signal through the removal of a DC component and the removal of an envelope; s2, performing short-time Fourier transform on the preprocessed vibration signal to obtain a time-frequency spectrum of the vibration signal; s3, dividing the time-frequency spectrum into 20 segments through a time sequence, and calculating an accumulated value of each frequency component in the whole time window based on the time-frequency spectrum, namely spectrum energy of the frequency; s4, taking the spectral energy of the initial time period as a reference, and calculating the correlation between the spectral energy of different time periods and the spectral energy of the reference time period by using a Pearson Correlation Coefficient (Pearson Correlation Coefficient); and S5, by setting a steady section threshold value, intercepting a time period of which the correlation is higher than the threshold value as a steady section. According to the invention, based on the correlation change of the vibration signal spectrum energy of a certain crown block under variable working conditions, the stable segment interception of the vibration signal is realized, and the fault diagnosis precision is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent operation and maintenance of rotating equipment, and in particular to a method for intercepting a stable segment of a variable working condition vibration signal based on time-frequency. Background Art

[0002] During mechanical operation, signals often exhibit complex nonstationary characteristics due to load variations, speed fluctuations, and environmental influences. Under variable operating conditions, signal nonstationarity manifests primarily as transient characteristics, frequency drift, and amplitude variations. These nonstationary segments can adversely affect fault detection. For example, strong noise and sudden operating condition changes within nonstationary segments can mask subtle bearing fault signatures, making feature extraction difficult. Among methods for extracting stationary vibration signals under variable operating conditions, the wavelet transform enhances local feature extraction through multi-scale decomposition. However, the selection of basis functions relies on empirical experience, resulting in high computational complexity. Furthermore, the wavelet coefficients of stationary vibration signals are easily overwhelmed by noise in strong noise backgrounds. While empirical mode decomposition can adaptively decompose nonlinear signals, modal aliasing and endpoint effects can introduce spurious components when operating condition changes suddenly, causing stationary frequencies (stable rotational frequencies) to be confused with noise components. While the Wigner-Ville distribution provides high-resolution time-frequency representation, cross-term interference and negative energy issues can produce spurious features in multi-component signals. Therefore, effectively extracting the stable segment of the vibration signal under variable working conditions is of great significance to improving the accuracy of fault diagnosis.

[0003] The present invention uses envelope analysis, time-frequency analysis, and spectral energy correlation analysis on collected vibration signals to determine the spectral energy correlation trends at different sampling times. Based on these correlation trends, a fixed threshold is set to effectively extract stable vibration data under varying operating conditions. Summary of the Invention

[0004] This invention aims to overcome the shortcomings of existing technologies by proposing a time-frequency-based method for extracting stationary segments from variable-condition vibration signals. This method analyzes the time-frequency characteristics of the vibration signal using time-frequency analysis and uses spectral energy similarity to determine stationary segments. This method is simple and reliable, requires no human intervention, and achieves high segmentation accuracy.

[0005] To achieve the above object, the present invention provides a method for extracting the stationary segment of a vibration signal under variable working conditions based on time-frequency, comprising the following steps:

[0006] S1. Collect vibration signals of target rotating equipment through vibration sensors ,in is the number of sampling points of vibration acceleration, is the sampling time length;

[0007] S2. Read a set of vibration signals The processed vibration signal is obtained by removing the DC component and envelope .

[0008] (1) Remove DC component: Subtract the average value of the signal to make the mean value of the signal zero. The signal after removing the DC component is:

[0009]

[0010] (2) De-envelope: Calculation The RMS envelope is passed through Remove the envelope of the signal by point division, and we get The specific steps are as follows:

[0011] Set the envelope window size , according to the sampling frequency Determine the window length:

[0012]

[0013] in, is the target frequency.

[0014] Compute the RMS envelope of a signal:

[0015]

[0016] Divide by the signal envelope , and obtain the dynamic normalized signal :

[0017]

[0018] S3, The time-frequency matrix is obtained by short-time Fourier transform (STFT). The mathematical expression of STFT is:

[0019]

[0020] in, represents the vibration signal after de-envelope, is the window function, is a negative exponential term, For time, is the frequency.

[0021] Time-frequency spectrum calculated by STFT Contains positive and negative frequency parts, extracts the positive frequency part Time spectrum TFR:

[0022] TFR= v STFT [f>0]

[0023] S4. Time window segmentation and TFR calculation. Divide the TFR sliding window into 20 segments based on the time sequence without overlapping, and calculate the cumulative amplitude of TFR in each time period within the entire time window. The cumulative value is the spectral energy of the frequency. .

[0024]

[0025] Among them, TFR mn Represents the complex amplitude of the spectral energy of the mth time period at the nth time point.

[0026] S5. Calculate the TFR correlation of the sub-windows. The spectrum energy of the vibration signal in the initial time period is used. As a benchmark, the Pearson Correlation Coefficient is used to calculate the correlation between the spectral energy of different time periods and the baseline time period. A higher correlation value indicates that the time-frequency characteristics of the signal segment are similar to those of the baseline segment signal, indicating that the signal is relatively stable. Conversely, a lower correlation value indicates that the signal may have large non-stationary changes. The correlation calculation formula is as follows:

[0027]

[0028] in, and Represent the spectral energy values of the i-th segment TFR and the benchmark TFR, and are the mean of the spectrum energy of the current time-frequency segment and the reference time-frequency segment, respectively, and N is the number of parts.

[0029] S6, vibration stable segment interception. Set a stable segment threshold When the spectral energy correlation corr is higher than the threshold, the vibration signal is judged to be a stationary segment, and the data of the stationary segment is retained for subsequent fault diagnosis and health status assessment operations; when the spectral energy correlation corr is lower than the threshold, the vibration signal is judged to be a non-stationary segment, and the signal of this segment will be intercepted and eliminated. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flow chart of the overall method

[0031] Figure 2 This is the original vibration signal diagram of a certain overhead crane when its traveling mechanism decelerates.

[0032] Figure 3 This is the STFT time-frequency diagram of a certain crane when its traveling mechanism is decelerating.

[0033] Figure 4 This is the energy correlation trend diagram of the deceleration spectrum of a certain overhead crane travel mechanism

[0034] Figure 5 It is the original vibration signal of the stable section intercepted when the overhead crane travel mechanism decelerates.

[0035] Figure 6 It is the original vibration signal of deceleration and acceleration of a certain crane travel mechanism

[0036] Figure 7 This is the STFT time-frequency diagram of a certain crane's travel mechanism during deceleration and acceleration

[0037] Figure 8 This is the energy correlation trend diagram of the deceleration-acceleration spectrum of a certain crane travel mechanism

[0038] Figure 9 It is the original vibration signal of the stable section captured when the overhead traveling mechanism decelerates and accelerates. DETAILED DESCRIPTION

[0039] S1: Collect vibration signals of target rotating equipment through vibration sensors ,in Indicates the number of sampling points of vibration acceleration, is the time sampling time;

[0040] In this embodiment, the target machine is a rotating device of an overhead crane. When the rotation speed of the target machine reaches a preset threshold, the vibration sensor starts collecting data. Under the preset rotation speed threshold, the collected vibration signal can be considered a stationary signal. Therefore, the vibration signal collected at the initial moment is determined to be a stationary vibration signal.

[0041] S2: Read a set of vibration signals And remove the DC component and signal envelope to obtain the processed vibration signal .

[0042] (1) Remove the DC component. Subtract the average value of the signal from the vibration signal to make the mean value of the signal zero. The signal after removing the DC component is:

[0043]

[0044] (2) De-envelopment. Perform envelope extraction and remove the envelope to obtain the signal after removing the envelope The specific steps are:

[0045] Set the envelope window size , according to the sampling frequency Determine the window length:

[0046]

[0047] in, is the target frequency.

[0048] Compute the RMS envelope of a signal:

[0049]

[0050] Will Divide by the envelope signal to get the dynamic normalized signal :

[0051]

[0052] In this embodiment, the DC component is removed from a selected set of vibration signals, wherein the sampling frequency of the vibration signal is , set the envelope window size , the envelope of the signal after removing the DC component is extracted and removed by the root mean square algorithm to obtain the preprocessed vibration signal.

[0053] S3: Vibration acceleration signal after de-envelopment The time-frequency spectrum is obtained by short-time Fourier transform The mathematical expression of STFT is:

[0054]

[0055] in, represents the vibration signal after de-envelope, is the window function, is a negative exponential term, t is time, and f is frequency.

[0056] Time-frequency spectrum calculated by STFT Contains positive and negative frequency parts. Extract the positive frequency part Time spectrum TFR:

[0057] TFR= v STFT [f>0]

[0058] In this embodiment, set the STFT parameters: select Kaiser window The length is , window shape parameters ; Segment length , overlap length , FFT points .

[0059] S4: Divide the TFR sliding window into 20 segments without overlapping through the time series, and calculate the cumulative amplitude of the spectrum energy of each time period in the entire time window based on TFR. The cumulative value is the spectrum energy of the frequency .

[0060]

[0061] Among them, TFR mn Represents the complex amplitude of the spectral energy of the mth time period at the nth time point.

[0062] In this embodiment, the time-frequency spectrum obtained by STFT is divided into 20 segments according to time periods, and the spectrum energy is calculated for each time period to obtain a spectrum energy sequence.

[0063] S5. Spectral energy of the vibration signal in the initial time period As a benchmark, the Pearson correlation coefficient is used to calculate the correlation between the spectral energy of different time periods and the reference time period. A higher correlation value indicates that the time-frequency characteristics of the signal segment are similar to those of the reference segment signal, indicating that the signal is relatively stable; conversely, a lower correlation value indicates that the signal may have large non-stationary changes. The correlation calculation formula is as follows:

[0064]

[0065] in, and Represent the spectral energy values of the i-th segment TFR and the benchmark TFR, and are the mean of the spectrum energy of the current time-frequency segment and the reference time-frequency segment, respectively, and N is the number of parts.

[0066] In this embodiment, the spectral energy of the vibration stability period is selected as the benchmark, and the correlation between the spectral energy of each period and the spectral energy of the initial period is calculated. This method can be used to evaluate the time-frequency stability of the signal in different time periods, and then used for signal stationarity analysis.

[0067] S6: Set a plateau threshold When the spectral energy correlation corr is higher than the threshold, the vibration signal is judged to be a stationary segment, and the data of the stationary segment is retained for subsequent fault diagnosis and health status assessment operations; when the spectral energy correlation corr is lower than the threshold, the vibration signal is judged to be a non-stationary segment, and the signal of this segment will be intercepted and eliminated.

[0068] In this example, select , to ensure that only stationary signal segments with characteristics similar to the baseline segment are retained for subsequent analysis.

Claims

1. A method for extracting the stationary segment of a variable working condition vibration signal based on time-frequency, characterized in that: The following steps are involved: S1. Collect vibration signals of target rotating equipment through vibration sensors ,in is the number of sampling points of vibration acceleration, is the time sampling length; S2. Read a set of vibration signals And remove the DC component and envelope to get the processed vibration signal . (1) Remove the DC component. Subtract the average value of the signal from the vibration signal to make the mean value of the signal zero. The signal after removing the DC component is: (2) De-envelope. By calculating the root mean square (RMS) Perform envelope extraction and remove the envelope to obtain The specific steps are: Set the envelope window size , according to the sampling frequency Determine the window length: in, is the target frequency. Compute the RMS envelope of a signal: Will Divide by the envelope signal to get the dynamic normalized signal : S3, vibration acceleration signal after de-envelopment The time-frequency spectrum is obtained by short-time Fourier transform (STFT). The mathematical expression of STFT is: in, represents the vibration signal after de-envelope, is the window function, is a negative exponential term, For time, is the frequency. Time-frequency spectrum calculated by STFT Contains positive and negative frequency parts. Extract the positive frequency part Time spectrum TFR: S4, through time series The sliding window is divided into 20 segments without overlapping, and the cumulative amplitude of the spectrum energy of each time period in the entire time window is calculated based on TFR. The cumulative value is the spectrum energy of the frequency . in, Represents the complex amplitude of the spectral energy of the mth time period at the nth time point. S5. Spectral energy of the vibration signal in the initial time period As a benchmark, the Pearson Correlation Coefficient is used to calculate the correlation between the spectral energy of different time periods and the reference time period. A higher correlation value indicates that the time-frequency characteristics of the signal segment are similar to those of the reference segment signal, indicating that the signal is relatively stable; conversely, a lower correlation value indicates that the signal may have large non-stationary changes. The correlation calculation formula is as follows: in, and Represents the i-th segment respectively and benchmarks The spectral energy value of and are the mean of the spectrum energy of the current time-frequency segment and the reference time-frequency segment, respectively, and N is the number of parts. S6. Set a plateau threshold When the spectral energy correlation corr is higher than the threshold, the vibration signal is judged to be a stationary segment, and the data of the stationary segment is retained for subsequent fault diagnosis and health status assessment operations; when the spectral energy correlation corr is lower than the threshold, the vibration signal is judged to be a non-stationary segment, and the signal of this segment will be intercepted and eliminated.

Images