Motor rolling bearing fault feature extraction method based on continuation autocorrelation and sliding time-frequency synchronous averaging

Abstract

The invention discloses a motor rolling bearing fault feature extraction method based on continuation self-correlation and sliding time-frequency synchronous averaging. The method comprises the following steps: firstly, carrying out enhanced periodic component processing on a received vibration signal with fault information by using continuation self-correlation; performing time-frequency analysis on the signals processed by the EACF by using short-time Fourier transform, and dividing a plurality of frequency bands on a time-frequency plane along a frequency axis; respectively taking the center frequency of each frequency band as characteristic line data; respectively extracting a normalized envelope spectrum of the feature data corresponding to each frequency band; and finally, by comparing the maximum energy of the center frequency of each frequency band on an envelope spectrum, selecting and extracting feature line data with the maximum energy, respectively performing sliding time domain synchronous averaging on the line and nearby data lines, and then combining multiple lines of data to realize sliding time frequency synchronous averaging, thereby extracting fault feature information. The method is visual, simple, high in calculation speed and good in precision.

Description

Technical Field

[0001] The invention belongs to the technical field of fault diagnosis, and in particular relates to a method for extracting fault features of a motor rolling bearing based on extended autocorrelation and sliding time-frequency synchronous averaging. Background Art

[0002] Motor rolling bearings are one of the key components commonly used in modern mechanical equipment and are widely present in various machines and motors. This type of bearing consists of an inner ring, an outer ring, rolling elements (such as balls or rollers) and a cage, which support and guide rotating parts, bear loads and reduce friction through these rolling elements. Although designed to extend the reliability and efficiency of equipment operation, motor rolling bearings are still at risk of wear, fatigue and damage when operating under high loads, high speeds and complex environments. Bearing failure can lead to reduced mechanical performance, increased energy consumption, increased maintenance costs, and even mechanical failure or system shutdown in severe cases. Therefore, it is particularly important to effectively monitor the health of motor rolling bearings and diagnose faults. Fault feature extraction is a key step in fault diagnosis, which aims to process and analyze the collected vibration signals in complex industrial environments in order to identify potential fault modes at an early stage. Summary of the invention

[0003] In order to overcome the shortcomings of the prior art, the present invention provides a method for extracting fault features of motor rolling bearings based on extended autocorrelation and sliding time-frequency synchronous averaging. First, the extended autocorrelation function (EACF) is used to strengthen the periodic component of the received vibration signal with fault information; then the signal processed by EACF is subjected to time-frequency analysis using short-time Fourier transform (STFT), and multiple frequency bands are divided along the frequency axis on the time-frequency plane; then the center frequency of each frequency band is taken as its characteristic row data. The normalized envelope spectrum of the characteristic data corresponding to each frequency band is extracted respectively; finally, by comparing the maximum energy of the center frequency of each frequency band on the envelope spectrum, the characteristic row data with the largest energy is selected and extracted, and the sliding time domain synchronous averaging (STSA) is performed on the row and the nearby data rows respectively, and then the multiple rows of data are combined to realize the sliding time-frequency synchronous averaging (STFSA), so as to extract the fault feature information. The method of the present invention utilizes extended autocorrelation to strengthen periodic components, time-frequency analysis and extraction of standardized envelope spectra to locate fault information and sliding time-frequency synchronous averaging to strengthen repeated fault components, and finally extracts bearing fault features. This feature extraction method is intuitive and simple, has fast calculation speed and good accuracy.

[0004] The technical solution adopted by the present invention to solve the technical problem is as follows:

[0005] Step 1: Use the extended autocorrelation EACF to process the received vibration signal with fault information;

[0006] Step 2: Perform time-frequency analysis on the signal after extended autocorrelation processing using short-time Fourier transform (STFT) and divide it into multiple frequency bands along the frequency axis on the time-frequency plane;

[0007] Step 3: Take the center frequency of each frequency band as its characteristic row data and extract the normalized envelope spectrum of the characteristic data corresponding to each frequency band;

[0008] Step 4: By comparing the maximum energy of the center frequency of each frequency band on the envelope spectrum, select and extract the feature row data with the largest energy, perform sliding time domain synchronous average (STSA) on the feature row data and the adjacent data rows, and then splice multiple rows of data to achieve sliding time-frequency synchronous average (STFSA), thereby extracting the fault feature information.

[0009] Preferably, the extended autocorrelation EACF is expressed as follows:

[0010] Step 1-1: Define X(t) as the received fault signal, and the autocorrelation ACF is defined as:

[0011]

[0012] Where τ is the time lag value, E(·) is the expected value calculation, t is the time, and T is the signal duration; the discrete data in X(t) is s[m], m = 1, 2, …, L, and L represents the length of the signal;

[0013] The formula for unbiased autocorrelation UACF is:

[0014]

[0015] In the formula, z is the serial number of the delayed sampling point;

[0016] Step 1-2: Introduce extension operation that keeps the signal length unchanged and establish the EACF method;

[0017] The EACF consists of a constant coefficient multiplied by the dot product calculation between the original signal and the extended signal of the same length as the original signal, and the formula is:

[0018]

[0019] Where s[m] is the original signal, s z ′[m] is the extended signal.

[0020] Preferably, in step 2, the vibration signal x(t) processed by EACF is subjected to time-frequency analysis using STFT and divided into multiple frequency bands. The formula of STFT is:

[0021]

[0022] Where τ is the time shift parameter, f is the frequency, and ω(t-τ) is the window function.

[0023] Preferably, in step 3, Hilbert transform HT is performed on the central feature row data of each frequency band and a normalized envelope spectrum is extracted. The formula of HT is:

[0024]

[0025] In the formula, * is the convolution symbol, and π is the circumference of a circle;

[0026] The formula for calculating the envelope env(x(t)) is:

[0027]

[0028] Where j is an imaginary unit, is the Hilbert transform; finally we get the envelope spectrum S e The formula for (f) is:

[0029]

[0030] In the formula, represents Fourier transform.

[0031] Preferably, step 4 comprises the following steps:

[0032] Step 4-1: Select the frequency band where the feature row data with the maximum energy is located, and the frequency band with the maximum energy R max The formula is:

[0033] R max =max{R[S e1 (f)],R[S e2 (f)],...,R[S ei (f)]} (8)

[0034] Where i is the frequency band number, R[S ei (f)] is the maximum energy of the characteristic line of each frequency band, and max{...} is the calculation of the extracted maximum value;

[0035] Step 4-2: After returning the feature row data with the largest energy, perform STFSA processing on the time-frequency plane;

[0036] First, we need to determine the number of synchronization cycles, the formula is:

[0037]

[0038] Where, T d is the period of the repetitive component;

[0039] After STTFSA, the length of the data will be shortened, and the shortened data length L STFSA for:

[0040] L STFSA =LL win (10)

[0041] Where, L win is the length of the sliding time window;

[0042] The formula for STFSA is:

[0043]

[0044] Where t∈[0,L STFSA ].

[0045] A computer program enables a computer to execute the above-mentioned method for extracting fault features of a motor rolling bearing.

[0046] An electronic device comprises: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device executes the above-mentioned motor rolling bearing fault feature extraction method.

[0047] A computer-readable storage medium stores a computer program, which implements the above-mentioned motor rolling bearing fault feature extraction method when executed by a processor.

[0048] A chip includes: a processor for calling and running a computer program from a memory, so that a device equipped with the chip executes the above-mentioned method for extracting fault features of a motor rolling bearing.

[0049] A computer program product, comprising a computer storage medium storing a computer program, wherein the computer program comprises instructions executable by at least one processor, and when the instructions are executed by the at least one processor, the above-mentioned method for extracting fault features of a motor rolling bearing is implemented.

[0050] The beneficial effects of the present invention are as follows:

[0051] The method of the present invention can accurately and significantly extract fault features in a complex noise environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flow chart of the motor rolling bearing fault feature extraction based on extended autocorrelation and sliding time-frequency synchronous averaging of the present invention;

[0053] Figure 2 is a simulation signal requiring fault feature extraction in an embodiment of the present invention;

[0054] Figure 3 It is a time-frequency image of the simulation signal of the embodiment of the present invention after being directly processed by STFT, and the brightness of the color indicates the magnitude of the energy;

[0055] Figure 4 is a process diagram of iteratively searching for the best extension point when performing EACF in an embodiment of the present invention;

[0056] Figure 5 This is the time-frequency plane of the signal processed by EACF and STFT in the embodiment of the present invention. The brightness of the color indicates the magnitude of the energy.

[0057] Figure 6It is the feature data corresponding to the center frequency of each frequency band after the frequency band is divided on the time-frequency plane in the embodiment of the present invention;

[0058] Figure 7 is the normalized envelope spectrum of the characteristic data of each frequency band in the embodiment of the present invention;

[0059] Figure 8 is a comparison diagram of the maximum amplitude in the envelope spectrum of each set of characteristic data in the embodiment of the present invention;

[0060] Fig. 9 This is a diagram showing the effect of performing STFSA in a frequency band close to the characteristic line after returning it in an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0062] The purpose of the present invention is to propose a method for extracting motor rolling bearing fault features based on extended autocorrelation and sliding time-frequency synchronous averaging, aiming to solve one of the technical problems in the related art at least to a certain extent; this method only needs to collect bearing vibration signals in industrial scenarios and does not require prior knowledge. It is easy to obtain and the calculation is simple and convenient.

[0063] To achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a method for extracting fault features of motor rolling bearings based on extended autocorrelation and sliding time-frequency synchronous averaging, comprising the following steps: step 1, using extended autocorrelation function (EACF) to process the received vibration signal with fault information; step 2, using short-time Fourier transform (STFT) to perform time-frequency analysis on the signal processed by EACF and divide it into multiple frequency bands along the frequency axis on the time-frequency plane; step 3, taking the center frequency of each frequency band as its feature row data and extracting the normalized envelope spectrum of the feature data corresponding to each frequency band; step 4, by comparing the maximum energy of the center frequency of each frequency band on the envelope spectrum, selecting and extracting the feature row data with the largest energy, performing sliding time domain synchronous averaging (STSA) on the row and the nearby data rows, and then splicing multiple rows of data to achieve sliding time-frequency synchronous averaging (STSA). Average,STFSA), thereby extracting fault feature information.

[0064] The motor rolling bearing fault feature extraction method of the present invention utilizes time-frequency analysis, enhances signal periodic components and extracts standardized envelope spectrum to extract bearing fault features. This feature extraction method is intuitive and simple, and has high sensitivity for early fault detection in a noisy environment, further enhancing the comprehensiveness and accuracy of motor rolling bearing state monitoring. In addition, the application of this method is not only limited to laboratory environments, but can also adapt to complex conditions in actual industrial sites, ensuring the practicality and reliability of real-time monitoring and fault warning.

[0065] Example:

[0066] Figure 1 The flowchart of the method for extracting the fault features of the motor rolling bearing based on the extended autocorrelation and the sliding time-frequency synchronous average is shown in the figure. In this embodiment, the method is implemented on the Matlab software platform. The method is demonstrated and explained based on the simulation model.

[0067] Figure 2 is the final input simulation signal, and the fault frequency in this simulation signal is 60Hz.

[0068] Based on this, in this embodiment, the method includes the following steps:

[0069] In the first step, the original simulated vibration signal is analyzed in time and frequency using STFT, which shows that the energy has strong randomness on the time-frequency plane. The formula of STFT is:

[0070]

[0071] Among them, X(t) is the synthetic simulated vibration signal, the window size of STFT is set to 100, the window overlap is set to 99, the sampling frequency fs is set to 5000, and the Fourier transform point number is set to 4096. The image of the original signal made by STFT is as follows Figure 3 As shown, this step is intended to show the randomness of the energy distribution of the original signal before processing.

[0072] Step 2: Calculate the received vibration signal using EACF. EACF is an extension of the traditional autocorrelation function ACF (Autocorrelation Function, ACF). It can not only strengthen the periodic component in the signal, but also solve the problem that when ACF enhances the periodic characteristics, as the effective length of the signal involved in the calculation decreases, the noise suppression will be weakened. The definition of ACF is:

[0073]

[0074] Where τ is the time lag value, E(·) is the expected value calculation, T is the time, and t is the signal duration. In general, the signal is discrete, and the discrete data in X(t) is s[m], m=1,…,L, where L represents the length of the signal and is set to 4630 in the simulation signal. The formula for the unbiased autocorrelation function (UACF) is:

[0075]

[0076] Where z is the serial number of the delayed sampling point. However, as the effective length of the signal involved in the calculation decreases, the noise suppression will weaken. On this basis, an extension operation that keeps the signal length unchanged is introduced to establish the EACF method. Therefore, the EACF can improve the denoising effect in the time domain. EACF is similar to UACF and also includes a constant coefficient multiplied by the dot product calculation between the original signal and the extended signal with the same length as the original signal. The formula of EACF is:

[0077]

[0078] Among them, s[n] is the original signal, s z ′[m] is the extended signal. The best extension point z found during the extension process is 46. The iterative optimization process is shown in the figure below: Figure 4 shown.

[0079] Step 3: Perform STFT time-frequency analysis on the signal processed by EACF and divide it into multiple frequency bands. The parameter settings are the same as described above. In order to show the strengthening effect of EACF on periodic components, the time-frequency plane image is as follows: Figure 5 As shown, the time-frequency plane effect of STFT is compared with the original signal ( Figure 3 ), it can be clearly seen that the periodic component has been strengthened.

[0080] Step 4: The center frequency of each frequency band is used as the feature row data of each frequency band. The extracted feature row data is as follows: Figure 6 As shown, the time domain signal on the characteristic line of each frequency band is processed by HT and its normalized envelope spectrum is extracted. The formula of HT is:

[0081]

[0082] Where, π = 3.14. The formula for calculating the envelope signal env(x(t)) using the HT method is:

[0083]

[0084] where j is the imaginary unit, is the Hilbert transform. Finally, we get the envelope spectrum S e The formula for (f) is:

[0085]

[0086] in, represents Fourier transform. The normalized envelope spectrum of each characteristic row signal is calculated as follows: Figure 7 shown.

[0087] Step 5: By comparing the maximum energy of the center frequency of each frequency band on the envelope spectrum, select the characteristic row with the largest energy and return it. The maximum amplitude corresponding to each frequency band is as follows: Figure 8 As shown, select the energy maximum R max The formula is:

[0088] R max =max{R[S e1 (f)],R[S e2 (f)],...,R[S ei (f)]} (19)

[0089] Among them, the maximum frequency band number i is 30, R[S ei (f)] is the maximum energy of the characteristic line of each frequency band, max{...} is the calculation for extracting the maximum value, the frequency band number of the maximum amplitude is 5, and the maximum normalized amplitude is 0.159207. After returning the characteristic line data with the largest energy, STFSA processing is performed on the time-frequency plane to provide more comprehensive fault information while maintaining the time-frequency characteristics of the signal. First, the number of synchronization cycles must be determined, and the formula is:

[0090]

[0091] Among them, the period of the repetitive component is T d is 167. Because of the involvement of the sliding time window, the data length after STFSA will inevitably be shortened. The sliding time window is set to the length of 13 repeated component cycles. The shortened data length L STFSA for:

[0092] L STFSA =LL win (twenty one)

[0093] Among them, the sliding time window length L win is the length of the 13-repetition component period, i.e. L win Take 2127. After this preliminary work, perform STFSA calculation to extract fault features. The formula of STFSA is:

[0094]

[0095] Among them, t∈[0,L STFSA ]. From the previous calculation, we can know the shortened data length L STFSA is 2277, the feature row data is close to the range i is 300, and the period of the repeated component is T d is 167, this method can effectively extract the fault signal repeated in the time dimension in the time-frequency plane. Fig. 9 The periodic fault pulses on the time-frequency plane can be significantly extracted. It can be seen that this method can accurately and significantly extract fault features in a complex noise environment.

Claims

1. A method for extracting fault features of motor rolling bearings based on extended autocorrelation and sliding time-frequency synchronous averaging, characterized in that: The steps include: Step 1: Use the extended autocorrelation EACF to process the received vibration signal with fault information; Step 2: Perform time-frequency analysis on the signal after extended autocorrelation processing using short-time Fourier transform (STFT) and divide it into multiple frequency bands along the frequency axis on the time-frequency plane; Step 3: Take the center frequency of each frequency band as its characteristic row data and extract the normalized envelope spectrum of the characteristic data corresponding to each frequency band; Step 4: By comparing the maximum energy of the center frequency of each frequency band on the envelope spectrum, select and extract the feature row data with the largest energy, perform sliding time domain synchronous average (STSA) on the feature row data and the adjacent data rows, and then splice multiple rows of data to achieve sliding time-frequency synchronous average (STFSA), thereby extracting the fault feature information.

2. The method for extracting fault features of motor rolling bearings based on extended autocorrelation and sliding time-frequency synchronous averaging according to claim 1 is characterized in that: Preferably, the extended autocorrelation EACF is expressed as follows: Step 1-1: Define X(t) as the received fault signal, and the autocorrelation ACF is defined as: Where τ is the time lag value, E(·) is the expected value calculation, t is the time, and T is the signal duration; the discrete data in X(t) is S[m], m = 1, 2, …, L, and L represents the length of the signal; The formula for unbiased autocorrelation UACF is: In the formula, z is the serial number of the delayed sampling point; Step 1-2: Introduce extension operation that keeps the signal length unchanged and establish the EACF method; The EACF consists of a constant coefficient multiplied by the dot product calculation between the original signal and the extended signal of the same length as the original signal, and the formula is: Where s[m] is the original signal, s z ′[m] is the extended signal.

3. The method for extracting fault features of motor rolling bearings based on extended autocorrelation and sliding time-frequency synchronous averaging according to claim 2 is characterized in that: In step 2, the vibration signal x(t) processed by EACF is subjected to time-frequency analysis using STFT and divided into multiple frequency bands. The formula of STFT is: Where τ is the time shift parameter, f is the frequency, and ω(t-τ) is the window function.

4. The method for extracting fault features of motor rolling bearings based on extended autocorrelation and sliding time-frequency synchronous averaging according to claim 3 is characterized in that: In step 3, the central feature row data of each frequency band is subjected to Hilbert transform HT and the normalized envelope spectrum is extracted. The formula of HT is: In the formula, * is the convolution symbol, and π is the circumference of a circle; The formula for calculating the envelope env(x(t)) is: Where j is an imaginary unit, is the Hilbert transform; finally we get the envelope spectrum S e The formula for (f) is: In the formula, represents Fourier transform.

5. The method for extracting fault features of motor rolling bearings based on extended autocorrelation and sliding time-frequency synchronous averaging according to claim 4 is characterized in that: The step 4 comprises the following steps: Step 4-1: Select the frequency band where the feature row data with the maximum energy is located, and the frequency band with the maximum energy R max The formula is: R max =max{R[S e1 (f)],R[S e2 (f)],...,R[S ei (f)]} (8) Where i is the frequency band number, R[S ei (f)] is the maximum energy of the characteristic line of each frequency band, and max{…} is the calculation of the extracted maximum value; Step 4-2: After returning the feature row data with the largest energy, perform STFSA processing on the time-frequency plane; First, we need to determine the number of synchronization cycles, the formula is: Where, T d is the period of the repetitive component; After STTFSA, the length of the data will be shortened, and the shortened data length L STFSA for: L STFSA =LL win (10) Where, L win is the length of the sliding time window; The formula for STFSA is: Where t∈[0,L STFSA ].

6. A computer program, characterized in that The computer program enables a computer to execute the method according to any one of claims 1 to 5.

7. An electronic device, characterized in that: include: Processor and memory; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device executes the method as claimed in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

9. A chip, characterized in that: include: A processor, configured to call and run a computer program from a memory, so that a device equipped with the chip executes a method as claimed in any one of claims 1 to 5.

10. A computer program product, characterized in that The computer program product comprises a computer storage medium storing a computer program, wherein the computer program comprises instructions executable by at least one processor, and when the instructions are executed by the at least one processor, the method according to any one of claims 1 to 5 is implemented.

Images