CN115456019A - Rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN - Google Patents
Rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN Download PDFInfo
- Publication number
- CN115456019A CN115456019A CN202211088353.2A CN202211088353A CN115456019A CN 115456019 A CN115456019 A CN 115456019A CN 202211088353 A CN202211088353 A CN 202211088353A CN 115456019 A CN115456019 A CN 115456019A
- Authority
- CN
- China
- Prior art keywords
- signal
- noise
- rolling bearing
- components
- fault diagnosis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
Landscapes
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention provides a rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN, firstly screening signal components of acquired vibration signals through a decomposition algorithm and a kurtosis value-correlation coefficient screening criterion and carrying out preliminary reconstruction, and dividing high-noise signals and low-noise signals with different noise components; decomposing a high-noise signal containing more noise components in a wavelet packet transformation mode, further reconstructing according to the kurtosis value of a signal component to achieve the effect of filtering background environment noise, and finally reconstructing with a low-noise signal divided by a screening criterion to generate a denoising signal; carrying out feature extraction on the generated de-noising signal in an envelope demodulation mode, and comparing the actual fault feature frequency with the theoretical fault feature frequency to carry out fault diagnosis; tests prove that the signal obtained by denoising through the method effectively filters background environment noise components, improves the diagnosis resolution of the traditional fault diagnosis method, has higher accuracy and has practical engineering application value.
Description
Technical Field
The invention belongs to the technical field of bearing fault diagnosis, and particularly relates to a rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN.
Background
The rotating machine is dominant in the industrial production development process, and the rolling bearing as an important component in the rotating machine has important significance for the running state of the rotating machine and even the modern industrial development. Therefore, it is of practical significance to realize accurate failure diagnosis of the rolling bearing. The rolling bearing comprises an inner ring, an outer ring, a rolling body, a retainer and the like; the rolling bodies comprise various types such as balls, cylindrical rollers, roller pins, tapered rollers and spherical rollers, all parts of the rolling bearing are affected by factors such as insufficient lubrication, improper assembly and overload operation in an operating state, so that faults are easy to occur, and the safety, precision and other performances of mechanical equipment are seriously affected in the operating process of the rolling bearing in a fault state, so that the rolling bearing can accurately monitor the daily operating state of the rolling bearing and perform fault diagnosis, and the rolling bearing plays an important role in guaranteeing the performance of the mechanical equipment. The rolling bearing vibration signal in the running state has the non-stationary nonlinear characteristic, and the most effective analysis method for the vibration signal with the characteristic is time-frequency analysis; common time-frequency analysis methods include fourier transform (FFT), wavelet transform, and Empirical Mode Decomposition (EMD).
The Fourier transform (FFT) has the problems of low characteristic frequency resolution and incapability of accurately positioning signals in the time-frequency analysis process of non-stationary signals; in the link of filtering background environmental noise, wavelet transformation has a good noise removal effect, but in principle, wavelet transformation can only transform low-frequency signals and ignore information of a high-frequency part, so that the wavelet transformation is optimized, wavelet packet transformation is provided, the high-frequency part and the low-frequency part are transformed, more effective information is reserved in the process of screening and reconstructing, and meanwhile, a good noise filtering effect can be achieved. Empirical Mode Decomposition (EMD) is a novel signal analysis method proposed in 1998 by WO Huang E of NASA, huang E, and the doctor considers that any complex signal is formed by overlapping a plurality of signal components, and essentially, a section of non-stationary vibration signal is subjected to smoothing processing, and the complex signal is decomposed step by step to generate a series of inherent modal components (IMF) containing different specific time scale scales, but the decomposition efficiency is low due to the lack of strict mathematical basis of the EMD decomposition principle, and the IMF component loses significance due to the problem that the IMF component generates signals with different time scales and frequencies in the same IMF component or modal aliasing with the same time scale and frequency in different IMF components; based on the problem, wu provides Ensemble Empirical Mode Decomposition (EEMD) through the research on the white Noise statistical characteristics in 2009, and solves the problem of modal aliasing, but the EEMD Decomposition has the problem of white Noise residue, which affects the subsequent signal analysis processing, torres et al improves the EEMD Decomposition method on the basis of the EEMD algorithm, and provides Complete Adaptive Noise set Empirical Mode Decomposition (cemdan), which is different from EEMD, and the cemdan Decomposition obtains an IMF component added with limited times of Adaptive white Noise, so as to reduce the times of Ensemble average calculation.
The fault feature extraction is a key step in the fault diagnosis process, the traditional fault diagnosis method comprises an expert diagnosis method, an envelope resonance demodulation method and other methods, the traditional fault diagnosis method is subjected to experience judgment depending too much on manual professional ability, therefore misdiagnosis caused by incomplete information is easy to occur, and the extraction accuracy of the fault feature is low. In the fault diagnosis process, the acquired vibration signals are greatly influenced by the environmental background noise on subsequent signal analysis and processing, so that the filtering work of the background environmental noise is very important in the fault diagnosis process, but the traditional noise filtering method such as wavelet packet transformation denoising can filter partial effective information while filtering the components of the background environmental noise. However, the traditional noise filtering method is optimized through multi-step decomposition, screening and division, so that background environment noise components can be filtered, and meanwhile, more effective information can be reserved, and a better fault diagnosis effect can be achieved.
Disclosure of Invention
Based on the problems, the invention provides a rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN, which comprises the following steps:
step 1: collecting vibration signals of a rolling bearing;
step 2: decomposing the acquired vibration signals of the rolling bearing to decompose different signal components;
the signal component is an IMF component, and the CEEMDAN is adopted to decompose the acquired vibration signal to decompose different IMF signal components;
and 3, step 3: combining the kurtosis value and the correlation coefficient, screening and dividing signal components generated after decomposition into a high-noise signal and a low-noise signal, and preliminarily reconstructing; the method comprises the following steps:
step 3.1: the kurtosis value ku for each signal component is calculated as follows:
where N represents the number of discrete sequence points, σ represents the standard deviation, x i Representing the vibration amplitude corresponding to the discrete sequence point of the time domain waveform, wherein x represents the average amplitude of the discrete sequence;
step 3.2: calculating kurtosis value ku of the acquired vibration signal 0 ；
Step 3.3: calculating the Pearson correlation coefficient r of each signal component IMF ；
Step 3.4: calculating a correlation coefficient threshold value T of the acquired vibration signal according to the Pearson correlation coefficient of each signal component as follows:
wherein, | r IMF | represents the absolute value of the correlation coefficient of each signal component,
representing the average value of the correlation coefficients of the signal components, n representing the number of the signal components;
step 3.5: and dividing the kurtosis value of each signal component with the kurtosis value of the acquired vibration signal to realize normalization, and summing the kurtosis values with the absolute values of Pearson correlation coefficients of the signal components to obtain a screening index K:
step 3.6: summing the normalized results of the threshold of the correlation number and the kurtosis value of the vibration signal to obtain a contrast parameter K 0 ：
Step 3.7: the signal components are sorted in a descending order according to the screening index K and compared with the comparison parameter K 0 Comparing to obtain a comparison parameter K 0 Part of the signal components are superposed and reconstructed to generate a high-noise signal which is smaller than a contrast parameter K 0 Part of the signal components are superposed and reconstructed to generate a low-noise signal;
and 4, step 4: decomposing the reconstructed high-noise signal through wavelet packet transformation to generate different WPT signal components, and reconstructing the signal components after noise filtering and the low-noise signal to obtain a final de-noising signal, wherein the method comprises the following steps:
step 4.1: carrying out one-dimensional three-order wavelet packet transformation on the high-noise signal obtained by reconstruction to generate a plurality of WPT signal components;
step 4.2: respectively calculating kurtosis values of all WPT signal components and arranging in a descending order according to the kurtosis values;
step 4.3: selecting two WPT signal components from the WPT signal components with the kurtosis value ranked in the top Q for superposition reconstruction and filtering noise components in a high-noise signal;
step 4.4: superposing and reconstructing the high-noise signal and the low-noise signal with the noise components removed to obtain a one-dimensional de-noising signal;
and 5: and carrying out feature extraction on the de-noised signal in an envelope demodulation mode, and carrying out fault diagnosis on the rolling bearing by using the extracted actual fault feature frequency and the calculated theoretical fault feature frequency.
The invention has the beneficial effects that:
1) The CEEMDAN decomposition adopted by the method provided by the invention effectively eliminates the problem of modal aliasing, has better local characteristics under the same time and frequency scale, and effectively improves the operation efficiency of the algorithm by adopting an overall average calculation method; the selected wavelet packet transformation has high operation efficiency, is used as an extension of the traditional wavelet transformation, has good effect of filtering environmental background noise components, and simultaneously transforms high-frequency detail parts to keep more effective information;
2) The kurtosis value-correlation coefficient screening criterion selected by the method provided by the invention is used for dividing IMF signal components generated by decomposition into different noise component contents, so that the one-sidedness of the IMF components selected by single index parameter screening is avoided, and the effects of dividing and primarily reconstructing to generate high-noise signals and low-noise signals are achieved;
3) Carrying out fault feature extraction on the processed de-noised signal in an envelope demodulation mode, and extracting modulation frequency which takes the fault feature frequency as a center and the rotating frequency as a sideband besides extracting the fault feature frequency and the rotating frequency, so that the fault diagnosis accuracy of the rolling bearing is improved;
4) The method provided by the invention is applied to the real vibration signal of the rolling bearing, and the result shows that the acquired vibration signal is denoised by the method, so that the background environmental noise component in the running state of the rolling bearing can be effectively filtered, the influence of the environmental background noise on the quality of the acquired vibration signal is eliminated, the fault diagnosis precision rate is higher, compared with the traditional wavelet packet transformation fault diagnosis method, the fault diagnosis precision rate is effectively improved, more effective information is reserved, and the method has practical engineering application value.
Drawings
FIG. 1 is a flowchart of a rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN in an embodiment of the present invention;
fig. 2 is a time domain diagram of a simple path experiment bearing raw signal of the university of west reservoir in an embodiment of the present invention, wherein (a) is a time domain diagram of a simple path experiment bearing inner ring fault sample raw vibration signal of the university of west reservoir in the embodiment, and (b) is a time domain diagram of a simple path experiment bearing outer ring fault sample raw vibration signal of the university of west reservoir in the embodiment;
fig. 3 is a time domain diagram of a denoised signal of a simple path experimental bearing of the university of west reservoir based on wavelet packet transformation and CEEMDAN in the embodiment of the present invention, wherein (a) is a time domain diagram of a denoised signal of a fault sample of an inner ring of the simple path experimental bearing of the university of west reservoir in the embodiment, and (b) is a time domain diagram of a denoised signal of a fault sample of an outer ring of the simple path experimental bearing of the university of west reservoir in the embodiment;
fig. 4 is a schematic diagram of a time domain comparison of a de-noised signal and an original vibration signal of an aircraft engine complex path intermediate bearing based on wavelet packet transformation and CEEMDAN in an embodiment of the present invention, wherein (a) is a schematic diagram of a time domain comparison of a de-noised signal and an original vibration signal of an aircraft engine complex path intermediate bearing inner ring fault sample in an embodiment, and (b) is a schematic diagram of a time domain comparison of a de-noised signal and an original vibration signal of an aircraft engine complex path intermediate bearing outer ring fault sample in an embodiment;
FIG. 5 is a frequency range envelope spectrum of a denoised signal 0-800Hz based on wavelet packet transform and CEEMDAN of a simple path experimental bearing of the West university in the embodiment of the present invention, wherein (a) is the frequency range envelope spectrum of a denoised signal 0-800Hz of a simple path experimental bearing inner ring fault sample of the West university in the embodiment, and (b) is the frequency range envelope spectrum of a denoised signal 0-800Hz of a simple path experimental bearing outer ring fault sample of the simple path experimental bearing of the West university in the embodiment;
FIG. 6 is a frequency range envelope spectrum of a denoised signal 0-200Hz based on wavelet packet transform and CEEMDAN of a simple path experimental bearing of the West university in the embodiment of the present invention, wherein (a) is the frequency range envelope spectrum of a denoised signal 0-200Hz of a simple path experimental bearing inner ring fault sample of the West university in the embodiment, and (b) is the frequency range envelope spectrum of a denoised signal 0-200Hz of a simple path experimental bearing outer ring fault sample of the simple path experimental bearing of the West university in the embodiment;
FIG. 7 is a frequency range envelope spectrum of a denoised signal 0-200Hz based on wavelet packet transformation and CEEMDAN for an aero-engine complex path intermediate bearing in an embodiment of the invention, wherein (a) is the frequency range envelope spectrum of a denoised signal 0-200Hz for an aero-engine complex path intermediate bearing inner ring fault sample in the embodiment, and (b) is the frequency range envelope spectrum of a denoised signal 0-200Hz for an aero-engine complex path intermediate bearing outer ring fault sample in the embodiment;
FIG. 8 is a 0-200Hz envelope spectrum of a denoised signal of an aircraft engine complex path intermediate bearing inner ring fault sample based on traditional wavelet packet transformation in the embodiment of the invention.
Detailed Description
The invention is further described with reference to the following figures and specific examples. The invention relates to a fault diagnosis method which decomposes a vibration signal through a CEEMDAN algorithm, screens and divides the vibration signal through kurtosis value-related coefficient multi-parameter indexes, and combines the kurtosis value-related coefficient multi-parameter indexes with a one-dimensional third-order wavelet packet transformation denoising method. Firstly, carrying out CEEMDAN decomposition on an acquired vibration signal, then dividing and preliminarily reconstructing an IMF component generated by decomposition by using a kurtosis value-correlation coefficient screening criterion to generate a high-noise signal and a low-noise signal, decomposing the high-noise signal generated by preliminary reconstruction by using one-dimensional third-order Wavelet Packet Transform (WPT) to generate 8 WPT signal components, screening according to the kurtosis values of the components, further reconstructing to realize the filtering of noise components of the high-noise signal, and finally reconstructing the denoised high-noise signal and the low-noise signal generated by preliminarily reconstructing the kurtosis value-correlation coefficient screening criterion to generate a denoised signal; finally, envelope demodulation is carried out on the de-noised signal, fault characteristic frequency is extracted, and fault diagnosis of the rolling bearing is completed. Compared with other rolling bearing fault diagnosis methods, the method effectively filters background environmental noise components to eliminate the influence on signal analysis, improves the accuracy of fault diagnosis and retains more effective information.
As shown in fig. 1, the rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN includes:
step 1: collecting vibration signals of a rolling bearing by adopting a vibration sensor;
step 2: decomposing the acquired vibration signals of the rolling bearing by adopting a CEEMDAN decomposition method to decompose different IMF components;
the principle of CEEMDAN is as follows: CEEMDAN is a novel decomposition method which is provided and formed by improvement on the basis of EMDAN, IMF components of adaptive white noise with limited times are obtained after CEEMDAN decomposition, the times of overall average calculation are reduced, in the decomposition process, the IMF components decomposed by the EMDAN are subjected to overall average, the CEEMDAN is directly subjected to the overall average calculation after first-order IMF is decomposed, and mathematically, a one-dimensional vibration signal is formed by adding different components generated after the CEEMDAN decomposition.
Firstly, adding white Gaussian noise z into an original vibration signal y (t) for N times i' (t)(i'＝1,2,…,N)；
y i' (t)＝y(t)+z i' (t)(i'＝1,2,...,N')
For vibration signal y added with Gaussian white noise i' (t) EMD decomposition to generate M IMFs m (t) (M =1,2, …, M), and a ensemble average calculation is performed to obtain a first-order intrinsic mode component IMF 1 Then adding the vibration signal y of Gaussian white noise i' (t) and IMF 1 The subtraction yields the margin h 1 (t) the following:
h 1 (t)＝y(t)-IMF 1
for the margin h generated after the first stage operation 1 (t) continuously adding Gaussian white noise and performing EMD to generate second-order intrinsic mode component IMF 2 And a second order margin h 2 (t), according to the calculation characteristics, performing EMD decomposition on the vibration signal for multiple times until the signal margin after the Mth decomposition can not be decomposed continuously, completing a CEEMDAN decomposition process, and finally performing CEEMDAN decomposition on the signal y' (t) as follows:
in the formula, IMF m (t) represents the mth order natural modal component, h, generated after the mth decomposition M And (t) represents the margin of the vibration signal after EMD decomposition for multiple times until the Mth decomposition is not continued in the calculation process.
And step 3: screening and preliminarily reconstructing IMF signal components generated after decomposition by combining kurtosis values and correlation coefficients, and realizing division of signal components with different noise component contents into high-noise signals and low-noise signals; the method comprises the following steps:
the process of screening the IMF signal components generated after decomposition by combining the kurtosis value and the correlation coefficient is as follows:
step 3.1: the kurtosis value ku for each IMF component is calculated as follows:
where N represents the number of discrete sequence points, σ represents the standard deviation, x i Representing the vibration amplitude corresponding to the discrete sequence point of the time domain waveform,
representing the average amplitude of the discrete sequence;
step 3.2: calculating kurtosis value ku of the acquired vibration signal 0 ；
Step 3.3: calculating the absolute value of the Pearson's correlation coefficient | r of each IMF component IMF |；
Step 3.4: calculating a correlation coefficient threshold value T of the acquired vibration signal according to the Pearson correlation coefficient of each IMF component as follows:
wherein, | r IMF | represents the absolute value of the correlation coefficient for each IMF signal component,
representing the average value of the correlation coefficients of the IMF signal components, and n represents the number of the IMF signal components;
step 3.5: normalizing the kurtosis value of each IMF signal component according to the acquired kurtosis value of the vibration signal and summing the kurtosis value of each IMF component with the Pearson correlation coefficient absolute value of each IMF component to obtain a screening index K:
step 3.6: summing the normalized results of the threshold T of the correlation coefficient method and the kurtosis value of the vibration signal, and calculating to obtain a contrast parameter K 0 ：
In the embodiment, a Pearson correlation coefficient is adopted to describe the correlation between each IMF component and the acquired vibration signal, the generated correlation coefficient range is [ -1,1], and the larger the absolute value of the correlation coefficient is, the larger the linear correlation degree between variables is; conversely, the smaller the correlation coefficient, the smaller the degree of linear correlation between the variables.
Let two samples be X and Y, respectively, and the correlation coefficients be:
where r is the Pearson correlation coefficient for two variables, cov (X, Y) is the covariance of X and Y,
the variance of the variable X is represented by,
represents the variance of Y.
In this embodiment, each IMF component is sorted in descending order according to the screening index K and compared with the comparison parameter K 0 And performing comparison, reconstructing the part larger than the comparison parameter to generate a high-noise signal, and reconstructing the part smaller than the comparison parameter to generate a low-noise signal.
And 4, step 4: and processing the one-dimensional high-noise signals preliminarily reconstructed and divided by a kurtosis value-correlation coefficient screening criterion by adopting a one-dimensional third-order wavelet packet transformation method, decomposing different WPT components, performing descending order arrangement according to the kurtosis values of 8 decomposed WPT components, selecting 2 WPT signal components from the WPT signal components with the kurtosis values arranged at the top Q for superposition reconstruction, filtering noise components in the high-noise signals, reconstructing the high-noise signals and the low-noise signals to finally generate de-noising signals, and realizing the filtering of the background environment noise.
The signal component is a WPT component, the collected vibration signals are decomposed by wavelet packet transformation, different WPT signal components are decomposed, appropriate components are screened out according to kurtosis values to be further reconstructed, the noise components in the high-noise signals are filtered out, the high-noise signals with the noise components being filtered out and the low-noise signals which are divided and preliminarily reconstructed by the kurtosis value-correlation coefficient screening criterion are finally reconstructed to generate the de-noising signals.
And 5: extracting the actual fault characteristic frequency (including the frequency of an inner ring, an outer ring, a rolling body and a retainer) of the rolling bearing by carrying out envelope demodulation on the de-noised signal, calculating the theoretical fault characteristic frequency according to the bearing parameters of the experimental rolling bearing fault sample, and carrying out fault diagnosis by comparing the extracted actual fault characteristic frequency with the theoretical fault characteristic frequency;
the theoretical characteristic frequency calculation formula of the rolling bearing fault is as follows:
wherein Z is the number of rolling elements, FFor the rotation frequency of the rolling bearing, D is the diameter of the ball, D is the pitch diameter of the bearing raceway, alpha is the contact angle of the bearing, f i For inner ring fault characteristic frequency, f o Is the outer ring fault characteristic frequency, f c For a characteristic frequency of rolling element failure, f r Is the cage failure signature frequency.
According to the fault diagnosis theory, the allowable actual fault characteristic frequency error range is within the frequency resolution range, and the frequency resolution is the ratio of the sampling frequency to the actual sampling point number as follows:
wherein e is max Indicating the maximum allowable error range, i.e. the frequency resolution, f s The sampling frequency is N ', and the number of sampling points is N';
and performing fault diagnosis explanation by taking the inner-circle fault characteristic judgment as an example. The allowable frequency error range of the inner ring fault set according to actual conditions is [ f i -e max ,f i +e max ]If the actual fault characteristic frequency extracted by the envelope demodulation mode is within the error range, the rolling bearing inner ring can be judged to be in fault.
The embodiment adopts real experimental data for analysis, and the real experimental data are respectively obtained from simple path experimental bearings of the western storage university and medium bearing data of complex paths of the aero-engine. The sampling frequency of the bearing vibration signal data of the simple path experiment of the university of western storage is 12000Hz; the sampling frequency of the vibration signal data of the medium bearing in the complex path of the aircraft engine is 25600Hz.
And (3) adopting the bearing fault sample data of the simple path experiment of the university of western reservoir, and selecting a fault bearing for analysis as a 6205-2RJEM SKF type deep groove ball bearing. The sampling frequency of the vibration data was 12000Hz. Selecting inner and outer ring simple path experiment bearing fault sample data for analysis, wherein the time domain signal of the original vibration signal of the collected inner ring fault sample data is shown in figure 2 (a); the time domain signal of the original vibration signal of the collected outer ring fault sample data is shown in fig. 2 (b).
Environmental background noise filtering is performed on simple path experiment bearing fault sample data through a rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN, wherein time domain signals of the simple path rolling bearing inner ring fault sample are shown in fig. 3 (a), and time domain signals of the simple path rolling bearing outer ring fault sample are shown in fig. 3 (b).
The comparison shows that the method provided by the patent has a good effect of filtering the background environmental noise of the rolling bearing with the simple path.
Envelope demodulation is carried out on the denoising signals of the inner fault sample and the outer fault sample of the simple path rolling bearing, firstly, the frequency range of 0-800Hz is intercepted, the characteristic frequency and the frequency multiplication of the denoising signals of the inner fault sample and the outer fault sample of the simple path rolling bearing are extracted, and the characteristic extraction result of the frequency range of 0-800Hz of the inner fault sample of the simple path rolling bearing is shown in figure 5 (a); the simple path rolling bearing outer ring fault sample 0-800Hz frequency range feature extraction result is shown in fig. 5 (b).
Reducing the frequency range, intercepting the frequency range of 0-200Hz, and extracting the modulation frequency of the denoising signals of the inner and outer ring fault samples of the simple path rolling bearing, wherein the extraction result of the frequency range of 0-200Hz of the inner ring fault sample of the simple path rolling bearing is shown in figure 6 (a); the simple path rolling bearing outer ring fault sample 0-200Hz frequency range feature extraction result is shown in FIG. 6 (b).
As can be seen from the extraction results in fig. 5 and 6, the rolling bearing fault diagnosis method based on the wavelet packet transform and the CEEMDAN can achieve a good fault diagnosis effect on the simple path rolling bearing, so that the theoretical feasibility of the rolling bearing fault diagnosis method based on the wavelet packet transform and the CEEMDAN can be verified.
In the embodiment, a simulation experiment table is built by adopting a 5-fulcrum intermediate bearing structure of a certain turbofan engine for data acquisition, the fault sample of the rolling bearing in the complex path of the aircraft engine is simulated, the highest rotating speed of the experiment table of the engine rotor system is 18000r/min, the maximum radial load is 20kN, the rotating directions of the inner ring and the outer ring can be controlled, and the sampling frequency of vibration data is 25600Hz.
The adopted rolling bearing is an intermediate bearing of the aero-engine, the inner ring and the outer ring are subjected to linear cutting processing to simulate faults of the intermediate bearing of the aero-engine, vibration signals of the rolling bearing are collected through a sensor, after noise filtering is carried out through a rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN, a time domain graph pair of an inner ring fault signal and a collected inner ring fault sample vibration signal is shown in a figure 4 (a); the time domain diagram of the outer ring fault signal and the collected outer ring fault sample vibration signal is shown in fig. 4 (b).
The comparison shows that the rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN has good effect on filtering background environmental noise of vibration signals collected by the rolling bearing under a complex path.
Envelope demodulation is carried out on the processed fault sample of the complex path rolling bearing, the frequency range of 0-200Hz is intercepted and analyzed, the characteristic extraction result in the frequency range of 0-200Hz of the noise removal signal of the fault sample of the inner ring of the complex path rolling bearing is shown in fig. 7 (a), and the characteristic extraction result in the frequency range of 0-200Hz of the noise removal signal of the fault sample of the outer ring of the complex path rolling bearing is shown in fig. 7 (b).
As shown in fig. 7, the frequency conversion, the fault characteristic frequency and the frequency doubling of the fault samples of the inner ring and the outer ring of the rolling bearing with the complex path can be clearly extracted within the frequency range of 0-200Hz, the fault frequency and the frequency doubling are taken as centers, the frequency conversion is taken as modulation frequency of a sideband, the vibration signal of the rolling bearing with the complex path is processed by a rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN, the characteristic of the inner ring is extracted, and the error between the actual characteristic frequency and the theoretical characteristic frequency is 1.2%; the error between the actual characteristic frequency and the theoretical characteristic frequency is 1.5% when the outer ring characteristic is extracted, and the rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN is verified to have good actual engineering application value.
The method comprises the steps of carrying out fault diagnosis on a complex path rolling bearing inner ring fault sample through a traditional wavelet packet transformation fault diagnosis method, and intercepting a frequency range of 0-200Hz to carry out feature extraction, wherein the feature extraction result is shown in figure 8.
Compared with the traditional wavelet packet transformation fault diagnosis method, the wavelet packet transformation and CEEMDAN-based rolling bearing fault diagnosis method is used for comparing the fault feature extraction results of the complex path rolling bearing inner ring fault sample with the complex path rolling bearing inner ring fault sample, so that the wavelet packet transformation and CEEMDAN-based rolling bearing fault diagnosis method can improve the resolution of fault diagnosis, can extract more fault feature frequencies, and has better rolling bearing fault diagnosis effect.
Claims (4)
1. A rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN is characterized by comprising the following steps:
step 1: collecting vibration signals of a rolling bearing;
step 2: decomposing the collected vibration signals of the rolling bearing to separate out different signal components;
and step 3: combining the kurtosis value and the correlation coefficient, screening and dividing the signal components generated after decomposition into a high-noise signal and a low-noise signal, and preliminarily reconstructing the signal components;
and 4, step 4: decomposing the reconstructed high-noise signal through wavelet packet transformation to generate different WPT signal components, and reconstructing the WPT signal components and the low-noise signal after noise filtering to obtain a final de-noising signal;
and 5: and carrying out feature extraction on the de-noised signal in an envelope demodulation mode, and carrying out fault diagnosis on the rolling bearing by using the extracted actual fault feature frequency and the calculated theoretical fault feature frequency.
2. The rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN as claimed in claim 1, wherein the signal component in step 2 is an IMF component, and CEEMDAN is used to decompose the acquired vibration signal to separate out different IMF signal components.
3. The rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN as claimed in claim 1, wherein the step 3 comprises:
step 3.1: the kurtosis value ku for each signal component is calculated as follows:
where N represents the number of discrete sequence points, σ represents the standard deviation, x i Representing the vibration amplitude corresponding to the discrete sequence point of the time domain waveform,
representing the average amplitude of the discrete sequence;
step 3.2: calculating kurtosis value ku of the acquired vibration signal 0 ；
Step 3.3: calculating the Pearson correlation coefficient r of each signal component IMF ；
Step 3.4: calculating a correlation coefficient threshold T of the acquired vibration signal according to the Pearson correlation coefficient of each signal component, as follows:
wherein, | r IMF | represents the absolute value of the correlation coefficient of each signal component,
representing the average value of the correlation coefficients of the signal components, n representing the number of the signal components;
step 3.5: and dividing the kurtosis value of each signal component with the kurtosis value of the acquired vibration signal to realize normalization, and summing the kurtosis values with the absolute values of Pearson correlation coefficients of the signal components to obtain a screening index K:
step 3.6: summing the normalized results of the threshold of the correlation number and the kurtosis value of the vibration signal to obtain a contrast parameter K 0 ：
Step 3.7: the signal components are sorted in descending order according to the screening index K and compared with the comparison parameter K 0 Comparing to obtain a comparison parameter K 0 Part of the signal components are superposed and reconstructed to generate a high-noise signal which is smaller than a contrast parameter K 0 And part of the signal components are superposed and reconstructed to generate a low-noise signal.
4. The rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN as claimed in claim 1, wherein the step 4 comprises:
step 4.1: carrying out one-dimensional three-order wavelet packet transformation on the high-noise signal obtained by reconstruction to generate a plurality of WPT signal components;
step 4.2: respectively calculating kurtosis values of all WPT signal components and arranging in a descending order according to the kurtosis values;
step 4.3: selecting two WPT signal components from the WPT signal components with the kurtosis value ranked in the top Q for superposition reconstruction and filtering noise components in a high-noise signal;
step 4.4: and superposing and reconstructing the high-noise signal and the low-noise signal with the noise components removed to obtain a one-dimensional de-noising signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211088353.2A CN115456019A (en) | 2022-09-07 | 2022-09-07 | Rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211088353.2A CN115456019A (en) | 2022-09-07 | 2022-09-07 | Rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115456019A true CN115456019A (en) | 2022-12-09 |
Family
ID=84303365
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211088353.2A Pending CN115456019A (en) | 2022-09-07 | 2022-09-07 | Rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115456019A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116202770A (en) * | 2023-03-21 | 2023-06-02 | 广东海洋大学 | Bearing fault diagnosis simulation experiment device |
| CN117030268A (en) * | 2023-10-07 | 2023-11-10 | 太原科技大学 | Rolling bearing fault diagnosis method |
-
2022
- 2022-09-07 CN CN202211088353.2A patent/CN115456019A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116202770A (en) * | 2023-03-21 | 2023-06-02 | 广东海洋大学 | Bearing fault diagnosis simulation experiment device |
| CN117030268A (en) * | 2023-10-07 | 2023-11-10 | 太原科技大学 | Rolling bearing fault diagnosis method |
| CN117030268B (en) * | 2023-10-07 | 2024-01-23 | 太原科技大学 | Rolling bearing fault diagnosis method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115456019A (en) | Rolling bearing fault diagnosis method based on wavelet packet transformation and CEEMDAN | |
| CN102607845B (en) | Bearing fault characteristic extracting method for redundantly lifting wavelet transform based on self-adaptive fitting | |
| CN111089726B (en) | Rolling bearing fault diagnosis method based on optimal dimension singular spectrum decomposition | |
| CN108444709B (en) | Rolling bearing fault diagnosis method combining VMD and FastICA | |
| CN109708891B (en) | Fault diagnosis method for raceway of flexible elliptical bearing | |
| CN110046476B (en) | Ternary binary fractal wavelet sparse diagnosis method for rolling bearing faults | |
| CN109765055B (en) | Rolling bearing fault detection method and system based on EWT, spectrum effective value and KNN | |
| CN110672327A (en) | Asynchronous motor bearing fault diagnosis method based on multilayer noise reduction technology | |
| CN112183259B (en) | Rolling bearing fault diagnosis method based on CEEMD and kurtosis weighted average threshold denoising | |
| CN107506710A (en) | A kind of rolling bearing combined failure extracting method | |
| CN112945546B (en) | Precise diagnosis method for complex faults of gearbox | |
| CN113176092A (en) | Motor bearing fault diagnosis method based on data fusion and improved experience wavelet transform | |
| Jiang et al. | A novel method for self-adaptive feature extraction using scaling crossover characteristics of signals and combining with LS-SVM for multi-fault diagnosis of gearbox | |
| CN112733612A (en) | Cross-domain rotating machinery fault diagnosis model establishing method and application thereof | |
| CN112067297A (en) | Bearing fault feature extraction method | |
| CN117571316A (en) | Composite fault diagnosis method and system | |
| CN115436053A (en) | WPT-SVD-based fault diagnosis method for rolling bearing of aero-engine | |
| CN115859054B (en) | MIC-CEEMDAN-based hydroelectric generating set draft tube pressure pulsation signal denoising method | |
| Yang et al. | Research on Fault Feature Extraction Method Based on FDM‐RobustICA and MOMEDA | |
| CN112747925B (en) | Rolling bearing fault diagnosis method based on composite morphological filtering | |
| CN114136604B (en) | Rotary equipment fault diagnosis method and system based on improved sparse dictionary | |
| CN116086807A (en) | Fault judging method for rolling bearing and related product | |
| CN115687892A (en) | Rolling bearing fault diagnosis method based on improved variational modal decomposition | |
| CN114705427A (en) | Rolling bearing intelligent fault diagnosis method under strong noise environment | |
| CN113065418A (en) | Rolling bearing fault diagnosis method based on SSA-WDCNN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |