CN105547627B - Rotating machinery feature extracting method based on WPT-CEEMD - Google Patents
Rotating machinery feature extracting method based on WPT-CEEMD Download PDFInfo
- Publication number
- CN105547627B CN105547627B CN201610060829.XA CN201610060829A CN105547627B CN 105547627 B CN105547627 B CN 105547627B CN 201610060829 A CN201610060829 A CN 201610060829A CN 105547627 B CN105547627 B CN 105547627B
- Authority
- CN
- China
- Prior art keywords
- signal
- wpt
- ceemd
- frequency
- decomposition
- 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.)
- Active
Links
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
- G01M7/00—Vibration-testing of structures; Shock-testing of structures
- G01M7/02—Vibration-testing by means of a shake table
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses the rotating machinery feature extracting methods based on WPT CEEMD, noise section is eliminated with WPT good partial analysis abilities or useful frequency band is extracted first, CEEMD adaptive decompositions are carried out again, it finally reuses WPT to a small amount of frequency alias part existing after decomposition to be corrected, so as to fulfill the extraction to characteristic frequency signal.WPT is combined by the present invention with CEEMD, the shortcomings of overcoming in existing method for diagnosing faults there are frequency alias, without adaptivity.
Description
Technical field
The invention belongs to mechanical fault diagnosis field, the more particularly to rotating machinery feature based on WPT-CEEMD
Extracting method.
Background technology
With the rapid development of modern industry and science and technology, industry, which has become, weighs what a national science technology developed
Standard.Mechanical equipment is the carrier of industrial development, provides key technology for industrial development, is played increasingly in national economy
Important role.Rotating machinery also increasingly develops towards the direction of enlargement, complication, precise treatment simultaneously, mechanical equipment
Function is more and more, and performance indicator is higher and higher, and composition becomes increasingly complex with structure, will certainly cause what failure occurred in this way
Probability greatly increases.
Rolling bearing is most widely used in rotating machinery and is also one of most flimsy machine components, many mechanical breakdowns
All related with rolling bearing, its work quality has a great impact to the working condition of machinery, therefore to the rolling bearing in work
It is particularly important to carry out fault diagnosis.Since mechanical equipment vibration signal is mostly nonlinear and non local boundary value problem, traditional Fourier point
Analysis has certain limitation unavoidably.Time-Frequency Analysis Method such as wavelet transformation (Wavelet transform, WT), wavelet package transforms
(Wavelet Package Transform, WPT), Hilbert-Huang transform (Hilbert-Huang transform, HHT),
Gabor transformation etc. due to that can provide the local message of signal time domain and frequency domain simultaneously, and obtains wide in mechanical fault diagnosis
General application.
Traditional WT and WPT in actual signal processing procedure due to itself Mallat algorithm the defects of, can be in frequency range point
It cuts place and serious frequency alias phenomenon occurs and without adaptivity, it is bad to be directly used in fault-signal extraction effect.Xi Er
Signal adaptive can be decomposed into limited a intrinsic mode function (Intrinsic Mode by the EMD in Bert Huang
Function, IMF), each IMF components decomposited include the local feature signal of original signal different time scales, but also deposit
In serious modal overlap shortcoming.For this shortcoming, population mean empirical mode decomposition is had also been proposed in the basis of EMD
The improved method of (Ensemble Empirical Mode Decomposition, EEMD) and EEMD --- it is complementary overall flat
Equal empirical mode decomposition (Complementary Ensemble Empirical Mode Decomposition, CEEMD).Two
Kind of method after adaptive decomposition, though effectively inhibit EMD frequency alias defect still remained in low frequency part it is a small amount of
Modal overlap phenomenon.
Invention content
In order to solve the technical issues of above-mentioned background technology proposes, the present invention is intended to provide the rotation based on WPT-CEEMD
Mechanical features extracting method, WPT and CEEMD are combined, and effectively inhibit the mode mixing after decomposing.
In order to realize above-mentioned technical purpose, the technical scheme is that：
Rotating machinery feature extracting method based on WPT-CEEMD, includes the following steps：
(1) characteristic of rotating machines vibration signal is acquired；
(2) FFT transform is carried out to the vibration signal of acquisition, reuses WPT transformation and denoising is carried out to signal；
(3) CEEMD adaptive decompositions are carried out to de-noising treated signal；
(4) frequency signal of IMF components decomposited is extracted, makes the spectrogram of each IMF components；
(5) serious modal overlap is checked for according to the correspondence of frequency and amplitude, i.e., multiple and different frequencies
Detail signal whether be broken down into same IMF components, if there is serious modal overlap, then return to step (3), otherwise
It is transferred to step (6)；
(6) if the signal of same frequency is by adaptive decomposition to adjacent IMF components, to there is the portion of modal overlap
Divide and be modified using WPT transformation, extract characteristic signal.
Further, the detailed process of denoising is carried out in step (2) to signal using WPT transformation：
(a) after selecting a wavelet basis and determining the level decomposed, WAVELET PACKET DECOMPOSITION is carried out to signal；
(b) for given entropy standard, best wavelet packet basis is determined；
(c) threshold value quantizing is carried out to the coefficient of best wavelet packet basis；
(d) wavelet packet coefficient after threshold value quantizing is reconstructed, obtains reconstruction signal.
Further, the detailed process of step (3)：
(I) add in the signal x (t) after denoising respectively one group of equal length, given standard deviation, symbol it is positive and negative
Opposite normal distribution white noise, so as to form two new signals；
(II) carries out EMD decomposition respectively to two signals after adding in white noise, obtains its respective IMF component；
(III) repeats step (I), (II) n times, and requirement adds in new normal state point every time according to the decomposition frequency n of setting
Cloth white noise；
The obtained IMF components of (IV) decomposition amount to 2n groups, after this 2n group IMF component is added divided by 2n, it are asked to be averaged, is obtained
To 1 group of IMF component.
Further, the amplitude of the normal distribution white noise is that the standard deviation of signal x (t) is multiplied by coefficient f, wherein 0<f<
1。
Further, when decomposition frequency n takes 100, the value range of coefficient f is 0.01<f<0.5.
Further, it in step (5), if there is serious modal overlap, at return to step (3), needs again
It determines to decompose frequency n.
The advantageous effect brought using above-mentioned technical proposal：
The present invention decomposes signal using CEEMD, the modal overlap problem that traditional EMD is effectively inhibited to decompose, and
The part there are modal overlap after decomposition is modified again with WPT good partial analysis abilities, so as to reach to thin
Save the accurate extraction of signal.
Description of the drawings
Fig. 1 is the flow chart of the present invention；
Fig. 2 includes (a), (b), (c) 3 width figure, respectively signal y1, y2 and emulation signal s (t) oscillogram；
Fig. 3 includes (a), (b) 2 width figure, IMF1, IMF2 oscillogram that respectively s (t) is decomposed through EMD；
Fig. 4 includes (a), (b), (c), (d) 4 width figure, IMF1, IMF2 that respectively s (t) is decomposed through CEEMD, IMF3,
IMF4 oscillograms；
Fig. 5 is the vibration signal time-domain diagram that actual acquisition obtains；
Fig. 6 is the vibration signals spectrograph figure that actual acquisition obtains；
Fig. 7 is to carry out the pretreated spectrograms of WPT to the vibration signal that actual acquisition obtains；
Fig. 8 includes (a), (b), (c) 3 width figure, what the vibration signal that respectively actual acquisition obtains was decomposed through CEEMD
The time-domain diagram of IMF6, IMF7, IMF8；
Fig. 9 includes (a), (b), (c) 3 width figure, what the vibration signal that respectively actual acquisition obtains was decomposed through CEEMD
The spectrogram of IMF6, IMF7, IMF8；
Figure 10 includes (a), (b) 2 width figure, the time-domain diagram and spectrogram of working frequency component that WPT is extracted after correcting.
Specific embodiment
Below with reference to attached drawing, technical scheme of the present invention is described in detail.
The flow chart of the present invention as shown in Figure 1, the rotating machinery feature extracting method based on WPT-CEEMD, including following
Step：
(1) characteristic of rotating machines vibration signal is acquired；
(2) FFT transform is carried out to the vibration signal of acquisition, reuses WPT transformation and denoising is carried out to signal, inhibit letter
Nonuseable part in number, enhances useful part in signal, and detailed process is as follows：
(a) after selecting a wavelet basis and determining the level decomposed, WAVELET PACKET DECOMPOSITION, wavelet packet first layer are carried out to signal
It decomposes and signal is divided into two frequency ranges of height by signal with the 1/2 of sample frequency for midpoint, first layer is decomposed in second layer decomposition
Two frequency ranges are decomposed to obtain four frequency ranges respectively using 1/4 and 3/4 sample frequency as midpoint again, and so on, Ke Yigen
Factually border signal extraction need select Decomposition order；
(b) for given entropy standard, best wavelet packet basis is determined；
(c) threshold value quantizing is carried out to the coefficient of best wavelet packet basis；
(d) wavelet packet coefficient after threshold value quantizing is reconstructed, obtains reconstruction signal.
(3) CEEMD adaptive decompositions are carried out to de-noising treated signal, detailed process is as follows：
(I) add in the signal x (t) after denoising respectively one group of equal length, given standard deviation, symbol it is positive and negative
Opposite normal distribution white noise, so as to form two new signals；
(II) carries out EMD decomposition respectively to two signals after adding in white noise, obtains its respective IMF component；
(III) repeats step (I), (II) n times, and requirement adds in new normal state point every time according to the decomposition frequency n of setting
Cloth white noise；
The obtained IMF components of (IV) decomposition amount to 2n groups, after this 2n group IMF component is added divided by 2n, it are asked to be averaged, is obtained
To 1 group of IMF component.
The amplitude of normal distribution white noise is that the standard deviation of signal x (t) is multiplied by coefficient f, wherein 0<f<1, when decomposition frequency n
100 are taken, the value range of coefficient f is 0.01<f<0.5.In practical applications, the coefficient of the amplitude of white noise is added according to pre- place
The size of signal institute Noise is adjusted after reason, the white noise width added when the bigger CEEMD of institute's Noise is decomposed in signal
The coefficient of value is bigger；
(4) frequency signal of IMF components decomposited is extracted, makes the spectrogram of each IMF components；
(5) serious modal overlap is checked for according to the correspondence of frequency and amplitude, i.e., multiple and different frequencies
Detail signal whether be broken down into same IMF components, if there is serious modal overlap, then return to step (3), at this time
It needs to redefine decomposition frequency n, is otherwise transferred to step (6)；
(6) if the signal (especially in low-band signal) of same frequency is by adaptive decomposition to adjacent IMF components,
It is then modified to there is the part of modal overlap using WPT transformation, the signal of the different frequency in same IMF components is carried out
Separation, then will be added in the signal time domain of identical frequency, you can extract required frequency signal.
As shown in Fig. 2, it is by low frequency sinusoidal component y1 (t) and high frequency intermittent signal y2 to emulate a signal s (t) first
(t) i.e. s (t)=y1+y2 is formed by stacking, wherein
Y1=sin (20 π * t) 0≤t≤1 (1)
Emulation signal sampling frequencies are 1kHz, sampling time 1s.EMD and CEEMD is respectively adopted to emulation signal s (t)
Decomposed, it is 50 times to decompose number, and addition white noise acoustic amplitude standard deviation be to emulate 0.1 times of signal, as a result respectively such as Fig. 3,
Shown in Fig. 4, as seen from Figure 3, the very big signal of two frequency phase-differences is decomposed by EMD in same component i.e. IMF1, is occurred
Modal overlap phenomenon.Fig. 4 can be seen that the IMF1 that CEEMD is decomposed extracts high frequency intermittent signal well and amplitude only has
The change of very little；IMF2, IMF3 contain two a small amount of source signals, a small amount of modal overlap phenomenon occur；IMF4 is extracted
Low Frequency Sine Signals.97.12% can be reached by doing correlation analysis to IMF1 and y2 (t).Effectively press down so as to demonstrate CEEMD
The modal overlap shortcoming of EMD is made.
Again by taking actual vibration signal as an example, using the dynamic signal analyzer of OROS R3X systems to horizontal spiral centrifuge
The vibration signal of (abbreviation decanter centrifuge) in the working speed operation phase carries out test experiments.Sample frequency is 12.8kHz, is adopted
The sample time is 20s, and decanter centrifuge relevant parameter is as shown in table 1.Fig. 5 is the time-domain signal collected.
Table 1
In order to be extracted to decanter centrifuge working frequency i.e. 1 order component 55Hz, selected first to collected vibration
Signal carries out FFT transform and makes its spectrogram as shown in fig. 6, carrying out WPT pretreatments according to the spectrogram of sample frequency and signal.
The pretreated signal spectrum figures of WPT are as shown in Figure 7, it can be seen that and high frequency noise components are eliminated after pretreatment,
Useful frequency band is extracted, and it is corresponding with amplitude to obtain from spectrogram each frequency content for next step CEEMD decomposition
Relationship.
Pretreated signal is subjected to CEEMD adaptive decompositions, by test of many times, selection, which is decomposed, repeats step time
Number.It is 50 to take and decompose frequency n, and the amplitude standard deviation of the white noise of addition is pretreated 0.1 times of WPT.Since CEEMD is adaptive
IMF components after should decomposing are to be arranged in order from high to low according to frequency band, Fig. 8, Fig. 9 decompose respectively after IMF6,
The time-domain diagram and spectrogram of tri- components of IMF7, IMF8.
Frequency and amplitude correspondence in the spectrogram of IMF components, carry out with pretreated signal spectrum Fig. 7
Comparison, it can be seen that after CEEMD adaptive decompositions, still there are a small amount of modal overlaps in the adjacent IMF components of low frequency part
Part, the signal of 55Hz are decomposed by adaptive in two components of IMF7, IMF8.
It is right using the partial analysis ability of WPT to existing a small amount of modal overlap part again with WPT processing
Frequency band in IMF7 is decomposed again, and 55Hz is detached, then the signal of the 55Hz isolated is added to 2 frequency multiplication 110Hz
So as to reach the extraction to 1 order component in IMF8.The time-domain spectral figure of the 55Hz components extracted is as shown in Figure 10.
Although can be seen that CEEMD adaptive decompositions by the decomposition result with emulation experiment and actual vibration signal has
Effect improves EMD containing the modal overlap problem in discontinuous point signal, but a small amount of frequency is still remained in low frequency part
Aliasing recombines WPT for this problem and is corrected, and the frequency as obtained by with calculating is compared, good to realize
Bearing failure diagnosis.
Above example is merely illustrative of the invention's technical idea, it is impossible to protection scope of the present invention is limited with this, it is every
According to technological thought proposed by the present invention, any change done on the basis of technical solution each falls within the scope of the present invention
Within.
Claims (6)
1. the rotating machinery feature extracting method based on WPT-CEEMD, which is characterized in that include the following steps：
(1) characteristic of rotating machines vibration signal is acquired；
(2) FFT transform is carried out to the vibration signal of acquisition, reuses WPT transformation and denoising is carried out to signal；
(3) CEEMD adaptive decompositions are carried out to de-noising treated signal；
(4) frequency signal of IMF components decomposited is extracted, makes the spectrogram of each IMF components；
(5) if the detail signal of multiple and different frequencies is broken down into same IMF components, illustrate that there are serious mode to mix
Folded, return to step (3), are otherwise transferred to step (6) at this time；
(6) if the signal of same frequency is by adaptive decomposition to adjacent IMF components, to there is the part of modal overlap profit
After being modified with WPT transformation, characteristic signal is extracted.
2. the rotating machinery feature extracting method based on WPT-CEEMD according to claim 1, which is characterized in that step (2)
The middle detailed process for carrying out denoising to signal using WPT transformation：
(a) after selecting a wavelet basis and determining the level decomposed, WAVELET PACKET DECOMPOSITION is carried out to signal；
(b) for given entropy standard, best wavelet packet basis is determined；
(c) threshold value quantizing is carried out to the coefficient of best wavelet packet basis；
(d) coefficient of the best wavelet packet basis after threshold value quantizing is reconstructed, obtains reconstruction signal.
3. the rotating machinery feature extracting method based on WPT-CEEMD according to claim 1, which is characterized in that step (3)
Detailed process：
(I) add in the signal x (t) after denoising respectively one group of equal length, given standard deviation, symbol it is positive and negative opposite
Normal distribution white noise, so as to form two new signals；
(II) carries out EMD decomposition respectively to two signals after adding in white noise, obtains its respective IMF component；
(III) repeats step (I), (II) n times according to the decomposition frequency n of setting, and to add in new normal distribution every time white for requirement
Noise；
The obtained IMF components of (IV) decomposition amount to 2n groups, after this 2n group IMF component is added divided by 2n, it are asked to be averaged, obtains 1
Group IMF components.
4. the rotating machinery feature extracting method based on WPT-CEEMD according to claim 3, it is characterised in that：It is described just
The standard deviation that the amplitude of state distribution white noise is signal x (t) is multiplied by coefficient f, wherein 0<f<1.
5. the rotating machinery feature extracting method based on WPT-CEEMD according to claim 4, it is characterised in that：Work as decomposition
When frequency n takes 100, the value range of coefficient f is 0.01<f<0.5.
6. the rotating machinery feature extracting method based on WPT-CEEMD according to claim 3, it is characterised in that：In step
(5) it in, if there is serious modal overlap, at return to step (3), needs to redefine decomposition frequency n.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610060829.XA CN105547627B (en) | 2016-01-28 | 2016-01-28 | Rotating machinery feature extracting method based on WPT-CEEMD |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610060829.XA CN105547627B (en) | 2016-01-28 | 2016-01-28 | Rotating machinery feature extracting method based on WPT-CEEMD |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105547627A CN105547627A (en) | 2016-05-04 |
| CN105547627B true CN105547627B (en) | 2018-06-12 |
Family
ID=55826971
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610060829.XA Active CN105547627B (en) | 2016-01-28 | 2016-01-28 | Rotating machinery feature extracting method based on WPT-CEEMD |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105547627B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105891707A (en) * | 2016-05-05 | 2016-08-24 | 河北工业大学 | Opening-closing fault diagnosis method for air circuit breaker based on vibration signals |
| CN106568607A (en) * | 2016-11-04 | 2017-04-19 | 东南大学 | Rub-impact sound emission fault diagnosis method based on empirical wavelet transformation |
| CN106568589A (en) * | 2016-11-04 | 2017-04-19 | 东南大学 | Rubbing acoustic emission denoise method based on empirical wavelet transform |
| CN108229382A (en) * | 2017-12-29 | 2018-06-29 | 广州供电局有限公司 | Vibration signal characteristics extracting method, device, storage medium and computer equipment |
| CN111337767B (en) * | 2020-02-25 | 2023-07-18 | 河北工业大学 | Fault analysis method for resonant wave speed reducer |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102394658A (en) * | 2011-10-16 | 2012-03-28 | 西南科技大学 | Composite compression method oriented to mechanical vibration signal |
| CN103091096A (en) * | 2013-01-23 | 2013-05-08 | 北京信息科技大学 | Extraction method for early failure sensitive characteristics based on ensemble empirical mode decomposition (EEMD) and wavelet packet transform |
| CN104198183A (en) * | 2014-09-17 | 2014-12-10 | 重庆大学 | Wind turbine drive chain vibration noise suppression and incipient fault feature extraction method |
-
2016
- 2016-01-28 CN CN201610060829.XA patent/CN105547627B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN105547627A (en) | 2016-05-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105547627B (en) | Rotating machinery feature extracting method based on WPT-CEEMD | |
| Wang et al. | Subband averaging kurtogram with dual-tree complex wavelet packet transform for rotating machinery fault diagnosis | |
| Feng et al. | Time-varying demodulation analysis for rolling bearing fault diagnosis under variable speed conditions | |
| CN110186682B (en) | Rolling bearing fault diagnosis method based on fractional order variation modal decomposition | |
| Li et al. | Independence-oriented VMD to identify fault feature for wheel set bearing fault diagnosis of high speed locomotive | |
| Chen et al. | High-accuracy fault feature extraction for rolling bearings under time-varying speed conditions using an iterative envelope-tracking filter | |
| Wang et al. | A demodulation method based on improved local mean decomposition and its application in rub-impact fault diagnosis | |
| Wang et al. | A novel procedure for diagnosing multiple faults in rotating machinery | |
| CN108168891A (en) | The extracting method and equipment of rolling bearing Weak fault signal characteristic | |
| Han et al. | Second-order synchroextracting wavelet transform for nonstationary signal analysis of rotating machinery | |
| Tong et al. | A fault diagnosis approach for rolling element bearings based on RSGWPT-LCD bilayer screening and extreme learning machine | |
| CN113340640B (en) | Rotating machinery axis track purification method, device, equipment and storage medium | |
| Lopatinskaia et al. | Monitoring varying speed machinery vibrations—II recursive filters and angle domain | |
| CN112485028B (en) | Feature spectrum extraction method of vibration signal and mechanical fault diagnosis analysis method | |
| CN110398364A (en) | Epicyclic gearbox method for diagnosing faults based on resonance sparse decomposition and FastICA algorithm | |
| Wang et al. | A piecewise hybrid stochastic resonance method for early fault detection of roller bearings | |
| Sharma | Gear fault detection based on instantaneous frequency estimation using variational mode decomposition and permutation entropy under real speed scenarios | |
| Lv et al. | Longitudinal synchroextracting transform: A useful tool for characterizing signals with strong frequency modulation and application to machine fault diagnosis | |
| Zhang et al. | Bearing fault diagnosis based on iterative 1.5-dimensional spectral kurtosis | |
| CN113761466A (en) | Method and device for constructing vibration signal order ratio spectrum of rotary machine | |
| Li et al. | Revision of bearing fault characteristic spectrum using LMD and interpolation correction algorithm | |
| Li et al. | Fault feature extraction of rolling bearing based on an improved cyclical spectrum density method | |
| Li et al. | Theoretical analysis of adaptive harmonic window and its application in frequency extraction of vibration signal | |
| CN112345248B (en) | Fault diagnosis method and device for rolling bearing | |
| Hai et al. | Application of Feature Extraction Method Based on Empirical Fourier Decomposition in Rotor Rub-impact Fault |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP02 | Change in the address of a patent holder |
Address after: Room 420, block C, Kechuang headquarters building, No. 320, pubin Road, Jiangpu street, Pukou District, Nanjing, Jiangsu Patentee after: NANJING University OF INFORMATION SCIENCE & TECHNOLOGY Address before: 210044 Nanjing City, Pukou Province, Nanjing Road, No. 219, No. six, No. Patentee before: NANJING University OF INFORMATION SCIENCE & TECHNOLOGY |
|
| CP02 | Change in the address of a patent holder |