CN110940522A - Bearing fault periodic pulse sparse separation and diagnosis method under strong background noise - Google Patents

Bearing fault periodic pulse sparse separation and diagnosis method under strong background noise Download PDF

Info

Publication number
CN110940522A
CN110940522A CN201910279626.3A CN201910279626A CN110940522A CN 110940522 A CN110940522 A CN 110940522A CN 201910279626 A CN201910279626 A CN 201910279626A CN 110940522 A CN110940522 A CN 110940522A
Authority
CN
China
Prior art keywords
fault
periodic
bearing
vibration acceleration
vertical direction
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.)
Withdrawn
Application number
CN201910279626.3A
Other languages
Chinese (zh)
Inventor
李庆
梁越昇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Donghua University
National Dong Hwa University
Original Assignee
Donghua University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Donghua University filed Critical Donghua University
Priority to CN201910279626.3A priority Critical patent/CN110940522A/en
Publication of CN110940522A publication Critical patent/CN110940522A/en
Withdrawn legal-status Critical Current

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)

Abstract

The invention relates to a bearing fault periodic pulse sparse separation and diagnosis method under strong background noise, which is characterized by comprising the following steps of: respectively installing acceleration sensors in the horizontal direction and the vertical direction of a bearing seat to be detected, and acquiring a vibration acceleration signal in the horizontal direction and a vibration acceleration signal in the vertical direction of the bearing; and establishing a target cost function, and respectively calculating the horizontal direction periodic fault pulse signals and the vertical direction periodic fault pulse signals of the horizontal direction vibration acceleration signals and the vertical direction vibration acceleration signals through iteration of an alternating direction multiplier method. The invention can successfully extract the period information of the hidden unknown periodic pulse in the noisy signal without any priori knowledge, and has high pulse separation accuracy and strong stability. The invention can greatly reduce the interference of background working conditions and system structure noise, and the extracted fault periodic transient pulse sequence is clear and has no aliasing.

Description

Bearing fault periodic pulse sparse separation and diagnosis method under strong background noise
Technical Field
The invention relates to a mechanical fault diagnosis and signal processing method, in particular to a bearing fault periodic pulse sparse separation and diagnosis method under strong background noise.
Background
The bearing is used as a key part of rotary mechanical equipment and widely applied to the fields of various commercial and industrial engineering, such as wind power generation, automobile gearboxes, high-speed rails, ships, aerospace engines and the like. The health state monitoring of the bearing during the operation of the equipment is related to the performance, safety and service life of the whole system, the degradation state of the bearing in service is accurately and timely monitored, and the method has extremely important research significance on the reliable operation of the system, the production of enterprises and the life safety of operators.
Generally, when a bearing element has local faults, such as inner ring pitting, outer ring peeling, ball abrasion and the like, a series of periodic fault pulse impact signals are excited along with the operation of a system, and the periodic fault pulse signals are key information for monitoring the health state of the bearing. However, in practical engineering, due to the interference of external noise, the influence of system structure, and the like, useful fault pulse signals are often submerged in the external noise, and therefore, it is a difficult point in the field of mechanical fault diagnosis in recent years to effectively separate periodic fault pulse signals from actual observation signals.
Disclosure of Invention
The purpose of the invention is: the periodic fault pulse signal is effectively isolated from the actual observed signal.
In order to achieve the purpose, the technical scheme of the invention is to provide a bearing fault periodic pulse sparse separation and diagnosis method under strong background noise, which is characterized by comprising the following steps of:
step 1, respectively installing acceleration sensors in the horizontal direction and the vertical direction of a bearing seat to be detected, and collecting a horizontal direction vibration acceleration signal and a vertical direction vibration acceleration signal of the bearing
Step 2, establishing a target cost function:
Figure RE-GDA0002366358540000011
in the formula (I), the compound is shown in the specification,
Figure RE-GDA0002366358540000012
representing a wavelet transform coefficient to be estimated; ω represents a wavelet transform coefficient; f (ω) represents the objective cost function; y represents the vibration acceleration signal in the horizontal direction or the vertical direction obtained in step 1A vibration acceleration signal; lambda [ alpha ]jRepresenting the regularization parameter at the time scale j; continuous wavelet transform of transform coefficients x: omegaj，k＝Wj，kx, in the formula, Wj，kThe wavelet transform is performed on a translation scale j and a time scale k, y is Ax + w, A represents a matrix with Toeplitz, and w represents background noise or interference components; a isjRepresenting a penalty function scale coefficient; phi (omega)j，k；aj) β denotes regularization parameters, D denotes a first order differential matrix;
step 3, inputting the horizontal direction vibration acceleration signals and the vertical direction vibration acceleration signals obtained in the step 1 into the target cost function established in the step 2 respectively, and obtaining the horizontal direction vibration acceleration signals and the vertical direction vibration acceleration signals through iterative calculation of an alternating direction multiplier method
Figure RE-GDA0002366358540000021
Obtaining a horizontal direction periodic fault pulse signal and a vertical direction periodic fault pulse signal through wavelet inverse transformation calculation, and defining the periodic fault pulse signals as
Figure RE-GDA0002366358540000022
Then there is
Figure RE-GDA0002366358540000023
Preferably, after the step 3, the method further comprises:
and carrying out envelope demodulation on the horizontal direction periodic fault pulse signal and the vertical direction periodic fault pulse signal by using a Hilbert envelope demodulation method, extracting bearing fault characteristic frequency and frequency multiplication thereof, comparing the extracted bearing fault characteristic frequency with theoretical fault characteristic frequency, and identifying the fault position of the bearing.
The method provided by the invention can effectively filter the interference component and the serious background noise, and simultaneously separates the periodic fault pulse signal hidden in the noise.
Compared with the prior art, the invention has the following advantages:
compared with the traditional signal filtering method, the method has the advantages that any priori knowledge is not needed to be known in advance, the period information of the hidden unknown periodic pulses in the noisy signals can be successfully extracted, and the pulse separation accuracy is high and the stability is high.
Secondly, the regularization target cost function is constructed by utilizing a non-convex penalty Daubechies wavelet function and a total variation algorithm, and the target cost function has strict convexity.
The method can greatly reduce the interference of background working conditions and system structure noise, the extracted fault periodic instantaneous pulse sequence is clear and has no aliasing, the amplitude of the fault characteristic frequency obtained by demodulation is high, the fault harmonic component is obvious, the method is suitable for real-time fault routing inspection of other rotating mechanical equipment such as a bearing or a gear box and the like, and sudden faults are avoided, and an important theoretical basis can be provided for the health management of enterprise equipment.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
fig. 2 is a schematic view of an experimental platform according to an embodiment of the present invention, in which 1 denotes a non-driving end of an input shaft, 2 denotes a faulty bearing, 3 denotes a non-driving end of an output shaft, 4 denotes a driving end of the input shaft, and 5 denotes a driving end of the output shaft;
FIG. 3(a) is a time domain waveform of an original vibration acceleration signal acquired in a horizontal direction;
fig. 3(b) is an envelope spectrum of an original vibration acceleration signal acquired in a vertical direction;
FIG. 3(c) is a time domain waveform of the original vibration acceleration signal in the horizontal direction;
FIG. 3(d) is an envelope spectrum of the original vibration acceleration signal in the vertical direction;
FIG. 4(a) is a horizontal direction fault periodic pulse obtained using the method of the present invention;
FIG. 4(b) is an envelope spectrum of a horizontal fault periodic pulse signal;
FIG. 4(c) is a vertical direction fault periodic pulse obtained using the method of the present invention;
fig. 4(d) is an envelope spectrum of a vertical direction fault periodic pulse signal.
Detailed Description
The invention will be further illustrated with reference to the following specific examples. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention may be made by those skilled in the art after reading the teaching of the present invention, and such equivalents may fall within the scope of the present invention as defined in the appended claims.
The invention provides a bearing fault periodic pulse sparse separation and diagnosis method under strong background noise, which comprises the following steps:
step 1, respectively installing acceleration sensors in the horizontal direction and the vertical direction of a bearing seat to be detected, and collecting vibration acceleration signals of the bearing.
Step 2, separating out the fault periodic transient pulse signal by using the method of the invention, and removing external interference noise, which comprises the following steps:
1) typically, the vibration sensor collects a device fault observation signal
Figure RE-GDA0002366358540000031
Can be expressed as y ═ x0+ w is Ax + w, wherein,
Figure RE-GDA0002366358540000032
for the purpose of background noise or interference components,
Figure RE-GDA0002366358540000033
in order to be a periodic fault pulse signal,
Figure RE-GDA0002366358540000034
for transforming coefficients, matrices
Figure RE-GDA0002366358540000035
With Toeplitz matrix. The conventional method of minimizing the L1 norm utilizes a linear least squares model to estimate the periodic fault pulse signal, i.e.
Figure RE-GDA0002366358540000036
In the formula (I), the compound is shown in the specification,
Figure RE-GDA0002366358540000037
representing a periodic fault pulse signal, f (x) is a target cost function,
Figure RE-GDA0002366358540000038
is a secondary data fidelity item, wherein
Figure RE-GDA0002366358540000039
Is a penalty function (i.e., L1-norm), λ0And λ1A regularization parameter is represented as a function of,
Figure RE-GDA00023663585400000310
is a first order differential matrix, i.e.
Figure RE-GDA0002366358540000041
The matrix D determines the sparsity of the fault pulse signal. The traditional minimization L1-norm method can be solved through a full-variational model and a soft threshold algorithm, namely
Figure RE-GDA0002366358540000042
Wherein Soft (·,) is a Soft threshold function of
Figure RE-GDA0002366358540000043
Tvd (·,. cndot.) is a fully-variational model with the expression
Figure RE-GDA0002366358540000044
Where Prox (-) is the neighbor of the then-current signal y.
2) Given a signal x, its continuous wavelet transform
Figure RE-GDA0002366358540000045
Where i and j are the mother wavelet translation factor and time factor, respectively. The continuous wavelet transform of the signal x can be simplified to ωi，j＝Wi，jx, wherein Wi，jFor wavelet transformation at a translation scale i and a time scale j, omegai，jRepresenting translationWavelet transform coefficients at scale i and time scale j. According to the shift invariant property and Parseval theorem WTThe wavelet transform coefficient W can be modeled by
Figure RE-GDA0002366358540000046
And calculating to obtain the final product of the formula,
Figure RE-GDA0002366358540000047
representing wavelet transform coefficients, λ, to be estimatedjAnd β are regularization parameters, ajRepresenting the penalty function scaling factor. Under the translation scale i and the time scale j, the two norms of the wavelet transform coefficient omega satisfy
Figure RE-GDA0002366358540000048
Final periodic fault pulse signal
Figure RE-GDA0002366358540000049
Can be calculated by inverse wavelet transform, i.e.
Figure RE-GDA00023663585400000410
3) Established objective cost function
Figure RE-GDA00023663585400000411
In the model, where phi (x; a) is phi (omega)jK, k; a) and a is not less than 0 and meets the following 7 conditions:
(I) penalty functions phi (x; a) in
Figure RE-GDA00023663585400000412
Continuously; (II) the penalty function phi (x; a) is in
Figure RE-GDA00023663585400000413
Upper and lower order conductibility; (III) the penalty function phi (x; a) is a symmetric function, i.e., phi (-x; a) is phi (x; a); (IV) φ' (x; a) > 0,
Figure RE-GDA00023663585400000414
(V)φ″(x；a)≤0，
Figure RE-GDA00023663585400000415
(VI)φ′(0+；a)＝1；(VII)
Figure RE-GDA00023663585400000416
assuming that the non-convex penalty function φ (x; a) satisfies the above-mentioned 7 conditions, for each scale j, a ≦ a is satisfied if 0j＜1/λjThen, a target cost function is proposed
Figure RE-GDA0002366358540000051
Has strict convexity.
4) Established objective cost function
Figure RE-GDA0002366358540000052
Has strict convexity, and the process is proved as follows:
the proposed objective cost function may be further expressed as
Figure RE-GDA0002366358540000053
Due to the last item L1-norm β | | | DWTω||1Is a convex function, so to prove that the objective cost function has strict convexity, it is only necessary to prove that the first term in the formula F (ω) is a convex function. Optimizing the model according to linear least squares
Figure RE-GDA0002366358540000054
Which can be expressed as
Figure RE-GDA0002366358540000055
Where phi (x; a) is a penalty function. The function θ (y, λ, a) has a strict convexity if and only if φ' (0)+)≥φ′(0-) That is, 1+ λ φ "(x) > 0, for allx ≠ 0 is obtained
Figure RE-GDA0002366358540000056
For standard penalty functions, e.g. arctangent penalty function
Figure RE-GDA0002366358540000057
And logarithmic penalty function
Figure RE-GDA0002366358540000058
Due to the fact that
Figure RE-GDA0002366358540000059
Its second derivative satisfies
Figure RE-GDA00023663585400000510
Then there is
Figure RE-GDA00023663585400000511
Further, due to
Figure RE-GDA00023663585400000512
The term can be expressed as θ ([ W ]y]j,k；λj，aj) When it is satisfied
Figure RE-GDA00023663585400000513
θ([Wy]j，k；λj，aj) Is a convex function. Therefore when the coefficient a isjSatisfy the requirement of
Figure RE-GDA00023663585400000514
Proposed objective cost function
Figure RE-GDA00023663585400000515
Has strict convexity.
5) The proposed objective cost function, the solving algorithm of which can be obtained by iterative calculation through an alternative direction multiplier method, specifically:
for any j, assume parameter ajSatisfy the requirement of
Figure RE-GDA0002366358540000061
For evaluating periodic fault pulse signals
Figure RE-GDA0002366358540000062
Proposed objective cost functionCan be decomposed into
Figure RE-GDA0002366358540000063
Wherein
Figure RE-GDA0002366358540000064
g2(u)＝β||DWTu||1To decomposition formula g1(omega) and g2(u) introducing a Lagrangian multiplier mu > 0 to obtain
Figure RE-GDA0002366358540000065
According to the principle of the alternative direction multiplier method, the decomposition formula can be:
Figure RE-GDA0002366358540000066
Figure RE-GDA0002366358540000067
d ═ d- (u- ω). When it is satisfied with
Figure RE-GDA0002366358540000068
It can be seen that
Figure RE-GDA0002366358540000069
And
Figure RE-GDA00023663585400000610
all have strict convexity.
For arbitrary j and k, equation
Figure RE-GDA00023663585400000611
Can be converted into a solution model
Figure RE-GDA00023663585400000612
Wherein p ═ y [ Wy + mu (u-d)]/(. mu. + 1). Therefore, using a soft threshold function and a full variational algorithm, equation
Figure RE-GDA00023663585400000613
Is solved as
Figure RE-GDA00023663585400000614
For arbitrary j and k, equation
Figure RE-GDA00023663585400000615
Can be calculated by a neighbor operator method, and the definition function q (v) is
Figure RE-GDA00023663585400000616
Consider the literature [ S.F.Kuang, H.Y.Chao, Q.Li, Matrix composition with a confined nuclear norm via a patterned simulation, neuro-molding.316 (2018)190-]Proposed semi-orthogonal linear transformation of neighborhood operators if
Figure RE-GDA00023663585400000617
LL＝vI，v＞0，
Figure RE-GDA00023663585400000618
Is provided with
Figure RE-GDA0002366358540000071
According to the theorem W of Daubechies waveletsTW ═ I, formula proxf(x) Can be changed into
Figure RE-GDA0002366358540000072
Wherein
Figure RE-GDA0002366358540000073
To obtain
Figure RE-GDA0002366358540000074
Wherein
Figure RE-GDA0002366358540000075
Thus, the formula
Figure RE-GDA0002366358540000076
Can be further expressed as
Figure RE-GDA0002366358540000077
6) According to the step 5, the iterative solution step of the proposed objective cost function can be summarized as the following algorithm:
Figure RE-GDA0002366358540000078
7) the proposed objective cost function, the function model parameters include: coefficient ajβ, and a regularization parameter λjThe three parameter method settings are as follows: satisfies a 0. ltoreq. a in accordance with strict convexity of the objective cost functionj＜1/λjWithout setting the coefficient ajSatisfies aj＝0.97/λjFor coefficient β and regularization parameter λjNote that when β is 0 and λjNot equal to 0, the target cost function is degenerated into a single wavelet de-noising model, when β not equal to 0 and lambdajThe objective cost function degenerates to a single fully-variant denoising model, 0. According to the literature [ l.dumbgen, a.kovac.extensions of smoothening via tau strings.electron.j.statist.3(2009)41-75.]Coefficient β with a regularization parameter λjCan be arranged as
Figure RE-GDA0002366358540000079
β＝(1-η)βTVRWherein, in the step (A),
Figure RE-GDA00023663585400000710
for wavelet de-noising coefficients, βTVRThe weight coefficient satisfies 0 < η < 1 for the fully variant de-noising coefficient, the weight is set to 0.9 < η < 1 for maximum reduction of noise interference, η is not set to 0.96, according to the Daubechies wavelet Parseval theorem condition WTW is equal to I, and the wavelet denoising coefficient satisfies
Figure RE-GDA0002366358540000081
According to the literature [ l. Dumbgen, a. kovac. extensions of smoothening via tau strings.electron. j. statist.3(2009)41-75.]In conclusion, the total variation denoising coefficient is set to
Figure RE-GDA0002366358540000082
Where σ is the standard deviation of the background noiseAnd N is the number of sample points. Thus having λj＝2.5ησ/2j/2，
Figure RE-GDA0002366358540000083
In practical engineering application, the standard deviation σ of the background noise can be calculated by the observation signal and the healthy vibration signal acquired under the same working environment. However, in many cases, the healthy vibration signals of the equipment are not collected in advance or are difficult to obtain, and in such cases, the healthy vibration signals can be estimated by using the principle of the absolute median difference of the wavelet denoising algorithm, namely
Figure RE-GDA0002366358540000084
Wherein, MAD (y) is the Median Absolute Difference (MAD) with few observed signals, which is expressed as: mad (y) ═ mean [ | yi-median(y)|]，i＝1，2，...，N。
And 3, carrying out envelope demodulation on the periodic instantaneous pulse signal by using a Hilbert envelope demodulation method, extracting the bearing fault characteristic frequency and frequency multiplication thereof, comparing the extracted bearing fault characteristic frequency with the theoretical fault characteristic frequency, and identifying the position of the bearing fault.
The invention is further illustrated below with reference to specific experimental data:
the invention provides a bearing fault periodic pulse sparse separation and diagnosis method under strong background noise, which comprises the following steps:
1) acceleration sensors are installed in the horizontal direction and the vertical direction of a fault bearing seat of the speed reducer, original vibration signals of the fault bearing in the horizontal direction and the vertical direction are collected, and a schematic diagram of an experiment platform related to the specific embodiment of the invention is shown in fig. 2.
The invention utilizes the outer ring fault of the tapered roller bearing of the input shaft (non-driving end) in the four-stage speed reducer gearbox to verify the effectiveness of the proposed method.
Before the vibration acceleration signal is collected, a groove with the length being about 30% of the width of the tooth root is machined on the outer ring of the detected tapered roller bearing (bearing model FAG-32212-A) by utilizing a linear cutting machining technology. The experimental sampling frequency is 5120Hz, the rotating speed is 1000rpm (16.67Hz), and the sampling sample length is5120 points, according to a bearing outer ring fault frequency formula fBPO＝1/2×N×[1-d/D×cosα]Wherein N is the number of the balls, D is the inner diameter of the bearing, D is the outer diameter of the bearing, α is the contact angle, the theoretical fault frequency of the outer ring of the tapered roller bearing can be calculated to be 118.8Hz, and the detailed geometric parameters of the tested bearing are shown in Table 1.
TABLE 1 Experimental bearing geometry parameters tested
Figure RE-GDA0002366358540000091
2) And randomly selecting a group of vibration acceleration signals in the horizontal direction and the vertical direction as signals to be analyzed. Fig. 3(a) is a time domain waveform of an original vibration acceleration signal in a horizontal direction, and fig. 3(b) is a Hilbert envelope spectrum of the vibration acceleration signal in the horizontal direction; fig. 3(c) is a time domain waveform of the original vibration acceleration signal in the vertical direction, and fig. 3(d) is a Hilbert envelope spectrum of the vibration acceleration signal in the vertical direction. It can be seen from the time domain signal waveform that the fault pulse is submerged in strong external interference noise, transient and periodic cycle characteristics of the fault pulse cannot be observed, and the extraction of the fault frequency is affected by the noise interference frequency distributed seriously around the bearing fault frequency (such as 119.4Hz, 238.8Hz, 120Hz and 239.4Hz) obtained in fig. 3(b) and fig. 3 (d).
3) The original vibration signals in the horizontal direction and the vertical direction are processed by the proposed sparse Daubechies wavelet pulse separation method, and the parameters of the target cost function are set as follows, namely a coefficient β is set to be β ═ 1- η (β)TVRWherein
Figure RE-GDA0002366358540000092
Regularization parameter λjIs arranged as
Figure RE-GDA0002366358540000093
Wherein
Figure RE-GDA0002366358540000094
The weighting coefficient is set to η -0.96, the wavelet transform scale is set to j-8, and the coefficient ajIs aj＝0.97/λjThe algorithm was executed 50 times. The standard deviation sigma of the noise can be obtained
Figure RE-GDA0002366358540000095
And MAD (y) mean [ | yi-median(y)|]N is estimated, i 1, 2. Finally, horizontal fault periodic pulses and vertical fault periodic pulses are obtained as shown in fig. 4(a) and 4(c), and it can be seen that an obvious instantaneous fault pulse sequence is sequentially separated from an original vibration signal to obtain a pulse interval of 0.0084s, and the pulse interval is matched with the fault frequency of a bearing outer ring. Fig. 4(b) and 4(d) are Hilbert envelope spectrums of the fault periodic pulse signals in the horizontal direction and the vertical direction, and it can be seen that noise interference frequencies around the fault frequency are greatly suppressed, and the fault frequency is more prominent.
The above results show that: the bearing fault periodic pulse sparse separation and diagnosis method under the strong background noise can effectively filter background interference noise irrelevant to fault frequency, can accurately separate out a periodic instantaneous pulse sequence relevant to bearing outer ring faults, and can clearly and prominently detect the bearing outer ring fault characteristic frequency and harmonic frequency of the bearing outer ring fault characteristic frequency by the extracted envelope spectrum, so that the extraction and fault position diagnosis of the periodic fault pulse are realized, and the theoretical effectiveness and the engineering practicability of the method are also proved.

Claims (2)

1. A bearing fault periodic pulse sparse separation and diagnosis method under strong background noise is characterized by comprising the following steps:
step 1, respectively installing acceleration sensors in the horizontal direction and the vertical direction of a bearing seat to be detected, and acquiring a vibration acceleration signal in the horizontal direction and a vibration acceleration signal in the vertical direction of the bearing;
step 2, establishing a target cost function:
Figure FDA0002021227920000011
in the formula (I), the compound is shown in the specification,
Figure FDA0002021227920000012
representing a wavelet transform coefficient to be estimated; ω represents a wavelet transform coefficient; f (ω) represents the objective cost function; y represents the vibration acceleration signal in the horizontal direction or the vibration acceleration signal in the vertical direction obtained in the step 1; lambda [ alpha ]jRepresenting the regularization parameter at the time scale j; continuous wavelet transform of transform coefficients x: omegaj,k＝Wj,kx, in the formula, Wj,kThe wavelet transform is performed on a translation scale j and a time scale k, y is Ax + w, A represents a matrix with Toeplitz, and w represents background noise or interference components; a isjRepresenting a penalty function scale coefficient; phi (omega)j,k；aj) β denotes regularization parameters, D denotes a first order differential matrix;
step 3, inputting the horizontal direction vibration acceleration signals and the vertical direction vibration acceleration signals obtained in the step 1 into the target cost function established in the step 2 respectively, and obtaining the horizontal direction vibration acceleration signals and the vertical direction vibration acceleration signals through iterative calculation of an alternating direction multiplier method
Figure FDA0002021227920000013
Obtaining a horizontal direction periodic fault pulse signal and a vertical direction periodic fault pulse signal through wavelet inverse transformation calculation, and defining the periodic fault pulse signals as
Figure FDA0002021227920000014
Then there is
Figure FDA0002021227920000015
2. The periodic impulse sparse separation and diagnosis method for bearing faults under strong background noise as claimed in claim 1, further comprising after said step 3:
and carrying out envelope demodulation on the horizontal direction periodic fault pulse signal and the vertical direction periodic fault pulse signal by using a Hilbert envelope demodulation method, extracting bearing fault characteristic frequency and frequency multiplication thereof, comparing the extracted bearing fault characteristic frequency with theoretical fault characteristic frequency, and identifying the fault position of the bearing.
CN201910279626.3A 2019-04-09 2019-04-09 Bearing fault periodic pulse sparse separation and diagnosis method under strong background noise Withdrawn CN110940522A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910279626.3A CN110940522A (en) 2019-04-09 2019-04-09 Bearing fault periodic pulse sparse separation and diagnosis method under strong background noise

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910279626.3A CN110940522A (en) 2019-04-09 2019-04-09 Bearing fault periodic pulse sparse separation and diagnosis method under strong background noise

Publications (1)

Publication Number Publication Date
CN110940522A true CN110940522A (en) 2020-03-31

Family

ID=69905719

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910279626.3A Withdrawn CN110940522A (en) 2019-04-09 2019-04-09 Bearing fault periodic pulse sparse separation and diagnosis method under strong background noise

Country Status (1)

Country Link
CN (1) CN110940522A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111750978A (en) * 2020-06-05 2020-10-09 中国南方电网有限责任公司超高压输电公司广州局 Data acquisition method and system of power device
CN112098094A (en) * 2020-09-27 2020-12-18 上海数深智能科技有限公司 Method for diagnosing fault vibration of low-speed heavy-load bearing
CN113804439A (en) * 2020-06-01 2021-12-17 株式会社日立大厦系统 Bearing inspection device and bearing inspection method
WO2023015855A1 (en) * 2021-08-10 2023-02-16 江苏大学 Generalized autocorrelation method for feature extraction of faults in bearings under variable speed conditions

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113804439A (en) * 2020-06-01 2021-12-17 株式会社日立大厦系统 Bearing inspection device and bearing inspection method
CN111750978A (en) * 2020-06-05 2020-10-09 中国南方电网有限责任公司超高压输电公司广州局 Data acquisition method and system of power device
CN112098094A (en) * 2020-09-27 2020-12-18 上海数深智能科技有限公司 Method for diagnosing fault vibration of low-speed heavy-load bearing
WO2023015855A1 (en) * 2021-08-10 2023-02-16 江苏大学 Generalized autocorrelation method for feature extraction of faults in bearings under variable speed conditions

Similar Documents

Publication Publication Date Title
CN110940522A (en) Bearing fault periodic pulse sparse separation and diagnosis method under strong background noise
CN108805059B (en) Sparse regularization filtering and self-adaptive sparse decomposition gearbox fault diagnosis method
Feng et al. Rolling element bearing fault detection based on optimal antisymmetric real Laplace wavelet
CN110046476B (en) Ternary binary fractal wavelet sparse diagnosis method for rolling bearing faults
CN113375939B (en) Mechanical part fault diagnosis method based on SVD and VMD
CN109708891B (en) Fault diagnosis method for raceway of flexible elliptical bearing
Hao et al. Morphological undecimated wavelet decomposition for fault diagnostics of rolling element bearings
CN109655266B (en) Wind turbine generator bearing fault diagnosis method based on AVMD and spectrum correlation analysis
CN109883706A (en) A kind of rolling bearing local damage Weak fault feature extracting method
CN112945546B (en) Precise diagnosis method for complex faults of gearbox
CN104215456A (en) Plane clustering and frequency-domain compressed sensing reconstruction based mechanical fault diagnosis method
CN112985809A (en) Rolling bearing fault diagnosis method based on signal multi-dimensional fine image
CN107966287B (en) Weak fault feature extraction method for self-adaptive electromechanical equipment
Lin et al. A review and strategy for the diagnosis of speed-varying machinery
CN111665050B (en) Rolling bearing fault diagnosis method based on clustering K-SVD algorithm
CN112098093A (en) Bearing fault feature identification method and system
CN114813123A (en) Rolling bearing weak fault diagnosis method based on PSO-VMD-MCKD
Li et al. A fault diagnosis approach for gears using multidimensional features and intelligent classifier
CN113758708B (en) Rolling bearing signal frequency domain fault diagnosis method based on L1 norm and group norm constraint
CN116086807A (en) Fault judging method for rolling bearing and related product
Xia et al. Convolutional sparse coding with periodic overlapped group sparsity for rolling element bearing fault diagnosis
Yu et al. A clustering K-SVD-based sparse representation method for rolling bearing fault diagnosis
CN113639985A (en) Mechanical fault diagnosis and state monitoring method based on optimized fault characteristic frequency spectrum
CN116028840A (en) Marine rotor fault diagnosis method for maximum overlapped discrete wavelet packet transformation time spectrum
CN112161805A (en) Bearing tiny fault diagnosis method based on time series scale analysis and CNN

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
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20200331