CN105445022B - A kind of planetary gear method for diagnosing faults based on dual-tree complex wavelet transform entropy Fusion Features - Google Patents
A kind of planetary gear method for diagnosing faults based on dual-tree complex wavelet transform entropy Fusion Features Download PDFInfo
- Publication number
- CN105445022B CN105445022B CN201510791644.1A CN201510791644A CN105445022B CN 105445022 B CN105445022 B CN 105445022B CN 201510791644 A CN201510791644 A CN 201510791644A CN 105445022 B CN105445022 B CN 105445022B
- Authority
- CN
- China
- Prior art keywords
- entropy
- dual
- wavelet transform
- planetary gear
- gear
- 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
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000004927 fusion Effects 0.000 title claims abstract description 28
- 239000013598 vector Substances 0.000 claims abstract description 29
- 238000004458 analytical method Methods 0.000 claims abstract description 24
- 239000000284 extract Substances 0.000 claims abstract description 21
- 230000009467 reduction Effects 0.000 claims abstract description 12
- 238000004088 simulation Methods 0.000 claims abstract description 8
- 238000001228 spectrum Methods 0.000 claims description 43
- 239000011159 matrix material Substances 0.000 claims description 18
- 102000008297 Nuclear Matrix-Associated Proteins Human genes 0.000 claims description 16
- 108010035916 Nuclear Matrix-Associated Proteins Proteins 0.000 claims description 16
- 210000000299 nuclear matrix Anatomy 0.000 claims description 16
- 230000003595 spectral effect Effects 0.000 claims description 13
- 238000005299 abrasion Methods 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 9
- 208000037656 Respiratory Sounds Diseases 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 238000003909 pattern recognition Methods 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 2
- 238000003745 diagnosis Methods 0.000 abstract description 12
- 230000008878 coupling Effects 0.000 abstract description 4
- 238000010168 coupling process Methods 0.000 abstract description 4
- 238000005859 coupling reaction Methods 0.000 abstract description 4
- 238000000354 decomposition reaction Methods 0.000 description 17
- 230000000694 effects Effects 0.000 description 5
- 239000000203 mixture Substances 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000010338 mechanical breakdown Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002592 echocardiography Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 210000004940 nucleus Anatomy 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
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
- G01M13/02—Gearings; Transmission mechanisms
- G01M13/021—Gearings
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a kind of planetary gear method for diagnosing faults based on dual-tree complex wavelet transform entropy Fusion Features, gathers integrated simulation experiment bench data, obtains planet gear carrier original vibration signal；Original vibration signal is decomposed using dual-tree complex wavelet transform, extracts the signal component of each frequency band；Entropy Feature Selection Model is built from multi-angle, obtains higher-dimension primitive character；The primitive character set formed using kernel Fisher discriminant analysis method to a variety of entropy features carries out dimension-reduction treatment, determines one group of optimal discriminant vector, extracts projection of the primitive character in optimal discriminant vector as sensitive fault feature, and with this determination fault type；Verify from multi-angle, the necessity of more spatial description characteristic informations and the validity for carrying out Feature Dimension Reduction using KFDA methods on this basis.The present invention is applied to non-linear, non-stationary and the planetary gear vibration signal of close coupling characteristic, can effectively extract sensitive fault feature, realize planetary Accurate Diagnosis.
Description
Technical field
The invention belongs to planetary gear fault diagnosis technology field, and in particular to one kind is based on dual-tree complex wavelet transform-entropy
The planetary gear method for diagnosing faults of Fusion Features.
Background technology
Planetary gear is frequently used in low speed, heavy duty, high intensity, the workplace of high pollution, and the generation of its failure is
As influenceing, equipment is reliable, principal element of stable operation.Because planetary gear itself is a strongly non-linear system, while
Disturbed in the course of work by external working environment, cause planetary gear to break down subtle, the vibration signal gathered
Show non-linear, non-stationary, close coupling characteristic.Research is adapted to Non-stationary vibration signal caused by processing planetary gear failure
Signal processing method, the Accurate Diagnosis for realizing planetary gear failure are the difficult point and focus studied at present.
At present, vibration signal processing method largely promotes the development of planetary gear fault diagnosis technology, when
Frequency analysis are the focuses of its research, and main Time-Frequency Analysis Method has：STFT conversion, Wigner-Ville distribution, EMD decompose,
Wavelet decomposition etc..Wherein STFT conversion, Wigner-Ville distribution have good analytical effect for stationary signal, but simultaneously
Complicated Non-stationary vibration signal is not applied to.EMD decompose be a kind of adaptive signal decomposition method, its by complexity non-stationary
Signal decomposition is into simple assertive evidence mode function, but EMD is decomposed the shortcomings that having 2 clearly, i.e. modal overlap and end
Point leakage.Wavelet decomposition passes through years development, is successfully applied to complicated Non-stationary vibration signal resolution process field, but
It is the deficiencies of wavelet decomposition equally exists frequency leakage, frequency alias, few translation sensitiveness and set direction, these shortcomings cause
The consequences such as signal detail information is lost or result is inaccurate.With the small wave direction complex field small echo development of real number field, dual-tree complex wavelet
Conversion is suggested, and compared with common wavelet transformation, dual-tree complex wavelet transform has translation invariance, anti-aliasing effect and multi-direction
The properties such as selection.
The content of the invention
It is an object of the invention to provide a kind of planetary gear failure based on dual-tree complex wavelet transform-entropy Fusion Features to examine
Disconnected method, adapts to non-linear, non-stationary and the planetary gear vibration signal of close coupling characteristic, can effectively extract sensitive fault spy
Sign, realize the Accurate Diagnosis of planetary gear failure.
To achieve the above object, the technical solution adopted by the present invention is：
A kind of planetary gear method for diagnosing faults based on dual-tree complex wavelet transform-entropy Fusion Features, comprises the following steps：
(1) integrated simulation experiment bench data are gathered, obtain planet gear carrier original vibration signal；
(2) original vibration signal is decomposed using dual-tree complex wavelet transform, extracts the signal component of each frequency band；
(3) entropy Feature Selection Model is built from multi-angle, obtains higher-dimension primitive character；
(4) the primitive character set formed using kernel Fisher discriminant analysis method to a variety of entropy features is carried out at dimensionality reduction
Reason, one group of optimal discriminant vector is determined, extract projection of the primitive character in optimal discriminant vector as sensitive fault feature, and with
This determination fault type；
(5) checking uses from multi-angle, the necessity of more spatial description characteristic informations and on this basis KFDA methods
Carry out the validity of Feature Dimension Reduction.
Further, planet gear carrier original vibration signal is determined with acceleration transducer in the step (1), described
Original vibration signal includes Gear Planet Transmission sun gear normal condition, broken conditions, few dentation state, tooth surface abrasion and tooth root crackle five
Type.
Further, the dual-tree complex wavelet transform in the step (2) is decomposed into 6 layers to original vibration signal, and extracts
The signal component of each frequency band.
Further, the entropy Feature Selection Model in the step (3) includes singular spectrum entropy, time domain energy entropy, power spectrum
Entropy and Sample Entropy, the primitive character of foundation is 28 dimensions.
Further, the meter of singular spectrum entropy, time domain energy entropy, Power Spectral Entropy and Sample Entropy is also included in the step (3)
Calculate；Wherein, the analysis length of phase space constructed in the extraction process of singular spectrum entropy be K=7000, delay constant for τ=
15, the singular value number of acquisition is 7000；During the asking for of Sample Entropy, the pattern dimension of setting is τ=1, similar tolerance limit
For r=0.15sd, sd is that signal standards is poor.
Further, dimensionality reduction is carried out to primitive character set using kernel Fisher discriminant analysis method in the step (4)
Handle and realize concretely comprising the following steps for Fault Identification：
(4-1) selects kernel function, and arrange parameter；
Primitive character set is converted into nuclear matrix by (4-2) using kernel function；
(4-3) asks for the within class scatter matrix and inter _ class relationship matrix of nuclear matrix；
(4-4) determines one group of optimal discriminant vector according to the within class scatter matrix and inter _ class relationship matrix of nuclear matrix；
(4-5) projects nuclear matrix to optimal discriminant vector, realizes dimension-reduction treatment, obtains sensitive fault feature；
(4-6) carries out Fault Pattern Recognition according to sensitive fault feature, obtains diagnostic result.
Further, the kernel function selected in the kernel Fisher discriminant analysis method in the step (4) is Gauss radial direction
Basic function k (x, y)=exp [- | | x-y | |2/2σ2], selected parameter is σ=0.1；Determine one group of optimal discriminant vector be
Preceding 4 optimal discriminant vectors.
Further, step (5) concretely comprise the following steps：
(5-1) is respectively to singular spectrum entropy and time domain energy entropy this 2 kinds of entropy features, singular spectrum entropy, time domain energy entropy and power
This 3 kinds of entropy features of spectrum entropy carry out Fusion Features；
(5-2) carries out dimension-reduction treatment using KFDA methods to the higher-dimension entropy feature of above-mentioned fusion, extracts sensitive fault spy
Reference ceases；
(5-3) is carried out to score with the sensitive fault characteristic information for using 4 kinds of entropy features extract after Fusion Features
The necessity of comprehensive, complete characteristic set is established in analysis, checking.
The beneficial effects of the invention are as follows：The method of the present invention is applied to non-linear, non-stationary and the planet of close coupling characteristic
Gearbox vibration signal, can effectively extract sensitive fault feature, realize planetary Accurate Diagnosis, to a certain extent enrich and
Perfect planetary gear method for diagnosing faults, as a result accurately and reliably, supervised suitable for the epicyclic gearbox state of large-scale heavy duty machinery
Survey and diagnose.
Brief description of the drawings
Fig. 1 is the flow chart of the method for the invention；
Fig. 2 is the time-domain diagram of five kinds of state gears of collection of the embodiment of the present invention；
Fig. 3 is the common wavelet transform decomposition result of the embodiment of the present invention；
Fig. 4 is dual-tree complex wavelet transform decomposition result of the embodiment of the present invention；
Fig. 5 a are the singular spectrum entropy curve of each band signal of five kinds of gear-types of the embodiment of the present invention；
Fig. 5 b are the time domain energy entropy curve of each band signal of five kinds of gear-types of the embodiment of the present invention；
Fig. 5 c are the frequency spectrum entropy curve of each band signal of five kinds of gear-types of the embodiment of the present invention；
Fig. 5 d are the Sample Entropy curve of each band signal of five kinds of gear-types of the embodiment of the present invention；
Fig. 6 a~6d are preceding 4 sensitive fault features when primitive character of embodiment of the present invention set 28 is tieed up；
Fig. 7 a~7d are preceding 4 sensitive fault features when primitive character of embodiment of the present invention set 14 is tieed up；
Fig. 8 a~8d are preceding 4 sensitive fault features when primitive character of embodiment of the present invention set 21 is tieed up.
Embodiment
The present invention is further described below in conjunction with the accompanying drawings.
Embodiments of the present invention are described in detail below in conjunction with accompanying drawing.It should be noted that retouched in following embodiments
The combination of the technical characteristic or technical characteristic stated is not construed as isolated, and they can be combined with each other and mutual group
Close so as to reach superior technique effect.
As shown in figure 1, the planetary gear fault diagnosis side based on dual-tree complex wavelet transform-entropy Fusion Features of the present invention
Method, comprise the following steps that：
Step 1, gather integrated simulation experiment bench data using acceleration vibrating sensor, including normal, broken teeth, few tooth,
Abrasion and the vibration signal of five kinds of state gears of tooth root crackle.
Step 2,6 layers of dual-tree complex wavelets decomposition are carried out to original vibration signal, and extract the signal component of each frequency band, walked
It is rapid as follows：
1) wavelet basis of two real number Wavelet representation for transient dual-tree complex wavelets is utilized：
In formula,WithFor 2 real number small echos, i is complex unit, andWithHilbert becomes each other
Change pair.
2) according to wavelet transformation theory, the high frequency coefficient and low frequency coefficient of real part are calculated：
Similarly, the high frequency coefficient and low frequency coefficient of imaginary part are calculated：
3) high frequency coefficient and low frequency coefficient of dual-tree complex wavelet are calculated according to step 2)：
4) by other wavelet coefficient zero setting, the signal of each frequency band is tried to achieve：
Step 3, entropy Feature Selection Model, including singular spectrum entropy, time domain energy entropy, Power Spectral Entropy and sample are built from multi-angle
This entropy, 28 dimension primitive characters are obtained, for the D obtained by dual-tree complex wavelet transformi, i=1,2 ..., J+1, if Di=(x1,
x2,...,xN), per the N number of data point of layer signal, the calculation procedure difference of multi-angle entropy feature is as follows：
1) calculating of singular spectrum entropy.By selecting data analysis length K and delay constant τ to build phase space, the phase of structure
Space carries out singular value decomposition, decomposes and obtains K singular value λ1≥λ2≥...≥λK, K singular value forms singular spectrum, by each
Singular value is regarded as is to one kind effectively division, the summation of K singular value of singular spectrum：Then singular spectrum entropy defines
For：
2) calculating of time domain energy entropy.Calculate and form each DiThe energy of data pointWith vibration gross energyWill
Each point vibration signal energy is regarded as to be divided to one kind of signal gross energy, then defining time domain energy entropy is：
3) calculating of Power Spectral Entropy.To each DiCarry out discrete Fourier transform and obtain FDi(ω), then power spectrum is expressed as：
Then { S (1), S (2) ..., S (N) } regards one kind effectively division to power energy in frequency domain as, vibration signal
General power energy isBy the use of pro rate of the different frequency in power spectrum as information distribution probability, then power
Composing entropy is：
4) calculating of Sample Entropy.Shown in structural matrix such as formula (10)：
Often the distance between row is for definition：
d(Xi,Xj)=max (| Xi(l)-Xj(l) |) l=1,2 ..., M (11)
Calculate each X in structural matrix (10)iWith other row vectors XjDistance.Given threshold values r, statistics d (Xi,Xj)<r
Number, calculates its ratio with row vector number N-K, and meter is done
All i, which are averaged, to be had：
τ=τ+1 is made, repeating above procedure can obtain：
The Sample Entropy of definition signal is：
HSE=-ln [Bτ+1/Bτ] (15)
The analysis length of constructed phase space is K=7000 in the extraction process of described singular spectrum entropy, delay constant
For τ=15, the singular value number of acquisition is 7000；During the asking for of Sample Entropy, the pattern dimension of setting is τ=1, similar appearance
R=0.15sd is limited to, sd is that signal standards is poor.
Step 4, the primitive character set formed using kernel Fisher discriminant analysis method to a variety of entropy features carry out dimensionality reduction
Processing, one group of optimal discriminant vector is determined, extract projection of the primitive character in optimal discriminant vector as sensitive fault feature, and
With this determination fault type, it is comprised the following steps that：
1) suitable kernel function, and arrange parameter are selected；
2) primitive character set is converted into nuclear matrix using kernel function；
3) within class scatter matrix and inter _ class relationship matrix of nuclear matrix are asked for；
4) according to the within class scatter matrix of nuclear matrix and inter _ class relationship matrix, one group of optimal discriminant vector is determined；
5) nuclear matrix is projected to optimal discriminant vector, realizes dimension-reduction treatment, obtain sensitive fault feature；
6) Fault Pattern Recognition is carried out according to sensitive fault feature, obtains diagnostic result.
The kernel function selected in described kernel Fisher discriminant analysis method for Gaussian radial basis function k (x, y)=exp [-
||x-y||2/2σ2], selected parameter is σ=0.1；The one group of optimal discriminant vector determined is preceding 4 optimal discriminant vectors.
Step 5, checking use from multi-angle, the necessity of more spatial description characteristic informations and on this basis KFDA side
Method carries out the validity of Feature Dimension Reduction, and it specifically comprises the following steps：
1) respectively to 2 kinds of entropy features (singular spectrum entropy and time domain energy entropy) and 3 kinds of entropy features (singular spectrum entropy, time domain energies
Entropy and Power Spectral Entropy) carry out Fusion Features；
2) dimension-reduction treatment is carried out to the higher-dimension entropy feature of above-mentioned fusion using KFDA methods, extracts sensitive fault feature letter
Breath；
3) the sensitive fault characteristic information with using 4 kinds of entropy features extract after Fusion Features is analyzed,
The necessity of comprehensive, complete characteristic set is established in checking.
The embodiment Binding experiment integrated use above method carries out planetary gear fault diagnosis.Planetary gear fault simulation is real
Test on the DDS mechanical breakdown comprehensive simulation test platforms of Spectra Quest companies of the U.S. and carry out.This experiment measures planet tooth altogether
Five kinds of sun gear normal condition, broken conditions, few dentation state, tooth surface abrasion and tooth root crackle states are taken turns, by being shaken to gathering
Dynamic signal is analyzed, the planetary gear method for diagnosing faults that inspection institute establishes.
Sun gear fault simulation experiment is carried out on DDS mechanical breakdown integrated simulation experiment bench, takes normal gear, broken teeth altogether
Gear, few gear, the vibration signal of five kinds of malfunctions of tooth surface abrasion and tooth root crackle, the motor output speeds of setting are
40Hz, planetary gear basic parameter are：The tooth of sun gear 28, the tooth of planetary gear 40, the tooth of ring gear 100, corresponding fault signature frequency
Rate is respectively：Sun gear 20.83Hz, planetary gear 4.05Hz, ring gear 5.83Hz, five kinds of sun gear gear conditions of collection shake
Dynamic signal is as shown in Fig. 2 from figure 2 it can be seen that by the comparison to five kinds of Gearbox vibration signals, because planetary gear is being made
The influence of make, assemble etc., the vibration signal of normal gear equally exist certain impact characteristics, the impact composition of introducing compared with
It is complicated.Few tooth failure gear because sun gear lacks a gear teeth, cause the impact composition that is introduced in engagement process with it is normal
Gear is different, and the vibration signal of few gear shows certain periodicity.Few gear vibration signal produces every 0.1502s
Once similar waveform, that is, there are 6.66Hz frequency contents.Similar waveform is the higher impact of two amplitudes, during this 2 impacts
Between at intervals of 0.05s.Abrased gear is because each gear teeth of sun gear are subject to different degrees of abrasion, therefore in running
In, more radio-frequency component is introduced, stronger, more complicated amplitude modulation is produced and frequency modulation phenomenon, normal gear is more original
Impact composition is blanked.Because the fault degree of broken teeth gear is more serious, its vibration signal shows certain periodicity, cycle
For 0.15s, frequency 6.67Hz.Tooth root Gear with Crack due to its wheel root portion have crackle, cause gear tooth rigidity to change,
Gear tooth rigidity shows strong non-linear and time variation in engagement process, therefore generates more random high frequency letter
Number, change the time domain waveform of normal gear.Planet tooth is found by carrying out time-domain analysis to five kinds of gear condition vibration signals
The generation of wheel failure will produce the signal component of other frequencies, but can not have been extracted by carrying out analysis to time domain waveform
The characteristic information of effect, the fault type of gear can not be accurately distinguished out.
Below using the planetary gear method for diagnosing faults pair based on dual-tree complex wavelet transform and entropy Fusion Features proposed
Said gear vibration signal is handled.By taking broken teeth Gearbox vibration signal as an example, by many experiments, by original vibration signal point
Solve as 6 layers, be reconstructed respectively for decomposing obtained each layer wavelet coefficient, that is, obtain each layer signal of wavelet decomposition.Commonly
Wavelet transform exploded view and dual-tree complex wavelet transform exploded view are as shown in Figure 3 and Figure 4.Xc6 is approximate part coefficient in figure
The signal of reconstruct, xd1-xd6 are the signal of detail coefficients reconstruct.Wherein xc6 corresponding frequency bands are [0Hz, 100Hz], xd1-xd6 pairs
Answer frequency range respectively [100Hz, 200Hz], [200Hz, 400Hz], [400Hz, 800Hz], [800Hz, 1600Hz], [1600Hz,
3200Hz], [3200Hz, 6400Hz].The signal decomposition figure and dual-tree complex wavelet transform obtained from common wavelet transform obtains
To signal decomposition figure compare as can be seen that common wavelet transformation signal decomposition due to originals such as discrete sampling and signal interferences
Cause, frequency alias phenomenon is than more serious, such as the frequency content that red elliptic region marks in Fig. 3.Dual-tree complex wavelet transform is preferable
Each frequency band of extraction signal, frequency alias phenomenon suppressed well, and the useful information of each frequency band is fully shown.It is right
After planetary gear original vibration signal carries out dual-tree complex wavelet transform decomposition, carried for each band signal of decomposition from multiple angles
Take a variety of entropy features, respectively singular spectrum entropy, time domain energy entropy, Power Spectral Entropy and Sample Entropy.The wherein extraction pin of singular spectrum entropy
It is the phase space of τ=15 to each band signal creation analysis length K=7000 and delay constant, obtains 7000 singular values, root
Singular spectrum entropy is obtained according to formula (6), the singular spectrum entropy of each band signal of five kinds of gear-types is as shown in Figure 5 a.Time domain energy entropy
As shown in Figure 5 b, frequency spectrum entropy is as shown in Figure 5 c.During the asking for of Sample Entropy, Setting pattern dimension τ=1, similar tolerance limit r=
0.15sd, sd are that signal standards is poor, and the Sample Entropy of five kinds of each frequency bands of gear is as fig 5d.
Singular spectrum entropy reflects the very inhomogeneous complexity of time-domain signal and uncertainty.It can be seen that abrasion from Fig. 5 a
Unusual composition is most complicated in each frequency band distribution caused by Gearbox vibration signal, and its singular spectrum entropy is in d1-d6 layer highests, in c6
Layer, broken teeth gear singular spectrum entropy is maximum, and other four kinds of gear entropy are very close to five kinds of gear singular spectrum entropy of d4-d5 layers do not have substantially
Have any different, d1-d3 layers are normal, the singular spectrum entropy of few tooth, abrased gear relatively, the singular spectrum entropy ratio of broken teeth and tooth root crackle
It is closer to, wherein tooth, the singular spectrum entropy of abrased gear are bigger than the singular spectrum entropy of broken teeth and tooth root Gear with Crack normally, less.Institute
Cogged singular spectrum entropy is in increase tendency.Time domain energy entropy reflects the complexity and uncertainty of time domain energy distribution,
It can be seen that five kinds of gear time domain energy entropy of d1-c6 layers are distributed no obvious rules from Fig. 5 b, tooth in each layer band signal
Root Gear with Crack time domain energy entropy is maximum, minimum in d2 and c6 layer broken teeth gear time domain energies entropy, d6 layer abrased gear time domain energy
It is minimum to measure entropy, the few gear time domain energy entropy of d3 layers is minimum.The cogged time domain energy entropy of institute is in reduction trend.Power Spectral Entropy is anti-
Frequency domain energy complex distribution and uncertainty are reflected, it can be seen that the Power Spectral Entropy of five kinds of gear of c5 and d6 layers is poor from Fig. 5 c
Different smaller, five kinds of gear of d1-d5 layers differ greatly, and d3-d6 layer abrased gears Power Spectral Entropy is maximum, d3-d4 layer tooth root crackle work(
Rate spectrum entropy is minimum, and d1-d2 layer broken teeth gears Power Spectral Entropy is minimum.The cogged Power Spectral Entropy of institute is in increase tendency.Sample Entropy from
The sample of the complexity and uncertainty, as can be seen from the figure five kinds of gear of d1-d3 layers of signal itself approximate angle reflection signal
This entropy difference is not too big, and Sample Entropy difference is mainly reflected in d4-c6 layer signals, and the normal gear Sample Entropy of c6 layers is minimum.Abrasion
Gear Sample Entropy is maximum.Minimum in the few gear Sample Entropy of d4-d6 layers, normal gear Sample Entropy is in increased dramatically trend, in d4-
The normal gear Sample Entropy of d5 layers is maximum, and increase tendency is all presented in c6-d4 layers in the vibration signal of broken teeth and tooth root Gear with Crack, and
And each layer Sample Entropy of broken teeth gear is consistently greater than each layer Sample Entropy of tooth root Gear with Crack.The cogged Sample Entropy of institute is in increase
Trend.In the entropy feature that each layer band signal for decomposing to obtain to dual-tree complex wavelet transform is asked for, some has to Fault-Sensitive
The difference as caused by failure and unobvious, therefore more invalid feature be present, the presence of these invalid features can be to feature
Selective extraction produces interference effect, finally influences diagnostic result, it is therefore necessary to entropy feature set is merged, realizes that feature drops
Dimension processing, extracts the characteristic quantity to Fault-Sensitive, realizes planetary gear fault diagnosis.
Because primary signal is decomposed into 6 layers by dual-tree complex wavelet, and for per layer signal respectively from four angle extraction entropys
Characteristic information, the entropy characteristic dimension of every kind of gear reach 28 dimensions, and intrinsic dimensionality is excessive.Therefore, kernel Fisher discriminant analysis is utilized
Method is realized carries out dimension-reduction treatment to the primitive character set that a variety of entropy features are formed, and determines one group of optimal discriminant vector, extracts
Primitive character optimal discriminant vector projection as sensitive fault feature, realize that fault type is classified.Planet tooth is gathered respectively
Every kind of 20 groups of central gear state sample is taken turns, gathers five kinds of gear conditions altogether, forms 100 groups of training samples.For every group of training
Sample, dual-tree complex wavelet decomposition being carried out respectively, obtaining 6 layer signals, 4 kinds of entropy features are extracted to each layer signal of decomposition, formed former
Beginning characteristic set, form 28 × 100 dimension matrixes.The target analyzed using primitive character set as KFDA, by more seed nucleus letters
Number is tested, selection gaussian radial basis function k (x, y)=exp [- | | x-y | |2/2σ2], parameter σ is set by test of many times
It is set to σ=0.1.Primitive character set is converted into nuclear matrix using kernel function, ask for nuclear matrix within class scatter matrix and
Inter _ class relationship matrix.
According to the within class scatter matrix of nuclear matrix and inter _ class relationship matrix, one group of optimal discriminant vector is determined, is extracted
Preceding 4 optimal discriminant vectors form new feature space, and nuclear matrix is extracted after the projection of optimal discriminant vector is dimensionality reduction
Sensitive fault feature.When in this direction projection so that away from minimum, class spacing in the overall class of the point corresponding to projection coordinate
Maximum, 28 original characteristic attributes can be so replaced with 4 sensitive fault features, eliminate invalid feature, extracted main special
Sign, dimension reduction 24.The sensitive fault feature extracted is as shown in Fig. 6 a~6d, and 1-25 groups are normal gear in figure, 26-50 groups
For few gear, 51-75 groups are abrased gear, and 76-100 groups are broken teeth gear, and 101-125 groups are tooth root Gear with Crack.From figure
In it can be seen that the sensitive fault feature of five kinds of gear-types of extraction has significant difference, although same type gear sample
Sensitive fault characteristic value there is certain fluctuation, but fluctuate in the range of very little.It is although several in some sensitive features
The feature of kind of gear-type occurs overlapping, such as in first sensitive features, abrased gear and tooth root Gear with Crack are easy
Coincidence is produced, in the second sensitive features, few gear and broken teeth gear easily produce coincidence.But in the third and fourth feature
In, abrased gear and tooth root Gear with Crack are very easy to distinguish, in the first, third and fourth feature, few gear and broken teeth
Gear is easily distinguished, and the numerical value difference of feature is larger., can Accurate Diagnosis planet tooth by four sensitive features of comprehensive analysis
Take turns fault type.
The primitive character that four kinds of entropy features are formed realizes the multi-angle description to signal, expressing information more comprehensively, it is more accurate
Really.Feature Dimension Reduction is realized using KFDA analysis methods fusion primitive character, makes full use of multi-angle, the feature in more spaces letter
Breath, remains the characteristic information most sensitive to failure, realizes planetary gear fault diagnosis.In order to illustrate from multi-angle, more spaces
The necessity of Expressive Features information and on this basis KFDA realize the validity of Feature Dimension Reduction application.Below with KFDA methods
Respectively to 2 kinds of entropy features (singular spectrum entropy and time domain energy entropy) and 3 kinds of entropy features (singular spectrum entropy, time domain energy entropy and power spectrum
Entropy) merged respectively, extract the fault characteristic information of sensitivity.
The result of 2 kinds of entropy Fusion Features is as shown in Fig. 7 a~7d, it can be seen that in the 3rd sensitive features and the 4th
In sensitive features, the characteristic value aliasings of five kinds of gear-types is not used to judge gear-type than more serious；It is sensitive special first
In second sensitive features of seeking peace, the characteristic value of gear-type is equally easier aliasing occur, in the second sensitive features
Few tooth and broken teeth are it is also possible to there is the possibility that characteristic value is completely superposed.Four sensitive features of same type gear sample simultaneously
Value fluctuation is larger, it is impossible to accurately distinguishes the fault type that planetary gear occurs.3 kinds of entropy Fusion Features results such as Fig. 8 a~8d institutes
Show, it can be seen that abrasion, broken teeth and tooth root Gear with Crack type are substantially completely overlapped one in the 4th sensitive features
Rise；In the 3rd sensitive features, tooth root Gear with Crack type can be distinguished substantially, and other gear-type characteristic values overlap more serious；
In the second sensitive features, abrased gear type is easier to distinguish, and the characteristic value of other gear-types has smaller difference, and
And easily produce the coincidence of characteristic value；In the first sensitive features, broken teeth and tooth root Gear with Crack type are easier to distinguish, few tooth tooth
There is situation about being completely superposed in the characteristic value of wheel and abrased gear.By it was found that, utilize four kinds of entropy features (28 original spies of dimension
Sign) carry out kernel-based Fisher discriminant analysis, the projection value of the primitive characters of five kinds of planetary gear types in optimal discriminant vector direction
With notable difference, and every kind of gear is directed to, the fluctuation of projection value is smaller, is most preferably being reflected according to the core eigenmatrix extracted
Projection value in other vector direction can accurately distinguish planetary gear fault type.When 2 kinds of entropy features of utilization (14 dimension primitive character)
When carrying out kernel-based Fisher discriminant analysis with 3 kinds of entropy features (21 dimension primitive character), the primitive character of five kinds of planetary gear types exists
The projection value difference and unobvious in optimal discriminant vector direction, while projection value fluctuation is larger, in a discriminant vectorses direction, no
Five kinds of gear-types can accurately be distinguished, serious coincidence phenomenon occurs for the Projection Character value for there are multiple gear-types.Therefore, it is former
Comprehensive, the complete foundation of beginning feature set is advantageous to realize that the accurate of planetary gear failure is examined using kernel-based Fisher discriminant analysis
It is disconnected.
Described above is only the preferred embodiment of the present invention, it should be pointed out that：For the ordinary skill people of the art
For member, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications also should
It is considered as protection scope of the present invention.
Claims (8)
- A kind of 1. planetary gear method for diagnosing faults based on dual-tree complex wavelet transform-entropy Fusion Features, it is characterised in that：Including Following steps：(1) integrated simulation experiment bench data are gathered, obtain planet gear carrier original vibration signal；(2) original vibration signal is decomposed using dual-tree complex wavelet transform, extracts the signal component of each frequency band；(3) entropy Feature Selection Model is built from multi-angle, obtains higher-dimension primitive character；(4) the primitive character set formed using kernel Fisher discriminant analysis method to a variety of entropy features carries out dimension-reduction treatment, really Fixed one group of optimal discriminant vector, extraction primitive character optimal discriminant vector projection as sensitive fault feature, it is and true with this Determine fault type；(5) checking is carried out from multi-angle, the necessity of more spatial description characteristic informations and using KFDA methods on this basis The validity of Feature Dimension Reduction.
- 2. the planetary gear method for diagnosing faults as claimed in claim 1 based on dual-tree complex wavelet transform-entropy Fusion Features, its It is characterised by：Planet gear carrier original vibration signal is determined with acceleration transducer in the step (1), the original vibration Signal includes Gear Planet Transmission sun gear normal condition, broken conditions, few dentation state, tooth surface abrasion and the type of tooth root crackle five.
- 3. the planetary gear method for diagnosing faults as claimed in claim 1 based on dual-tree complex wavelet transform-entropy Fusion Features, its It is characterised by：Dual-tree complex wavelet transform in the step (2) is decomposed into 6 layers to original vibration signal, and extracts each frequency band Signal component.
- 4. the planetary gear method for diagnosing faults as claimed in claim 1 based on dual-tree complex wavelet transform-entropy Fusion Features, its It is characterised by：Entropy Feature Selection Model in the step (3) includes singular spectrum entropy, time domain energy entropy, Power Spectral Entropy and sample Entropy, the primitive character of foundation is 28 dimensions.
- 5. the planetary gear method for diagnosing faults as claimed in claim 4 based on dual-tree complex wavelet transform-entropy Fusion Features, its It is characterised by：Also include the calculating of singular spectrum entropy, time domain energy entropy, Power Spectral Entropy and Sample Entropy in the step (3)；Wherein, The analysis length of constructed phase space is K=7000 in the extraction process of singular spectrum entropy, and delay constant is τ=15, is obtained Singular value number be 7000；During the asking for of Sample Entropy, the pattern dimension of setting is τ=1, and similar tolerance limit is r= 0.15sd, sd are that signal standards is poor.
- 6. the planetary gear method for diagnosing faults as claimed in claim 1 based on dual-tree complex wavelet transform-entropy Fusion Features, its It is characterised by：Using kernel Fisher discriminant analysis method to primitive character set progress dimension-reduction treatment and real in the step (4) Existing Fault Identification concretely comprises the following steps：(4-1) selects kernel function, and arrange parameter；Primitive character set is converted into nuclear matrix by (4-2) using kernel function；(4-3) asks for the within class scatter matrix and inter _ class relationship matrix of nuclear matrix；(4-4) determines one group of optimal discriminant vector according to the within class scatter matrix and inter _ class relationship matrix of nuclear matrix；(4-5) projects nuclear matrix to optimal discriminant vector, realizes dimension-reduction treatment, obtains sensitive fault feature；(4-6) carries out Fault Pattern Recognition according to sensitive fault feature, obtains diagnostic result.
- 7. the planetary gear method for diagnosing faults as claimed in claim 6 based on dual-tree complex wavelet transform-entropy Fusion Features, its It is characterised by：The kernel function selected in kernel Fisher discriminant analysis method in the step (4) is Gaussian radial basis function k (x, y)=exp [- | | x-y | |2/2σ2], selected parameter is σ=0.1；Determine one group of optimal discriminant vector for first 4 most Good discriminant vectorses.
- 8. the planetary gear method for diagnosing faults as claimed in claim 1 based on dual-tree complex wavelet transform-entropy Fusion Features, its It is characterised by：Step (5) concretely comprise the following steps：(5-1) is respectively to singular spectrum entropy and time domain energy entropy this 2 kinds of entropy features, singular spectrum entropy, time domain energy entropy and Power Spectral Entropy This 3 kinds of entropy features carry out Fusion Features；(5-2) carries out dimension-reduction treatment using KFDA methods to the higher-dimension entropy feature of above-mentioned fusion, extracts sensitive fault feature letter Breath；(5-3) is analyzed with the sensitive fault characteristic information for using 4 kinds of entropy features extract after Fusion Features, tests Card establishes the necessity of comprehensive, complete characteristic set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510791644.1A CN105445022B (en) | 2015-11-17 | 2015-11-17 | A kind of planetary gear method for diagnosing faults based on dual-tree complex wavelet transform entropy Fusion Features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510791644.1A CN105445022B (en) | 2015-11-17 | 2015-11-17 | A kind of planetary gear method for diagnosing faults based on dual-tree complex wavelet transform entropy Fusion Features |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105445022A CN105445022A (en) | 2016-03-30 |
| CN105445022B true CN105445022B (en) | 2017-11-14 |
Family
ID=55555482
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510791644.1A Active CN105445022B (en) | 2015-11-17 | 2015-11-17 | A kind of planetary gear method for diagnosing faults based on dual-tree complex wavelet transform entropy Fusion Features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105445022B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106124988A (en) * | 2016-06-28 | 2016-11-16 | 江苏科技大学 | A kind of motor multi-state fault detection method based on RBF, multilamellar FDA and SVDD |
| CN107180140B (en) * | 2017-06-08 | 2019-12-10 | 中南大学 | Shafting fault identification method based on dual-tree complex wavelet and AdaBoost |
| CN107608936B (en) * | 2017-09-22 | 2020-09-15 | 桂林电子科技大学 | Method for extracting compound fault characteristics of planetary gear box |
| CN107832687A (en) * | 2017-10-27 | 2018-03-23 | 武汉大学 | Fault diagnostic method for transformer winding based on wireless identification sensing |
| CN109864740B (en) * | 2018-12-25 | 2022-02-01 | 北京津发科技股份有限公司 | Surface electromyogram signal acquisition sensor and equipment in motion state |
| CN110134719B (en) * | 2019-05-17 | 2023-04-28 | 贵州大学 | Identification and classification method for sensitive attribute of structured data |
| CN110988511A (en) * | 2019-07-22 | 2020-04-10 | 华南理工大学 | Power electronic converter nonlinear identification method based on multiple entropy characteristic extraction |
| CN110844111B (en) * | 2019-10-11 | 2022-11-04 | 中国直升机设计研究所 | Multi-characteristic index bevel gear health state assessment method |
| CN111027577B (en) * | 2019-11-13 | 2023-03-31 | 湖北省纤维检验局 | Fabric abnormal texture type identification method and device |
| CN111351655A (en) * | 2020-03-25 | 2020-06-30 | 湖南铁路科技职业技术学院 | Method for diagnosing gear fault in locomotive gear box |
| CN115371988A (en) * | 2022-10-27 | 2022-11-22 | 北谷电子有限公司 | Engineering machinery fault diagnosis method and system based on multi-feature fusion |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103776654A (en) * | 2014-02-21 | 2014-05-07 | 黑龙江省科学院自动化研究所 | Method for diagnosing faults of multi-sensor information fusion |
| CN104330255A (en) * | 2014-11-04 | 2015-02-04 | 徐州隆安光电科技有限公司 | Gear fault diagnosing method based on multi-sensor information fusion |
| CN104390781A (en) * | 2014-11-26 | 2015-03-04 | 中国矿业大学 | Gear fault diagnosis method based on LMD and BP neural network |
| CN104483127A (en) * | 2014-10-22 | 2015-04-01 | 徐州隆安光电科技有限公司 | Method for extracting weak fault characteristic information of planetary gear |
| CN104655423A (en) * | 2013-11-19 | 2015-05-27 | 北京交通大学 | Rolling bearing fault diagnosis method based on time-frequency domain multidimensional vibration feature fusion |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2008292288A (en) * | 2007-05-24 | 2008-12-04 | Mitsubishi Electric Engineering Co Ltd | Bearing diagnostic device for reduction gear |
-
2015
- 2015-11-17 CN CN201510791644.1A patent/CN105445022B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104655423A (en) * | 2013-11-19 | 2015-05-27 | 北京交通大学 | Rolling bearing fault diagnosis method based on time-frequency domain multidimensional vibration feature fusion |
| CN103776654A (en) * | 2014-02-21 | 2014-05-07 | 黑龙江省科学院自动化研究所 | Method for diagnosing faults of multi-sensor information fusion |
| CN104483127A (en) * | 2014-10-22 | 2015-04-01 | 徐州隆安光电科技有限公司 | Method for extracting weak fault characteristic information of planetary gear |
| CN104330255A (en) * | 2014-11-04 | 2015-02-04 | 徐州隆安光电科技有限公司 | Gear fault diagnosing method based on multi-sensor information fusion |
| CN104390781A (en) * | 2014-11-26 | 2015-03-04 | 中国矿业大学 | Gear fault diagnosis method based on LMD and BP neural network |
Non-Patent Citations (2)
| Title |
|---|
| 基于双树复小波变换和奇异谱熵的采煤机摇臂轴承故障诊断;吕栋鸿;《山东煤炭科技》;20151028(第10期);第103-104、107页 * |
| 基于核Fisher判别分析方法的非线性统计过程监控与故障诊断;赵旭等;《化工学报》;20070430;第58卷(第4期);第951-956页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105445022A (en) | 2016-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105445022B (en) | A kind of planetary gear method for diagnosing faults based on dual-tree complex wavelet transform entropy Fusion Features | |
| Zhang et al. | A novel Fast Entrogram and its applications in rolling bearing fault diagnosis | |
| Jiang et al. | A coarse-to-fine decomposing strategy of VMD for extraction of weak repetitive transients in fault diagnosis of rotating machines | |
| Li et al. | Fault diagnosis for rolling bearing based on VMD-FRFT | |
| Hemmati et al. | Roller bearing acoustic signature extraction by wavelet packet transform, applications in fault detection and size estimation | |
| CN105510023B (en) | Variable working condition wind power planetary gear box fault diagnosis method based on divergence index | |
| Jiang et al. | A time-frequency spectral amplitude modulation method and its applications in rolling bearing fault diagnosis | |
| Bin et al. | Early fault diagnosis of rotating machinery based on wavelet packets—Empirical mode decomposition feature extraction and neural network | |
| Guo et al. | A compound fault diagnosis method of rolling bearing based on wavelet scattering transform and improved soft threshold denoising algorithm | |
| Osman et al. | An enhanced Hilbert–Huang transform technique for bearing condition monitoring | |
| CN105928702B (en) | Variable working condition box bearing method for diagnosing faults based on form PCA | |
| Xu et al. | The Enfigram: A robust method for extracting repetitive transients in rolling bearing fault diagnosis | |
| CN103902844A (en) | Transformer vibration signal de-noising method based on EEMD kurtosis threshold value | |
| Sheng et al. | Applications in bearing fault diagnosis of an improved Kurtogram algorithm based on flexible frequency slice wavelet transform filter bank | |
| Li et al. | Incipient fault information determination for rolling element bearing based on synchronous averaging reassigned wavelet scalogram | |
| Cui et al. | Spectrum-based, full-band preprocessing, and two-dimensional separation of bearing and gear compound faults diagnosis | |
| Xu et al. | An adaptive spectrum segmentation method to optimize empirical wavelet transform for rolling bearings fault diagnosis | |
| CN110398364A (en) | Epicyclic gearbox method for diagnosing faults based on resonance sparse decomposition and FastICA algorithm | |
| Gao et al. | Fault diagnosis of rolling bearings based on improved energy entropy and fault location of triangulation of amplitude attenuation outer raceway | |
| CN112485028B (en) | Feature spectrum extraction method of vibration signal and mechanical fault diagnosis analysis method | |
| Ma et al. | An improved intrinsic time-scale decomposition method based on adaptive noise and its application in bearing fault feature extraction | |
| Wang et al. | An order tracking-free method for variable speed fault diagnosis based on adaptive chirp mode decomposition | |
| Xu et al. | Rolling bearing fault feature extraction via improved SSD and a singular-value energy autocorrelation coefficient spectrum | |
| Chen et al. | Rolling bearing fault feature extraction method using adaptive maximum cyclostationarity blind deconvolution | |
| Li et al. | Bearing fault diagnosis method based on maximum noise ratio kurtosis product deconvolution with noise conditions |
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 |