CN107255563A - Realize gear-box mixed fault signal blind source separation method - Google Patents
Realize gear-box mixed fault signal blind source separation method Download PDFInfo
- Publication number
- CN107255563A CN107255563A CN201710504367.0A CN201710504367A CN107255563A CN 107255563 A CN107255563 A CN 107255563A CN 201710504367 A CN201710504367 A CN 201710504367A CN 107255563 A CN107255563 A CN 107255563A
- Authority
- CN
- China
- Prior art keywords
- signal
- fault
- blind source
- mixed
- mrow
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- G01M13/028—Acoustic or vibration analysis
Abstract
Gear-box mixed fault signal blind source separation method is realized the invention discloses one kind, is related to gear-box mixed fault blind source separating diagnostic method technical field.Methods described comprises the following steps：EMD pretreatments are carried out to the single channel gear-box mixed fault signal collected；IMF components are chosen using white noise statistical nature and kurtosis value associated methods, effective vibration mode component is used as；Packet is reconstructed to selected IMF components, as the input signal of blind source separating, mixed signal blind source separating is carried out with reference to CICA algorithms, extracts fault-signal interested；Hilbert Envelope Analysis is carried out to separating the fault-signal interested drawn through CICA methods, its envelope spectrum is obtained, is diagnosed to be fault signature interested in mixed fault, realizes blind source separating.Methods described can realize that blind source separating single channel extends, and extract fault-signal interested.
Description
Technical field
The present invention relates to gear-box mixed fault blind source separating diagnostic method technical field, more particularly to one kind are achievable blind
Source separation single channel extension, extracts the gear-box mixed fault blind source separation method of fault-signal interested.
Background technology
Plant equipment is complicated, is made up of many parts, in order to obtain the running status that the equipment is complete, frequently with many
Point layout sensor gathers signal, but the vibration signal that each sensor is picked up not is to survey part in the position
Indeed vibrations, it includes the synthesis oscillation of all parts, and the transmitting path of the vibration source arrival sensor of all parts generation
Also different, which increase the complexity of sensor measured signal.Obviously there are two unknown factors in vibration signal here：
Vibration source is unknown, and the transmission hybrid parameter of vibration source to sensor is unknown.
Blind source separating (BSS) is to realize one of effective ways of mechanical fault diagnosis, and it utilizes equipment fault signal
With the relative independentability of noise signal, the noise remove to observation signal is realized.BSS can be extracted to be fallen into oblivion in noise completely
Useful signal, therefore, is easier faint in discovery plant equipment than traditional wavelet analysis and Hilbert-Huang conversion
Failure.However, in general blind separation model, often requiring that the number of signal receiver is no less than the number of information source.But thing
In reality, because Cost Problems, and environment monitored are limited, possibly multiple sensors can not be installed simultaneously to equipment, sometimes very
To the situation for occurring being only capable of carrying out it single channel monitoring.Therefore, single channel blind source separating problem is increasingly becoming fanaticism in recent years
Focus in number research field.
Empirical mode decomposition (empirical mode decomposition, EMD) is one that Huang was proposed in 1998
Time-Frequency Analysis Method is planted, signal decomposition is several intrinsic mode functions by its time scale based on signal local feature
(intrinsic mode function, IMF) sum.Wherein each IMF comprises only single frequency content at each moment,
So that the instantaneous frequency of signal has physical meaning, it is consequently adapted to analyze non-stationary signal.Signal is obtained after being decomposed through EMD
The intrinsic mode function and a residual components of a series of frequencies from high to low, these intrinsic mode functions characterize signal institute
Some frequency distribution information.Inevitably there is high-frequency noise in the vibration signal collected on the rotating machinery of operation
Composition, this to decompose, and part IMF in obtained IMF sequences characterized is high frequency noise content in data, and EMD is decomposed in addition
False vibration mode is inevitably produced, the main contents of the mixed fault Signal Pretreatment based on EMD methods are exactly to pick
Except the high frequency noise content in IMF sequences and false vibration mode, the basic object of single channel extension and noise reduction is reached.
Independent component analysis (Independent Component Analysis, ICA) is to grow up in recent decades
A class signal processing method, be mainly used in solve under above-mentioned two unknown condition, according only to source signal statistical property and
The problem of independence is to recover source signal.But independent component analysis is only with the statistical independence of data, and actually
Project data also have other attributes, this can cause the utilization of data message not abundant enough.Constrained independent component is analyzed
(Constrained Independent Component Analysis, CICA) is exactly the prior information addition research object
Into parser, it grows up on the basis of independent component analysis development.Therefore can be by constrained independent component
Analyze for Research on Machine Fault Diagnosis, part early-stage weak fault feature can be believed from stronger ambient noise and other interference
Extracted in number, reach the purpose of fault diagnosis.
The content of the invention
The technical problems to be solved by the invention are how to provide a kind of achievable blind source separating single channel extension, extract sense
The gear-box mixed fault blind source separation method of the fault-signal of interest.
In order to solve the above technical problems, the technical solution used in the present invention is：One kind realizes that gear-box mixed fault is believed
Number blind source separation method, it is characterised in that comprise the following steps：
EMD pretreatments are carried out to the single channel gear-box mixed fault signal collected；
IMF components are chosen using white noise statistical nature and kurtosis value associated methods, effective vibration mode component is used as；
Packet is reconstructed to selected IMF components, as the input signal of blind source separating, mixed with reference to CICA algorithms
Signal blind source separating, extracts fault-signal interested；
Hilbert Envelope Analysis is carried out to separating the fault-signal interested drawn through CICA methods, its envelope is obtained
Spectrum, is diagnosed to be fault signature interested in mixed fault, realizes blind source separating.
Further technical scheme is：Single channel gear-box is gathered in methods described by single acceleration transducer to mix
Close fault-signal.
Further technical scheme is that the single channel mixed fault signal to collecting carries out EMD pretreatments
Method it is as follows：
Based on white noise statistic feature extraction mechanical oscillation pattern, to any normalization white noise time series nm, m=
1 ..., N, are expressed as after carrying out EMD decompositionN is natural number, and white noise sequence is divided into after EMD is decomposed
Solve as w IMF component, cl(m) it is l-th of IMF component.
Further technical scheme is that described utilization white noise statistical nature and kurtosis value associated methods choose IMF
Component, the method as effective vibration mode component is as follows：
Define cl(m) energy density is
cl(m) the average period corresponding to fourier spectra is
Show that EMD decomposes obtained any IMF energy density and the relation between average period is by the derivation of equationLogarithm using e the bottom of as is taken to formula both sides, the energy density and logarithm value ideal average period for obtaining each IMF are closed
It is formulaMake energy density and logarithm two-dimentional relation figure average period；
Average period and the energy density logarithm of each IMF components obtained after white noise EMD is decomposed are calculated, scatterplot is represented
In two-dimentional relation figure；
White noise IMF energy density Normal Distribution is normalized, the free degree is equal to average energy；Therefore, for returning
One changes any c of white noise time seriesl(m), lnElConfidential interval beInstitute
It is the confidential interval under the conditions of 99% to state and make confidence level with dotted line in two-dimentional relation figure, and this figure is united as based on white noise
Count the template that characteristic chooses IMF components；
The equal length mixed signal that single channel is collected carries out EMD decomposition, calculates average period and the energy of each IMF components
Metric density logarithm, scatter diagram is represented in white noise statistical property template；It will fall in confidential interval or borderline IMF points
Amount is removed as high-frequency noise and low frequency chaff component, and fall the representative of each IMF components beyond confidential interval is exactly actually
Mechanical oscillation pattern；
The kurtosis value of each IMF components is calculated, when the kurtosis value of IMF components is more than 3, is represented in IMF components containing more
Impact composition, this principle is combined with white noise statistical property, IMF components are chosen.
Further technical scheme is that described combination CICA algorithms carry out mixed signal blind source separating, extract sense
The method of the fault-signal of interest is as follows：
Fault-signal characteristic frequency based on known bearing and gear, construction reference burst signal r (t) is defined to be extracted
Independent element y and reference signal r (t) distance function be ε (y, r)；
Then have with lower inequality：
Wherein：w*It is intended to the mixed vector of the corresponding optimal solution of independent element extracted, wi(i=1,2 ..., l-1) it is other only
Vertical composition is corresponding to solve mixed vector；Then following constraint functions has and only in y=w*TIt is true during x：
G (y)=ε (y, r)-ξ≤0
Wherein：For threshold parameter, g (y) expression formula is substituted into, obtain constraint it is independent into
Divide algorithm, i.e. CICA algorithms are as follows：
max J(y)≈ρ{E[G(y)]-E[G(v)]}2
In formula：J (y) represents the object function on negentropy；G (y) is constraint function；H (y) and h (r) is to make constraint respectively
The independent element y and reference signal r of output have unit variance；Above formula is actually that a constrained optimization of object function is asked
Topic, can be solved by lagrange's method of multipliers, obtain the best estimate of target source signal, be letter interested by target source signal
Number extract.
Further technical scheme is that the ε (y, r) uses mean square error ε (y, r)=E { (y-r)2Measurement, or use phase
Close function of ε (y, r)=- E { yr } measurements.
It is using the beneficial effect produced by above-mentioned technical proposal：Methods described effectively realizes single channel mixed fault
The problem of pretreatment of signal blind source separating, the effect of single channel extension and signal de-noising can be reached simultaneously, with CICA blind source separatings
Algorithm is combined, and more accurately extracts fault-signal interested, improves the specific aim and effect of fault diagnosis, is that one kind has
The gear-box mixed fault diagnostic method of effect.
Brief description of the drawings
The present invention is further detailed explanation with reference to the accompanying drawings and detailed description.
The main flow chart of Fig. 1 methods describeds of the embodiment of the present invention；
Fig. 2 methods described axis bearing outer-ring fault simulation signal time-domain diagrams of the embodiment of the present invention；
Fig. 3 methods described middle gear fault simulation signal time-domain diagrams of the embodiment of the present invention；
Fig. 4 is random white noise time-domain diagram in methods described of the embodiment of the present invention；
Mixed signal time-domain diagram in Fig. 5 methods describeds of the embodiment of the present invention；
The white noise statistical property figure of each IMF components in Fig. 6 methods describeds of the embodiment of the present invention；
The kurtosis value distribution map of each IMF components in Fig. 7 methods describeds of the embodiment of the present invention；
Original signal and reconstruction signal comparison diagram in Fig. 8 methods describeds of the embodiment of the present invention；
The IMF component time domain beamformers extracted in Fig. 9 methods describeds of the embodiment of the present invention
Reference signal and the fault-signal time-domain diagram isolated in Figure 10 methods describeds of the embodiment of the present invention；
The simulated fault signal envelope spectrogram isolated in Figure 11 methods describeds of the embodiment of the present invention.
Embodiment
With reference to the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Ground is described, it is clear that described embodiment is only a part of embodiment of the present invention, rather than whole embodiments.It is based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of creative work is not made
Embodiment, belongs to the scope of protection of the invention.
Many details are elaborated in the following description to facilitate a thorough understanding of the present invention, still the present invention can be with
It is different from other manner described here using other to implement, those skilled in the art can be without prejudice to intension of the present invention
In the case of do similar popularization, therefore the present invention is not limited by following public specific embodiment.
As shown in figure 1, the present invention is a kind of method for realizing gear-box mixed fault signal blind source separating, methods described bag
Include following steps：
S101：Single channel gear-box mixed fault signal is gathered by single acceleration transducer, to the single-pass collected
Road gear-box mixed fault signal carries out EMD pretreatments, to realize that signal de-noising and single channel extend.
S102：IMF components are chosen using white noise statistical property and kurtosis value associated methods, effective vibration mode is used as
Component.
S103：Packet is reconstructed to selected IMF components, as the input signal of blind source separating, (constraint is only with reference to CICA
Vertical PCA) algorithm progress mixed signal blind source separating, it can extract out fault-signal interested.
Such as mixed fault signal includes bearing, gear distress signal and harmonic signal and noise signal, is carrying out
During CICA algorithms, corresponding reference signal is set up according to bearing fault characteristics frequency, just can be from input signal by bearing event
Barrier time-domain signal separation (extraction) comes out, and the bearing fault signal extracted is fault-signal interested.
S104：To separating the trouble unit drawn progress Hilbert Envelope Analysis through CICA methods, (CICA separation is drawn
Fault-signal is time domain waveform, it is necessary to which carry out that frequency-domain analysis obtains that failure-frequency just can verify that separating resulting obtains is interested
Fault-signal, carry out the envelope spectrum that Hilbert Envelope Analysis can obtain signal to it, envelope spectrum can reflect separation signal
Spectrum information), to obtain its envelope spectrum, it is diagnosable go out mixed fault in fault signature interested, realize blind source separating.
Specifically, being based on carrying out roller box mixed fault signal EMD pretreatments and using in vain in described step S101
Noise statisticses and kurtosis value associated methods choose IMF components, and the specific method as effective vibration mode component is as follows：
1) white noise statistic feature extraction mechanical oscillation pattern is based on, to any normalization white noise time series nm, m=
1 ..., N.It is expressed as after carrying out EMD decompositionN is natural number, and white noise sequence is decomposed by EMD
Afterwards, w IMF component, c are decomposed into altogetherl(m) it is l-th of IMF component.
2) c is definedl(m) energy density is
3)cl(m) the average period corresponding to fourier spectra is
4) it can show that EMD decomposes obtained any IMF energy density and the pass between average period by the derivation of equation
It is to beLogarithm using e the bottom of as is taken to formula both sides, the energy density and logarithm value average period for obtaining each IMF are managed
Think relational expressionMake energy density and logarithm two-dimentional relation figure average period.
5) average period and the energy density logarithm of each IMF components obtained after white noise EMD is decomposed are calculated, scatterplot is represented
In two-dimentional relation figure.
6) normalization white noise IMF energy density Normal Distribution, the free degree is equal to average energy.Therefore, for
Normalize any c of white noise time seriesl(m), lnElConfidential interval be
It is the confidential interval under the conditions of 99% to make confidence level with dotted line in above-mentioned X-Y scheme, and this figure counts special as based on white noise
Property choose IMF components template.
7) the equal length mixed signal collected single channel carries out EMD decomposition, calculate each IMF components average period and
Energy density logarithm, scatter diagram is represented in white noise statistical property template.It will fall in confidential interval or borderline IMF
Component is removed as high-frequency noise and low frequency chaff component, and falls the actual of the representatives of each IMF beyond confidential interval
Mechanical oscillation pattern.
8) each IMF kurtosis value is calculated, when IMF kurtosis value is more than 3, represents to contain more impact composition in IMF,
This principle is combined with white noise statistical property, more accurate choose is carried out to IMF components.
CICA algorithms progress mixed signal blind source separating is combined in described step S103 to comprise the following steps：
Fault-signal characteristic frequency based on known bearing and gear, construction reference burst signal r (t) is defined to be extracted
Independent element y and reference signal r (t) distance function be ε (y, r).ε (y, r) can use mean square error ε (y, r)=E { (y-r
)2Measurement, it is also possible to correlation function ε (y, r)=- E { yr } is measured.Then have with lower inequality：
Wherein：w*It is intended to the mixed vector of the corresponding optimal solution of independent element extracted, wi(i=1,2 ..., l-1) it is other only
Vertical composition is corresponding to solve mixed vector.Then following constraint functions has and only in y=w*TIt is true during x：
G (y)=ε (y, r)-ξ≤0 (5)
Wherein：For threshold parameter.Wushu (5) is substituted into, and can must constrain independent element
(CICA) algorithm is as follows：
In formula：J (y) represents that the object function g (y) on negentropy is constraint function；H (y) and h (r) is to make constraint respectively
The independent element y and reference signal r of output have unit variance；Above formula is actually that a constrained optimization of object function is asked
Topic, can be solved by lagrange's method of multipliers, obtain the best estimate of target source signal, target source extraction is come out into (mesh
Mark source signal be signal interested, such as source signal includes gear, bearing fault signal, if using bearing fault signal as
The echo signal extracted, i.e. bearing fault signal are wanted for signal interested).
Validity of the extracting method in analysis gearbox fault feature extraction in order to verify, so as to construct one group of emulation letter
Number.Wherein, s1For bearing outer ring fault simulation signal, s2Signal, s are emulated for gear distress3For the white noise signal being mixed into, s1
And s2As shown in formula (7).Fig. 2 is the time-domain diagram of bearing outer ring fault simulation signal, and Fig. 3 is the time domain that gear distress emulates signal
Figure, Fig. 4 is the time-domain diagram for the white noise signal being mixed into.
Wherein, sample frequency is 25600Hz, and a=800 is attenuation rate, A=2ms-2For impact amplitude, t=1s is emulation
Duration, f1=3kHz is resonant frequency, f caused by impact2=240Hz is failure gear mesh frequency, and the failure for emulating bearing is special
Levy frequency fmFor 55Hz, simulated fault gear amplitude modulation frequency fzFor 20Hz.
Single-channel data collection is emulated, above signal is mixed, the time-domain diagram of hybrid simulation signal is as shown in Figure 5.
EMD decomposition is carried out to the mixed signal after normalized, 13 IMF components is obtained, the energy of each IMF components is calculated respectively
Density and average period.The normalization white noise sample of construction and mixed signal equal length, will emulate each IMF that signal decomposition goes out
The energy density and logarithm value average period (corresponding point is represented with *) of component are contrasted with the confidential interval, as a result such as Fig. 6 institutes
Show, can substantially observe that the point for representing IMF2, IMF5, IMF6 and IMF7 falls outside confidential interval.Each IMF kurtosis value is calculated,
As shown in Figure 7, it can be seen that IMF2, IMF5, IMF6, IMF7 and IMF11 kurtosis value are more than 3, then calculate IMF2, IMF5 respectively,
IMF6, IMF7 and IMF11 and original signal correlation coefficient value, as shown in table 1, it is seen that IMF11 correlation coefficient value is relatively small,
Thus judge that IMF2, IMF5, IMF6 and IMF7 belong to vibration mode component (the i.e. effective component for including mechanical oscillation information).
The IMF components selected are reconstructed, its time-domain diagram is contrasted with original signal as shown in figure 8, it will be evident that to letter from figure
Number carry out EMD and decompose to reconstruct the noise reduction process that realizes to a certain extent again, impact composition is more obvious.
The coefficient correlation of table 1-IMF components and mixed signal
| IMF components | IMF2 | IMF5 | IMF6 | IMF7 | IMF11 |
| Coefficient correlation | 0.5800 | 0.2821 | 0.6258 | 0.2574 | 0.0368 |
Blind source separating is carried out to the IMF components extracted with CICA algorithms, bearing and gear distress are emulated signal by purpose
Separated from the IMF component signals of selection.Such as Fig. 9, IMF2, IMF5, IMF6 and IMF7 are regard as input signal, reference axis
Hold and gear simulated fault characteristic frequency, set up the reference signal of square, and the bearing that is extracted using CICA methods and
Gear simulated fault time domain plethysmographic signal is as shown in Figure 10.Hilbert Envelope Analysis is done respectively to the fault-signal isolated, obtained
To Hilbert envelope spectrums as shown in figure 11, corresponding bearing simulated fault frequency 55Hz and again is can clearly be seen that from figure
Frequently, and gear simulated fault modulating frequency 20Hz, demonstrate the correct of above-mentioned selection IMF components methods and CICA algorithms
Property.Test data result in the embodiment of the present invention demonstrates the validity of methods described.
Methods described effectively realizes the problem of pretreatment of single channel mixed fault signal blind source separating, and list can be reached simultaneously
Passage extends the effect with signal de-noising, is combined with CICA blind source separation algorithms, more accurately extracts failure letter interested
Number, the specific aim and effect of fault diagnosis are improved, is a kind of effective gear-box mixed fault diagnostic method.
Claims (6)
1. one kind realizes gear-box mixed fault signal blind source separation method, it is characterised in that comprise the following steps：
EMD pretreatments are carried out to the single channel gear-box mixed fault signal collected；
IMF components are chosen using white noise statistical nature and kurtosis value associated methods, effective vibration mode component is used as；
Packet is reconstructed to selected IMF components, as the input signal of blind source separating, mixed signal is carried out with reference to CICA algorithms
Blind source separating, extracts fault-signal interested；
Hilbert Envelope Analysis is carried out to separating the fault-signal interested drawn through CICA methods, its envelope spectrum is obtained, examines
Break and fault signature interested in mixed fault, realize blind source separating.
2. realize gear-box mixed fault signal blind source separation method as claimed in claim 1, it is characterised in that methods described
In pass through single acceleration transducer gather single channel gear-box mixed fault signal.
3. realize gear-box mixed fault signal blind source separation method as claimed in claim 1, it is characterised in that described pair
The method that the single channel mixed fault signal collected carries out EMD pretreatments is as follows：
Based on white noise statistic feature extraction mechanical oscillation pattern, to any normalization white noise time series nm, m=1 ..., N,
It is expressed as after carrying out EMD decompositionN is natural number, and white noise sequence is decomposed into w altogether after EMD is decomposed
Individual IMF components, cl(m) it is l-th of IMF component.
4. realize gear-box mixed fault signal blind source separation method as claimed in claim 3, it is characterised in that described profit
IMF components are chosen with white noise statistical nature and kurtosis value associated methods, the method as effective vibration mode component is as follows：
Define cl(m) energy density is
cl(m) the average period corresponding to fourier spectra is
Show that EMD decomposes obtained any IMF energy density and the relation between average period is by the derivation of equationLogarithm using e the bottom of as is taken to formula both sides, the energy density and logarithm value ideal average period for obtaining each IMF are closed
It is formulaMake energy density and logarithm two-dimentional relation figure average period；
Average period and the energy density logarithm of each IMF components obtained after white noise EMD is decomposed are calculated, scatterplot is represented in two dimension
In graph of a relation；
White noise IMF energy density Normal Distribution is normalized, the free degree is equal to average energy；Therefore, for normalization
Any c of white noise time seriesl(m), lnElConfidential interval beDescribed two
It is the confidential interval under the conditions of 99% to make confidence level in dimension graph of a relation with dotted line, and this figure is counted into special as based on white noise
Property choose IMF components template；
The equal length mixed signal that single channel is collected carries out EMD decomposition, and average period and the energy for calculating each IMF components are close
Logarithm is spent, scatter diagram is represented in white noise statistical property template；It will fall in confidential interval or borderline IMF components are made
Be that high-frequency noise and low frequency chaff component are removed, and fall the representative of each IMF components beyond confidential interval be exactly reality machine
Tool vibration mode；
The kurtosis value of each IMF components is calculated, when the kurtosis value of IMF components is more than 3, represents to contain more rush in IMF components
Composition is hit, this principle is combined with white noise statistical property, IMF components are chosen.
5. realize gear-box mixed fault signal blind source separation method as claimed in claim 1, it is characterised in that described knot
Close CICA algorithms and carry out mixed signal blind source separating, the method for extracting fault-signal interested is as follows：
Fault-signal characteristic frequency based on known bearing and gear, construction reference burst signal r (t) is defined to be extracted only
Vertical composition y and reference signal r (t) distance function are ε (y, r)；
Then have with lower inequality：
<mrow>
<mi>&epsiv;</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>w</mi>
<mrow>
<mo>*</mo>
<mi>T</mi>
</mrow>
</msup>
<mi>x</mi>
<mo>,</mo>
<mi>r</mi>
<mo>)</mo>
</mrow>
<mo><</mo>
<mi>&epsiv;</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>w</mi>
<mn>1</mn>
<mi>T</mi>
</msubsup>
<mi>x</mi>
<mo>,</mo>
<mi>r</mi>
<mo>)</mo>
</mrow>
<mo>&le;</mo>
<mo>...</mo>
<mo>&le;</mo>
<mi>&epsiv;</mi>
<mrow>
<mo>(</mo>
<msubsup>
<mi>w</mi>
<mrow>
<mi>l</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mi>T</mi>
</msubsup>
<mi>x</mi>
<mo>,</mo>
<mi>r</mi>
<mo>)</mo>
</mrow>
</mrow>
Wherein：w*It is intended to the mixed vector of the corresponding optimal solution of independent element extracted, wi(i=1,2 ..., l-1) be other independences into
Divide the corresponding mixed vector of solution；Then following constraint functions has and only in y=w*TIt is true during x：
G (y)=ε (y, r)-ξ≤0
Wherein：For threshold parameter, g (y) expression formula is substituted into, constraint independent element is obtained and calculates
Method, i.e. CICA algorithms are as follows：
max J(y)≈ρ{E[G(y)]-E[G(v)]}2
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
<mo>.</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mi>g</mi>
<mo>(</mo>
<mi>y</mi>
<mo>)</mo>
<mo>&le;</mo>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>h</mi>
<mo>(</mo>
<mi>y</mi>
<mo>)</mo>
<mo>=</mo>
<mi>E</mi>
<mo>(</mo>
<msup>
<mi>y</mi>
<mn>2</mn>
</msup>
<mo>)</mo>
<mo>-</mo>
<mn>1</mn>
<mo>=</mo>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>h</mi>
<mo>(</mo>
<mi>r</mi>
<mo>)</mo>
<mo>=</mo>
<mi>E</mi>
<mo>(</mo>
<msup>
<mi>r</mi>
<mn>2</mn>
</msup>
<mo>)</mo>
<mo>-</mo>
<mn>1</mn>
<mo>=</mo>
<mn>0</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
In formula：J (y) represents the object function on negentropy；G (y) is constraint function；H (y) and h (r) is to export constraint respectively
Independent element y and reference signal r there is unit variance；Above formula is actually a constrained optimization problem to object function,
It can be solved by lagrange's method of multipliers, obtain the best estimate of target source signal, be signal interested by target source signal
Extract.
6. realize gear-box mixed fault signal blind source separation method as claimed in claim 1, it is characterised in that：The ε (y,
R) with mean square error ε (y, r)=E { (y-r)2Measurement, or measured with correlation function ε (y, r)=- E { yr }.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710504367.0A CN107255563A (en) | 2017-06-28 | 2017-06-28 | Realize gear-box mixed fault signal blind source separation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710504367.0A CN107255563A (en) | 2017-06-28 | 2017-06-28 | Realize gear-box mixed fault signal blind source separation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107255563A true CN107255563A (en) | 2017-10-17 |
Family
ID=60024313
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710504367.0A Pending CN107255563A (en) | 2017-06-28 | 2017-06-28 | Realize gear-box mixed fault signal blind source separation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107255563A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107748734A (en) * | 2017-10-31 | 2018-03-02 | 电子科技大学 | One kind parsing empirical mode decomposition method |
| CN107817083A (en) * | 2017-10-31 | 2018-03-20 | 南京工业大学 | A kind of more vibration source fault signal of mechanical equipment separators and method |
| CN108507783A (en) * | 2018-03-14 | 2018-09-07 | 湖南大学 | A kind of combined failure of rotating machinery diagnostic method decomposed based on group |
| CN108801630A (en) * | 2018-06-22 | 2018-11-13 | 石家庄铁道大学 | The gear failure diagnosing method of single channel blind source separating |
| CN109029973A (en) * | 2018-06-22 | 2018-12-18 | 石家庄铁道大学 | The method for realizing the diagnosis of single channel gear-box mixed fault |
| CN112113784A (en) * | 2020-09-22 | 2020-12-22 | 天津大学 | Equipment state monitoring method based on equipment acoustic signals and EMD |
| CN112179653A (en) * | 2020-09-07 | 2021-01-05 | 神华铁路装备有限责任公司 | Rolling bearing vibration signal blind source separation method and device and computer equipment |
| CN112945546A (en) * | 2021-01-20 | 2021-06-11 | 南京航空航天大学 | Accurate diagnosis method for complex fault of gear box |
| CN113074935A (en) * | 2021-04-01 | 2021-07-06 | 西华大学 | Acoustic separation and diagnosis method for impact fault characteristics of gearbox |
| CN113792657A (en) * | 2021-09-15 | 2021-12-14 | 西华大学 | Method for extracting acoustic signal identification and blind deconvolution algorithm gear box fault |
| CN113834645A (en) * | 2021-11-18 | 2021-12-24 | 长沙理工大学 | LOD-ICA-based rotary machine fault diagnosis method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105928702A (en) * | 2016-04-29 | 2016-09-07 | 石家庄铁道大学 | Variable working condition gear case bearing fault diagnosis method based on form component analysis |
| WO2017007333A1 (en) * | 2015-07-09 | 2017-01-12 | Mhwirth As | Condition monitoring method |
| CN106338395A (en) * | 2016-10-27 | 2017-01-18 | 石家庄铁道大学 | Gear case fault diagnosis method based on six-order monostable system |
-
2017
- 2017-06-28 CN CN201710504367.0A patent/CN107255563A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017007333A1 (en) * | 2015-07-09 | 2017-01-12 | Mhwirth As | Condition monitoring method |
| CN105928702A (en) * | 2016-04-29 | 2016-09-07 | 石家庄铁道大学 | Variable working condition gear case bearing fault diagnosis method based on form component analysis |
| CN106338395A (en) * | 2016-10-27 | 2017-01-18 | 石家庄铁道大学 | Gear case fault diagnosis method based on six-order monostable system |
Non-Patent Citations (5)
| Title |
|---|
| 吴小涛等: "基于峭度准则EEMD及改进形态滤波方法的轴承故障诊断", 《振动与冲击》 * |
| 李云飞等: "基于EMD的ICA语音增强", 《华中师范大学学报（自然科学版）》 * |
| 王玉娟: "基于EMD和ICA-R的旋转机械混合故障诊断方法的研究", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
| 胡爱军等: "基于集成经验模态分解和峭度准则的滚动轴承故障特征提取方法", 《中国电机工程学报》 * |
| 苏文胜等: "EMD降噪和谱峭度法在滚动轴承早期故障诊断中的应用", 《振动与冲击》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107748734B (en) * | 2017-10-31 | 2021-08-27 | 电子科技大学 | Analytic-empirical mode decomposition method |
| CN107817083A (en) * | 2017-10-31 | 2018-03-20 | 南京工业大学 | A kind of more vibration source fault signal of mechanical equipment separators and method |
| CN107748734A (en) * | 2017-10-31 | 2018-03-02 | 电子科技大学 | One kind parsing empirical mode decomposition method |
| CN108507783A (en) * | 2018-03-14 | 2018-09-07 | 湖南大学 | A kind of combined failure of rotating machinery diagnostic method decomposed based on group |
| CN108801630A (en) * | 2018-06-22 | 2018-11-13 | 石家庄铁道大学 | The gear failure diagnosing method of single channel blind source separating |
| CN109029973A (en) * | 2018-06-22 | 2018-12-18 | 石家庄铁道大学 | The method for realizing the diagnosis of single channel gear-box mixed fault |
| CN112179653A (en) * | 2020-09-07 | 2021-01-05 | 神华铁路装备有限责任公司 | Rolling bearing vibration signal blind source separation method and device and computer equipment |
| CN112113784A (en) * | 2020-09-22 | 2020-12-22 | 天津大学 | Equipment state monitoring method based on equipment acoustic signals and EMD |
| CN112945546A (en) * | 2021-01-20 | 2021-06-11 | 南京航空航天大学 | Accurate diagnosis method for complex fault of gear box |
| CN113074935A (en) * | 2021-04-01 | 2021-07-06 | 西华大学 | Acoustic separation and diagnosis method for impact fault characteristics of gearbox |
| CN113074935B (en) * | 2021-04-01 | 2022-09-13 | 西华大学 | Acoustic separation and diagnosis method for impact fault characteristics of gearbox |
| CN113792657A (en) * | 2021-09-15 | 2021-12-14 | 西华大学 | Method for extracting acoustic signal identification and blind deconvolution algorithm gear box fault |
| CN113834645A (en) * | 2021-11-18 | 2021-12-24 | 长沙理工大学 | LOD-ICA-based rotary machine fault diagnosis method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107255563A (en) | Realize gear-box mixed fault signal blind source separation method | |
| CN104483127B (en) | A kind of planetary gear feature information of weak faults extracting method | |
| CN108168891B (en) | Method and equipment for extracting weak fault signal characteristics of rolling bearing | |
| CN107356432A (en) | Fault Diagnosis of Roller Bearings based on frequency domain window experience small echo resonance and demodulation | |
| CN104132250B (en) | A kind of pipe leakage characteristic vector pickup method based on improvement wavelet packet | |
| CN107013449B (en) | The method and system of voice signal identification compressor fault based on deep learning | |
| CN109670485B (en) | Rotary machine local fault remote diagnosis method based on multi-data compression tracking algorithm | |
| CN105092239A (en) | Method for detecting early stage fault of gear | |
| WO2021027579A1 (en) | Fault diagnosis method under the effect of center frequency convergence trend | |
| CN105928702B (en) | Variable working condition box bearing method for diagnosing faults based on form PCA | |
| CN105760839A (en) | Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine | |
| CN105445022B (en) | A kind of planetary gear method for diagnosing faults based on dual-tree complex wavelet transform entropy Fusion Features | |
| CN103962888A (en) | Tool abrasion monitoring method based on wavelet denoising and Hilbert-Huang transformation | |
| CN104390781A (en) | Gear fault diagnosis method based on LMD and BP neural network | |
| CN104316163B (en) | Gear-box coupling modulation signal separating method based on interior product transformation and correlation filtering | |
| CN102519726A (en) | Acoustic-based diagnosis (ABD) method for compound fault of rolling bearing | |
| CN103900816A (en) | Method for diagnosing bearing breakdown of wind generating set | |
| CN106096562A (en) | Separate based on vibration signal blind sources and the gearbox of wind turbine method for diagnosing faults of sparse component analysis | |
| CN104714075B (en) | A kind of electric network voltage flicker envelope parameters extracting method | |
| CN109827777A (en) | Rolling bearing fault prediction technique based on Partial Least Squares extreme learning machine | |
| CN102521502B (en) | Wavelet packet-assisted self-adaption anti-aliasing ensemble empirical mode decomposition method | |
| CN109374119A (en) | Transformer vibration signal Characteristic Extraction method | |
| CN112287796B (en) | Radiation source identification method based on VMD-Teager energy operator | |
| CN106872171A (en) | A kind of adaptive learning bearing calibration of Doppler's acoustic signal | |
| CN107478418A (en) | A kind of rotating machinery fault characteristic automatic extraction method |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171017 |
|
| RJ01 | Rejection of invention patent application after publication |