CN107543722A - The Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth - Google Patents
The Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth Download PDFInfo
- Publication number
- CN107543722A CN107543722A CN201710710318.2A CN201710710318A CN107543722A CN 107543722 A CN107543722 A CN 107543722A CN 201710710318 A CN201710710318 A CN 201710710318A CN 107543722 A CN107543722 A CN 107543722A
- Authority
- CN
- China
- Prior art keywords
- msub
- mrow
- msubsup
- dictionary
- dictionary learning
- 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
Landscapes
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
Abstract
The Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth, fully utilize dictionary learning and the common feature of deep learning, extraction and diagnosis applied to rolling bearing cyclic breakdown shock characteristic, this method is overlapped dictionary learning and the greedy algorithm that successively stacks, the construction of self-adapting dictionary is carried out in a manner of hierarchical alterative, realize the extraction and expression to rolling bearing fault signal period property shock characteristic under very noisy, overcome the shortcomings that original dictionary learning can not effectively extract mechanical fault signature, obtain good feature extraction and diagnosis effect.
Description
Technical field
The present invention relates to rolling bearing fault diagnosis technical field, and the rolling of dictionary learning is more particularly to stacked based on depth
Bearing fault characteristics extracting method.
Background technology
Rolling bearing is a kind of common universal machine part, has that running precision is high, substitutability is good, cheap etc.
Series of advantages, it is widely used in the industries such as oil, chemical industry, ship, mining.Rolling bearing is as connection rotation in plant equipment
" joint " of part and fixed component, its health status directly affect the running quality of whole set equipment.Once rolling bearing is sent out
Raw failure, gently then causes plant equipment to be shut down, heavy then bring huge property loss or even casualties, therefore, to rolling bearing
Carry out health monitoring and fault diagnosis is significant.Analysis of vibration signal is the effective hand for carrying out rolling bearing fault diagnosis
Section, in the presence of ambient noise and other interference, it is special that faint failure can be effectively extracted from test signal
Sign, and then whether accurate judgement bearing breaks down and fault type.
In recent years, substantial amounts of Feature Extraction Technology has been used for the fault diagnosis of rolling bearing, including：Referred to based on statistics
Target Time Domain Analysis, such as kurtosis, root-mean-square value, degree of skewness；Frequency-domain analysis method based on Fourier transformation；And with
Short Time Fourier Transform and wavelet transformation are Time-Frequency Analysis Method of representative etc..But strong ambient noise and non-stationary fortune
Row condition often influences the diagnosis capability of these classical ways, therefore many advanced signal processing methods are by numerous studies,
Such as compose kurtosis, minimum entropy deconvolution.Recently, a kind of dictionary learning method based on sparse representation is based on deep learning with a kind of
Artificial intelligence approach attracted substantial amounts of scholar, in image procossing, speech processes have been achieved for great achievement.
But substantially all dictionary learning models used are all individual layers at present, use it for signal noise silencing and feature carries
Taking to achieve satisfactory results, and deep learning model can successful wherein critically important factor be it
In a manner of successively stacking greediness, the more complicated characteristic information of more higher-dimension is constantly extracted, can be obtained more relative to single-layer model
Good effect.
The content of the invention
The shortcomings that in order to overcome above-mentioned existing dictionary learning technology, it is an object of the invention to propose to be based on depth stacking
The Rolling Bearing Fault Character extracting method of dictionary learning, the related spy of failure is adaptively extracted by way of stacking and learning
Reference ceases, and the extracted in self-adaptive of Rolling Bearing Fault Character can be achieved, effectively suppress the interference of ambient noise, hence it is evident that improve and roll
The accuracy of bearing failure diagnosis.
In order to achieve the above object, the technical scheme taken of the present invention is：
The Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth, is comprised the following steps：
Step 1, vibration acceleration sensor is adsorbed on the bearing block of tested rolling bearing, signal carried out high
Frequency sampling, obtain vibration data；
Step 2, according to sample frequency and bearing rotary speed, intercept the signal in a period of time and carry out bandpass filtering, make
For primary signal y (t)；
Step 3, multilayer dictionary learning algorithm is built, the parameter of different layers dictionary learning is set：Iterations, matching word
The signal segmentation length of allusion quotation, dictionary atom number, initialize dictionary；
Step 4：Primary signal y (t) is inputted, according to corresponding parameter setting application dictionary learning algorithm, through sparse volume
Code, by column renewal dictionary and coefficient matrix obtain final dictionary and corresponding sparse representation coefficient matrix, and pass through addition pair
Number likelihood item optimization, tries to achieve the time-domain signal after filtering de-noising
Step 5, by time-domain signalAs the input of next layer of dictionary learning algorithm, repeat identical with step 4
Algorithm, the like, the dictionary learning algorithm until completing whole layers；
Step 6, envelope spectrum analysis, the time domain output signal of last layer is obtained, done soon after Hilbert transform is done to it
Fast Fourier transformation, corresponding frequency spectrum is obtained, export diagnostic result.
Interception time in described step two includes multiple swing circles, is easy to highlight at intervals of t, period planted agent
Periodic shock feature.
The detailed process of described step four is：
Each layer uses K-SVD dictionary learning algorithms, solves following optimization problem：
In formula, y is the signal of input, and D be dictionary matrix, and x is sparse representation coefficient matrix, and i is x line number, xiFor x squares
I-th row of battle array, ε are approximate error, T0For the sparse constraint factor, | | | |FF norms are represented, | | | |0Represent 0 norm；First
It is assumed that D is fixed, above-mentioned optimization problem is converted to search for optimal coefficient matrix x, is solved such as using match tracing optimized algorithm
Lower problem obtains sparse representation coefficient：
Then dictionary d is updated by columnkAnd the coefficient x of corresponding rowk, it is as follows to rewrite formula (1)：
D in formulakRepresent D kth row, xkRepresent x row k, defined parametersφkFor
Size N × | ωk| matrix, i.e., in (ωk(i), i) value be 1, remaining is 0, and therefore, formula (3) multiplies φkIt can be written as
It is rightSingular value decomposition is carried out to obtainAnd then the dictionary d of the first row renewal respective column using Uk, profit
Coefficient of correspondence is updated with V first row and the product of Δ (1,1)The like, until the dictionary of K row has all updated；For
The further precision for improving characteristic signal, adds log-likelihood item, and newer (1) is
In formula, z is noise cancellation signal, and λ is Lagrange multiplier, RiIt is the corresponding operator extracted from z, Section 1 represents warp
Dictionary learning filtered data fidelity is crossed, Section 2 and Section 3 are noise cancellation signal z priori items；Pass through foregoing K-SVD
Algorithm obtain D andAfterwards, formula (5) simplification is as follows：
Solved using the optimization enclosed of quadratic problem
The present invention has the advantages that compared to prior art：
A) present invention proposes that depth stacks dictionary learning algorithm, fully combines the learning performance of different size dictionaries, has
The characteristic signal related beneficial to bearing fault is found.
B) instant invention overcomes individual layer dictionary learning Algorithm Learning pattern it is single the shortcomings that, there is the adaptive spy of multiple features
Point.
C) present invention does not need priori, directly handles primary signal, is advantageously implemented Rolling Bearing Fault Character certainly
Adapt to extraction and the automation of diagnostic monitoring.
Brief description of the drawings
Fig. 1 is test platform structure schematic diagram of the embodiment of the present invention.
Fig. 2 is rolling bearing inner ring failure of the embodiment of the present invention.
Fig. 3 is the flow chart of the inventive method.
Fig. 4 is the original vibration signal of the embodiment of the present invention.
Fig. 5 is the envelope spectrum of the primary signal of the embodiment of the present invention.
Fig. 6 is the primary signal that the embodiment of the present invention obtains.
Fig. 7 is the envelope spectrum signal that the embodiment of the present invention obtains.
Fig. 8 is the time-domain signal of embodiment individual layer K-SVD extractions.
Fig. 9 is the envelope spectrum signal of embodiment individual layer K-SVD extractions.
Embodiment
The present invention is described in detail with reference to the accompanying drawings and examples.
By taking the locomotive rolling bearing fault detect testing stand of certain rolling stock section as an example, the rolling bearing test table is by motor
1st, driving wheel 2, rolling bearing 3, wheel form to 4, rolling bearing 5, as shown in figure 1, motor 1 drives driving wheel 2 to rotate, drive
Driving wheel 2 contacts with the outer ring of tested rolling bearing 3 and drives outer ring to rotate, and rolling bearing 5 and wheel are fixed to 4.
Design parameter is as follows：1) contact angle of rolling bearing 3：9°；2) the rolling element diameter of rolling bearing 3：
23.775mm；3) the rolling element number of rolling bearing 3：20；4) pitch diameter of rolling bearing 3 is：180mm；5) event of rolling bearing 3
Barrier type is inner ring spalling failure, as shown in Figure 2；6) vibrating sensor M is arranged on shaft end；7) test system is to vibration signal
High frequency sampling and data storage are carried out, the frequency of sampling process is 76800Hz, sampling time 15s.
Fault diagnosis is carried out to rolling bearing, determines bearing fault type and and list to primary data analysis using the present invention
Layer dictionary learning method is contrasted.
As shown in figure 3, stacking the bearing fault characteristics extracting method of dictionary learning based on depth, comprise the following steps：
Step 1, vibration acceleration meter is installed on to the fixed position of the tested shaft end of rolling bearing 3, vibration signal is entered
The high frequency sampling of row；
Step 2, according to rotating speed and sample frequency, 0.5s data are intercepted, and carry out bandpass filtering, filter out bearing mnanufacture
Low-frequency vibration caused by rigging error and the interference component of other structures vibration, obtain primary signal y (t), as shown in figure 4,
Primary signal y (t) envelope spectrum from this two figure as shown in figure 5, can not find obvious fault signature；
Step 3, multilayer dictionary learning algorithm is built, the parameter of different layers dictionary learning is set：Iterations, matching word
The signal segmentation length of allusion quotation, dictionary atom number, initialize dictionary；
Step 4, input primary signal y (t), according to corresponding parameter setting, using dictionary learning algorithm, through sparse volume
Code, renewal dictionary and coefficient matrix two benches obtain final dictionary and corresponding sparse representation coefficient matrix by column, and pass through
Add the optimization of log-likelihood item and try to achieve the time-domain signal after filtering de-noisingSpecially：
Each layer uses K-SVD dictionary learning algorithms, solves following optimization problem：
In formula, D is dictionary matrix, and x is sparse representation coefficient matrix.Assume first that dictionary D is fixed, convert above-mentioned optimization
Problem solves following problem to search for optimal coefficient matrix x, using match tracing optimized algorithm can obtain sparse representation coefficient：
Then dictionary d is updated by columnkAnd the coefficient x of corresponding rowk, it is as follows to rewrite formula (1)：
DefinitionφkFor size N × | ωk| matrix, i.e., in (ωk(i), i) value be 1,
Remaining is 0, and therefore, formula (3) can be written as again
It is rightCarrying out singular value decomposition can obtainAnd then the dictionary d of the first row renewal respective column using Uk,
Coefficient is updated using V first row and the product of Δ (1,1)The like, until the dictionary of K row has all updated；Add
Log-likelihood item further improves the precision of the signal characteristic of extraction, and newer (1) is
In formula, Section 1, which represents, passes through dictionary learning filtered data fidelity, and Section 2 and Section 3 are de-noising letters
Number z priori item, RiIt is the corresponding operator extracted from z, λ is Lagrange multiplier；By foregoing K-SVD algorithms obtain D and
Afterwards, formula (5), which can simplify, is as follows：
Solve and can obtain using the optimization enclosed of quadratic problem
Step 5, by time-domain signalAs the input of next layer of dictionary learning algorithm, repeat identical with step 4
Algorithm, the like, the dictionary learning algorithm until completing whole layers；
Step 6, envelope spectrum analysis, the time domain output signal of last layer is obtained, done soon after Hilbert transform is done to it
Fast Fourier transformation, corresponding frequency spectrum is obtained, export diagnostic result.
The time domain beamformer and envelope that method based on depth stacking dictionary learning extraction Rolling Bearing Fault Character obtains
As shown in Figure 6, Figure 7, envelope spectrum is the harmonic wave of inner ring fault characteristic frequency to spectrum, coincide, is realized to failure with fault signature
Accurate Diagnosis.And the result feature of conventional monolayers dictionary extraction rolling bearing fault is as shown in Figure 8 and Figure 9, fail to find inner ring
The failure of generation.
The extraction Rolling Bearing Fault Character method proposed by the present invention that dictionary learning is stacked based on depth overcomes individual layer
The defects of dictionary learning, fault characteristic information is extracted, efficient diagnosis has been carried out to failure, there is good robustness.
Claims (3)
1. the Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth, it is characterised in that comprise the following steps：
Step 1, vibration acceleration sensor is adsorbed on the bearing block of tested rolling bearing, carrying out high frequency to signal adopts
Sample, obtain vibration data；
Step 2, according to sample frequency and bearing rotary speed, intercept the signal in a period of time and carry out bandpass filtering, as original
Beginning signal y (t)；
Step 3, multilayer dictionary learning algorithm is built, the parameter of different layers dictionary learning is set：Iterations, match dictionary
Signal segmentation length, dictionary atom number, initialize dictionary；
Step 4：Input primary signal y (t), according to corresponding parameter setting application dictionary learning algorithm, through sparse coding, by
Row renewal dictionary and coefficient matrix obtain final dictionary and corresponding sparse representation coefficient matrix, and by adding log-likelihood
Item optimization, try to achieve the time-domain signal after filtering de-noising
Step 5, by time-domain signalAs the input of next layer of dictionary learning algorithm, repetition is calculated with identical in step 4
Method, the like, the dictionary learning algorithm until completing whole layers；
Step 6, envelope spectrum analysis, the time domain output signal of last layer is obtained, quick Fu is after Hilbert transform is done to it
In leaf transformation, obtain corresponding frequency spectrum, export diagnostic result.
2. the Rolling Bearing Fault Character extracting method according to claim 1 that dictionary learning is stacked based on depth, it is special
Sign is：Interception time in described step two includes multiple swing circles, is easy to highlight at intervals of t, period planted agent
Periodic shock feature.
3. the Rolling Bearing Fault Character extracting method according to claim 1 that dictionary learning is stacked based on depth, it is special
Sign is：The detailed process of described step four is：
Each layer uses K-SVD dictionary learning algorithms, solves following optimization problem：
<mrow>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<mrow>
<mi>D</mi>
<mo>,</mo>
<mi>x</mi>
</mrow>
</munder>
<mo>|</mo>
<mo>|</mo>
<mi>y</mi>
<mo>-</mo>
<mi>D</mi>
<mi>x</mi>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>&le;</mo>
<mi>&epsiv;</mi>
<mi>s</mi>
<mi>u</mi>
<mi>b</mi>
<mi>j</mi>
<mi>u</mi>
<mi>c</mi>
<mi>t</mi>
<mi> </mi>
<mi>t</mi>
<mi>o</mi>
<mo>&ForAll;</mo>
<mi>i</mi>
<mo>,</mo>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>0</mn>
</msub>
<mo>&le;</mo>
<msub>
<mi>T</mi>
<mn>0</mn>
</msub>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, D is dictionary matrix, and x is sparse representation coefficient matrix；Assume first that D is fixed, convert above-mentioned optimization problem to search
The optimal coefficient matrix x of rope, solve following problem using match tracing optimized algorithm and obtain sparse representation coefficient：
<mrow>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
</munder>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>Dx</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>&le;</mo>
<mi>&epsiv;</mi>
<mi>s</mi>
<mi>u</mi>
<mi>b</mi>
<mi>j</mi>
<mi>u</mi>
<mi>c</mi>
<mi>t</mi>
<mi> </mi>
<mi>t</mi>
<mi>o</mi>
<mo>&ForAll;</mo>
<mi>i</mi>
<mo>,</mo>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>0</mn>
</msub>
<mo>&le;</mo>
<msub>
<mi>T</mi>
<mn>0</mn>
</msub>
<mo>,</mo>
<mi>f</mi>
<mi>o</mi>
<mi>r</mi>
<mi> </mi>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<mi>N</mi>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
Then dictionary d is updated by columnkAnd the coefficient x of corresponding rowk, it is as follows to rewrite formula (1)：
<mrow>
<mo>|</mo>
<mo>|</mo>
<mi>y</mi>
<mo>-</mo>
<mi>D</mi>
<mi>x</mi>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>=</mo>
<mo>|</mo>
<mo>|</mo>
<mi>y</mi>
<mo>-</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>K</mi>
</munderover>
<msub>
<mi>d</mi>
<mi>j</mi>
</msub>
<msubsup>
<mi>x</mi>
<mi>T</mi>
<mi>j</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>=</mo>
<mo>|</mo>
<mo>|</mo>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>-</mo>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>&NotEqual;</mo>
<mi>k</mi>
</mrow>
</munder>
<msub>
<mi>d</mi>
<mi>j</mi>
</msub>
<msubsup>
<mi>x</mi>
<mi>T</mi>
<mi>j</mi>
</msubsup>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msub>
<mi>d</mi>
<mi>k</mi>
</msub>
<msubsup>
<mi>x</mi>
<mi>T</mi>
<mi>k</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>=</mo>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>E</mi>
<mi>k</mi>
</msub>
<mo>-</mo>
<msub>
<mi>d</mi>
<mi>k</mi>
</msub>
<msubsup>
<mi>x</mi>
<mi>T</mi>
<mi>k</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
DefinitionφkFor size N × | ωk| matrix, i.e., in (ωk(i), i) value be 1, remaining
For 0, therefore, formula (3) can be written as again
<mrow>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>E</mi>
<mi>k</mi>
</msub>
<msub>
<mi>&phi;</mi>
<mi>k</mi>
</msub>
<mo>-</mo>
<msub>
<mi>d</mi>
<mi>k</mi>
</msub>
<msubsup>
<mi>x</mi>
<mi>T</mi>
<mi>k</mi>
</msubsup>
<msub>
<mi>&phi;</mi>
<mi>k</mi>
</msub>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>=</mo>
<mo>|</mo>
<mo>|</mo>
<msubsup>
<mi>E</mi>
<mi>k</mi>
<mi>R</mi>
</msubsup>
<mo>-</mo>
<msub>
<mi>d</mi>
<mi>k</mi>
</msub>
<msubsup>
<mi>x</mi>
<mi>R</mi>
<mi>k</mi>
</msubsup>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mi>F</mi>
<mn>2</mn>
</msubsup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
It is rightSingular value decomposition is carried out to obtainAnd then the dictionary d of the first row renewal respective column using Uk, utilize V
First row and Δ (1,1) product renewal coefficientThe like, until the dictionary of K row has all updated；Add logarithm seemingly
Right item further improves the precision of the signal characteristic of extraction, and newer (1) is
<mrow>
<munder>
<mi>argmin</mi>
<mrow>
<mi>z</mi>
<mo>,</mo>
<msub>
<mrow>
<mo>{</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>}</mo>
</mrow>
<mi>i</mi>
</msub>
</mrow>
</munder>
<mi>&lambda;</mi>
<mo>|</mo>
<mo>|</mo>
<mi>y</mi>
<mo>-</mo>
<mi>z</mi>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msub>
<mi>&mu;</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>0</mn>
</msub>
<mo>+</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>Dx</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mi>z</mi>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>5</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, Section 1, which represents, passes through dictionary learning filtered data fidelity, and Section 2 and Section 3 are noise cancellation signal z
Priori item, RiIt is the corresponding operator extracted from z, λ is Lagrange multiplier；D and x is obtained by foregoing K-SVD algorithmskAfterwards, formula
(5) simplification is as follows：
<mrow>
<munder>
<mi>argmin</mi>
<mi>z</mi>
</munder>
<mi>&lambda;</mi>
<mo>|</mo>
<mo>|</mo>
<mi>z</mi>
<mo>-</mo>
<mi>y</mi>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<mo>|</mo>
<mo>|</mo>
<mi>D</mi>
<msub>
<mover>
<mi>x</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mi>z</mi>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
</mrow>
Solved using the optimization enclosed of quadratic problem
<mrow>
<mover>
<mi>z</mi>
<mo>^</mo>
</mover>
<mo>=</mo>
<msup>
<mrow>
<mo>(</mo>
<mi>&lambda;</mi>
<mi>I</mi>
<mo>+</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msubsup>
<mi>R</mi>
<mi>i</mi>
<mi>T</mi>
</msubsup>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mi>&lambda;</mi>
<mi>y</mi>
<mo>+</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msubsup>
<mi>R</mi>
<mi>i</mi>
<mi>T</mi>
</msubsup>
<mi>D</mi>
<msub>
<mover>
<mi>x</mi>
<mo>^</mo>
</mover>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>.</mo>
</mrow>
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710710318.2A CN107543722A (en) | 2017-08-18 | 2017-08-18 | The Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710710318.2A CN107543722A (en) | 2017-08-18 | 2017-08-18 | The Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107543722A true CN107543722A (en) | 2018-01-05 |
Family
ID=60957567
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710710318.2A Pending CN107543722A (en) | 2017-08-18 | 2017-08-18 | The Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107543722A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388692A (en) * | 2018-01-17 | 2018-08-10 | 西安交通大学 | Rolling Bearing Fault Character extracting method based on layering sparse coding |
| CN109060350A (en) * | 2018-09-04 | 2018-12-21 | 重庆大学 | A kind of Rolling Bearing Fault Character extracting method dictionary-based learning |
| CN109117896A (en) * | 2018-09-28 | 2019-01-01 | 西安交通大学 | A kind of Rolling Bearing Fault Character extracting method based on KSVD dictionary learning |
| CN110147637A (en) * | 2019-06-05 | 2019-08-20 | 厦门大学 | Based on the small impact-rub malfunction diagnostic method for involving the greedy sparse identification of harmonic components |
| CN110160790A (en) * | 2019-05-14 | 2019-08-23 | 中国地质大学（武汉） | A kind of rolling bearing fault impact signal extracting method and system based on improvement K-SVD |
| CN111582128A (en) * | 2020-04-30 | 2020-08-25 | 电子科技大学 | Mechanical fault sparse representation method based on wolf pack parameterized joint dictionary |
| CN112141837A (en) * | 2020-09-08 | 2020-12-29 | 金陵科技学院 | Intelligent voice elevator system based on multilayer dictionary learning |
| CN112710969A (en) * | 2020-12-18 | 2021-04-27 | 武汉大学 | Open-circuit fault diagnosis method for switching tube of single-phase half-bridge five-level inverter |
| CN113340598A (en) * | 2021-06-01 | 2021-09-03 | 西安交通大学 | Rolling bearing intelligent fault diagnosis method based on regularization sparse model |
| CN113743240A (en) * | 2021-08-12 | 2021-12-03 | 武汉钢铁有限公司 | Fault detection method, device and computer readable storage medium |
| CN113780055A (en) * | 2021-07-19 | 2021-12-10 | 北京交通大学 | MOMEDA and compressive sensing rolling bearing fault diagnosis method |
| CN113340598B (en) * | 2021-06-01 | 2024-05-28 | 西安交通大学 | Rolling bearing intelligent fault diagnosis method based on regularized sparse model |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101275702B1 (en) * | 2012-02-15 | 2013-06-17 | 목포대학교산학협력단 | Method and apparatus for fault diagnosis and fault detection of gearbox using vibration signal |
| CN103728130A (en) * | 2013-10-10 | 2014-04-16 | 西安交通大学 | Wind driven generator set failure feature extracting method based on sparse decomposition |
| CN105241666A (en) * | 2015-09-21 | 2016-01-13 | 华南理工大学 | Rolling bearing fault feature extraction method based on signal sparse representation theory |
| CN106441897A (en) * | 2016-10-27 | 2017-02-22 | 石家庄铁道大学 | Morphological component bearing failure diagnosis method based on dictionary study |
-
2017
- 2017-08-18 CN CN201710710318.2A patent/CN107543722A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101275702B1 (en) * | 2012-02-15 | 2013-06-17 | 목포대학교산학협력단 | Method and apparatus for fault diagnosis and fault detection of gearbox using vibration signal |
| CN103728130A (en) * | 2013-10-10 | 2014-04-16 | 西安交通大学 | Wind driven generator set failure feature extracting method based on sparse decomposition |
| CN105241666A (en) * | 2015-09-21 | 2016-01-13 | 华南理工大学 | Rolling bearing fault feature extraction method based on signal sparse representation theory |
| CN106441897A (en) * | 2016-10-27 | 2017-02-22 | 石家庄铁道大学 | Morphological component bearing failure diagnosis method based on dictionary study |
Non-Patent Citations (6)
| Title |
|---|
| AHARON, M.等: "K-SVD: An algorithm for designing overcomplete dictionaries for sparse representation", 《SIGNAL PROCESS》 * |
| MICHAEL ELAD: "《稀疏与冗余表示 理论及其在信号与图像处理中的应用》", 31 May 2015, 国防工业出版社 * |
| YONGZHI QU等: "Detection of Pitting in Gears Using a Deep Sparse Autoencoder", 《APPLIED SCIENCES》 * |
| 余发军等: "基于字典学习的轴承早期故障稀疏特征提取", 《振动与冲击》 * |
| 刘煜等: "《稀疏表示基础理论与典型应用》", 31 October 2014, 国防科学技术大学出版社 * |
| 张晗等: "基于稀疏分解理论的航空发动机轴承故障诊断", 《机械工程学报》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388692A (en) * | 2018-01-17 | 2018-08-10 | 西安交通大学 | Rolling Bearing Fault Character extracting method based on layering sparse coding |
| CN109060350A (en) * | 2018-09-04 | 2018-12-21 | 重庆大学 | A kind of Rolling Bearing Fault Character extracting method dictionary-based learning |
| CN109117896A (en) * | 2018-09-28 | 2019-01-01 | 西安交通大学 | A kind of Rolling Bearing Fault Character extracting method based on KSVD dictionary learning |
| CN109117896B (en) * | 2018-09-28 | 2020-07-24 | 西安交通大学 | Rolling bearing fault feature extraction method based on KSVD dictionary learning |
| CN110160790A (en) * | 2019-05-14 | 2019-08-23 | 中国地质大学（武汉） | A kind of rolling bearing fault impact signal extracting method and system based on improvement K-SVD |
| CN110160790B (en) * | 2019-05-14 | 2020-08-07 | 中国地质大学（武汉） | Rolling bearing fault impact signal extraction method and system based on improved K-SVD |
| CN110147637A (en) * | 2019-06-05 | 2019-08-20 | 厦门大学 | Based on the small impact-rub malfunction diagnostic method for involving the greedy sparse identification of harmonic components |
| CN111582128B (en) * | 2020-04-30 | 2022-05-03 | 电子科技大学 | Mechanical fault sparse representation method based on wolf pack parameterized joint dictionary |
| CN111582128A (en) * | 2020-04-30 | 2020-08-25 | 电子科技大学 | Mechanical fault sparse representation method based on wolf pack parameterized joint dictionary |
| CN112141837A (en) * | 2020-09-08 | 2020-12-29 | 金陵科技学院 | Intelligent voice elevator system based on multilayer dictionary learning |
| CN112710969A (en) * | 2020-12-18 | 2021-04-27 | 武汉大学 | Open-circuit fault diagnosis method for switching tube of single-phase half-bridge five-level inverter |
| CN113340598A (en) * | 2021-06-01 | 2021-09-03 | 西安交通大学 | Rolling bearing intelligent fault diagnosis method based on regularization sparse model |
| CN113340598B (en) * | 2021-06-01 | 2024-05-28 | 西安交通大学 | Rolling bearing intelligent fault diagnosis method based on regularized sparse model |
| CN113780055A (en) * | 2021-07-19 | 2021-12-10 | 北京交通大学 | MOMEDA and compressive sensing rolling bearing fault diagnosis method |
| CN113780055B (en) * | 2021-07-19 | 2024-02-09 | 北京交通大学 | MOMEDA and compression sensing rolling bearing fault diagnosis method |
| CN113743240A (en) * | 2021-08-12 | 2021-12-03 | 武汉钢铁有限公司 | Fault detection method, device and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107543722A (en) | The Rolling Bearing Fault Character extracting method of dictionary learning is stacked based on depth | |
| Han et al. | Multi-level wavelet packet fusion in dynamic ensemble convolutional neural network for fault diagnosis | |
| CN107526853B (en) | Rolling bearing fault mode identification method and device based on stacking convolutional network | |
| Jiao et al. | Hierarchical discriminating sparse coding for weak fault feature extraction of rolling bearings | |
| Zhao et al. | A weighted multi-scale dictionary learning model and its applications on bearing fault diagnosis | |
| Guo et al. | Envelope extraction based dimension reduction for independent component analysis in fault diagnosis of rolling element bearing | |
| Zhao et al. | Sparse classification based on dictionary learning for planet bearing fault identification | |
| Cui et al. | Matching pursuit of an adaptive impulse dictionary for bearing fault diagnosis | |
| Yang et al. | Bearing fault automatic classification based on deep learning | |
| CN111397896B (en) | Fault diagnosis method and system for rotary machine and storage medium | |
| CN108388692A (en) | Rolling Bearing Fault Character extracting method based on layering sparse coding | |
| Osman et al. | An enhanced Hilbert–Huang transform technique for bearing condition monitoring | |
| CN105760839A (en) | Bearing fault diagnosis method based on multi-feature manifold learning and support vector machine | |
| Liao et al. | Wavelet transform based convolutional neural network for gearbox fault classification | |
| Chen et al. | Fault feature extraction and diagnosis of gearbox based on EEMD and deep briefs network | |
| Zhang et al. | Sparsity-aware tight frame learning with adaptive subspace recognition for multiple fault diagnosis | |
| Chen et al. | Fault identification method for planetary gear based on DT-CWT threshold denoising and LE | |
| Grezmak et al. | Multi-stream convolutional neural network-based fault diagnosis for variable frequency drives in sustainable manufacturing systems | |
| Wang et al. | A deep neural network based on kernel function and auto-encoder for bearing fault diagnosis | |
| Yuan et al. | LW-Net: an interpretable network with smart lifting wavelet kernel for mechanical feature extraction and fault diagnosis | |
| CN113792597A (en) | Mechanical equipment abnormal sound detection method based on self-supervision feature extraction | |
| CN110044619B (en) | Multi-fault feature identification method based on sparse multi-cycle group lasso | |
| Wang et al. | Detrended fluctuation analysis and Hough transform based self-adaptation double-scale feature extraction of gear vibration signals | |
| CN114136604B (en) | Rotary equipment fault diagnosis method and system based on improved sparse dictionary | |
| Luo et al. | A fault diagnosis model based on LCD-SVD-ANN-MIV and VPMCD for rotating machinery |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180105 |