CN109029940A - Mechanical Multiple faults diagnosis approach based on signal atom driving small echo regeneration nuclear machine learning - Google Patents
Mechanical Multiple faults diagnosis approach based on signal atom driving small echo regeneration nuclear machine learning Download PDFInfo
- Publication number
- CN109029940A CN109029940A CN201810617661.7A CN201810617661A CN109029940A CN 109029940 A CN109029940 A CN 109029940A CN 201810617661 A CN201810617661 A CN 201810617661A CN 109029940 A CN109029940 A CN 109029940A
- Authority
- CN
- China
- Prior art keywords
- atom
- signal
- mechanical
- small echo
- machine 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
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Abstract
The present invention discloses a kind of mechanical multiple faults mode identification method based on signal atom driving small echo regeneration nuclear machine learning, the signal to be analyzed of this method collection machinery equipment, fault setting is carried out to mechanical equipment again, and acquires the unsteady mechanical vibration signal of faulty equipment；Unsteady mechanical vibration signal is tracked and decomposed using the Gabor atom orthogonal matching pursuit method based on ant colony search strategy, obtains the multiple atom components and residual component with signal best match to be analyzed；Machine learning is inputted using multiple atom components of acquisition as feature samples --- it is trained and tests study in small echo reproducing kernel support vector machine classifier, to identify mechanical multiple faults mode type.The present invention can be improved accuracy of identification.
Description
Technical field
The present invention relates to the mechanical Multiple faults diagnosis approach based on signal atom driving small echo regeneration nuclear machine learning, belong to
Technology for mechanical fault diagnosis field.
Background technique
Mechanical fault diagnosis is substantially the pattern recognition problem of state of runtime machine, and feature extraction and classifier design are
The key of pattern-recognition.
When carrying out diagnosis identification to mechanical breakdown, first have to carry out feature extraction to fault-signal.With signal processing
The development of technology, various new signal time frequency analysis are introduced into fault diagnosis field.Such as: Fourier transformation method, small echo become
Change method and Atomic Decomposition algorithm etc..Fourier transformation is to handle a kind of the most commonly used method of stationary signal, but use FF'
When the analysis of I' algorithm has the non-stationary signal of mechanical oscillation, then it is limited to the time-frequency resolution capability of algorithm, it cannot be accurate
Identify the feature of disturbing signal；Wavelet transformation is the local transformation of the time and frequency to signal, relatively for Fourier transformation,
Wavelet transformation can more extract effective information from signal, and practicability is stronger, but in wavelet transformation wavelet basis selection to point
It is very big to analyse result influence；Gabor Atomic Decomposition algorithm, it is the basic function in T/F-scale three-dimensional space, is combined
The characteristics of Gabor basic function and wavelet basis function, while avoiding the deficiency of the two.But how to construct for signal immanent structure
The atom dictionary of best match, improving computational efficiency is still current research hotspot.
In terms of Mechanical Fault Pattern Recognition, the machine learning based on Statistical Learning Theory --- support vector machines (SVM)
Show the unrivaled superior function of other learning machines.But it finds in the application, existing SVM kernel function cannot be passed through
Translation generates the one group of complete base of square integrable spatially, and the incompleteness of this base, which results in classification SVM, cannot approach this
Arbitrary classification interface spatially.On the other hand, SVM is a kind of based on kernel method for solving to be learnt from sample
New technology, and carrying out study from sample is an ill-posed problem (being mainly shown as multi-solution).Currently, by wavelet multiresolution analysis
Know, the flexible and translation of wavelet function can constitute one group of complete base in square integrable space.And regularization method is used to convert
It is solved for a well-posed problem, reproducing kernel and its corresponding Reproducing Kernel Hilbert Space are managed in function approximation and regularization
Important role is played in, and constructing Reproducing Kernel Function is the key that nuclear sparse expression disaggregated model, is not yet formed at present
Perfect theory.
Summary of the invention
Small echo is driven based on signal atom technical problem to be solved by the invention is to provide a kind of raising accuracy of identification
Regenerate the mechanical Multiple faults diagnosis approach of nuclear machine learning.
The technical proposal adopted by the invention to solve the above technical problems is that: small echo reproducing kernel machine is driven based on signal atom
The mechanical Multiple faults diagnosis approach of device study, comprising the following steps:
Step S10, the signal to be analyzed of collection machinery equipment, then fault setting is carried out to mechanical equipment, and acquire failure
The unsteady mechanical vibration signal of equipment；
Step S20, it is shaken using the Gabor atom orthogonal matching pursuit method based on ant colony search strategy to non-stationary machinery
Dynamic signal is tracked and is decomposed, and multiple atom components and residual component with signal best match to be analyzed are obtained；
Step S30, machine learning --- small echo reproducing kernel branch is inputted using multiple atom components of acquisition as feature samples
It holds and study is trained and tested in vector machine classifier, to identify mechanical multiple faults mode type.
Further embodiment is the detailed process of the step S20 are as follows:
Step S201, residue signal R is initialized0F, the value range of specified atom dictionary parameter, and initialization with Ant colony is excellent
Change algorithm, wherein initialization residue signal R0F is equal to unsteady mechanical vibration signal, current residue signal RnF is equal to letter to be analyzed
Number；
Step S202, ant colony optimization algorithm search and current residue signal R are utilizednThe Gabor atom of f best matchIts Gabor atomExpression formula are as follows:
In formula: u is shift factor, and m is scale factor, and ν is frequency factor, and ω is phase factor, and t is the time, and n is orthogonal
The current decomposition number of match tracing method；
Step S203, to Gabor atomIt is orthogonalized processing, obtains atom unWith current residue signal RnF is in original
Sub- unOn projection, wherein atom unWith residue signal Rn+1The formula of f are as follows:
In formula: n is the current decomposition number of orthogonal matching pursuit method, unFor atom, Rn+1F is residue signal.
Further embodiment is that the Gabor of ant colony optimization algorithm search best match is utilized in the step S202
AtomProcedural representation are as follows:
In formula: fitness indicates fitness function or objective function, u are shift factor, and m is scale factor, and ν is frequency
The factor, ω are phase factor, and t is the time, and n is the current decomposition number of orthogonal matching pursuit method；γ is a Gabor atom
Parameter group；RnF is current residue signal.
Further embodiment is that the signal to be analyzed and unsteady mechanical vibration signal pass through acceleration sensing
Device acquisition.
Further embodiment is the detailed process of the step S30 are as follows:
Step S301, according to the tensor product theorem of reproducing kernel, wavelet structure Reproducing Kernel Function K (x, y)；Its expression formula are as follows:
In formula: cjFor small echo ψjCoefficient, j, k are that morther wavelet ψ () is flexible and translation parameters respectively；
Step S302, machine learning is established --- small echo reproducing kernel support vector machine classifier decision model f (x), table
Show formula are as follows:
In formula: b*∈ R, α*For Lagrange multiplier；
Step S303, small echo reproducing kernel support vector cassification is inputted using obtained multiple atom components as feature samples
Study is trained and tested in device decision model f (x), identifies mechanical multiple faults mode type.
Beneficial effects of the present invention: the feature of present invention extraction stage incorporates ant colony optimization algorithm based on Gabor atom just
In match tracing (OMP) method of friendship, the atom with signal best match to be analyzed can be adaptively chosen from atom dictionary, directly
The signal atom that connecing will acquire is trained and tests as feature input machine learning, reduces and needs to calculate time domain, frequency domain
With time and frequency domain characteristics and carry out data for untreated link, substantially increase computational efficiency.Machine is directed in the pattern-recognition stage
Device study --- tradition commonly uses the incompleteness of kernel function and carries out study from sample and asks there are ill posed in support vector machines
Topic has studied the practical approach that a kind of orthogonal wavelet Symmlet small echo with compact schemes is used to construct Reproducing Kernel Function, and will
The kernel function introduces support vector machines and constructs a kind of referred to as small echo reproducing kernel support vector machine classifier.The classifier is applied to
Mechanical Fault Pattern Recognition, the results showed that small echo Reproducing Kernel Function has better than traditional Gauss radial direction base (RBF) kernel function
The application that identification precision is support vector machines in engineering provides new theoretical basis.
Detailed description of the invention
Fig. 1 is the flow chart of embodiment；
Fig. 2 is machine learning --- support vector machines Kernel waveform in embodiment；
Fig. 3 is the time domain waveform of crack fault original vibration signal in embodiment；
Fig. 4 is the time domain waveform of pitting fault original vibration signal in embodiment；
Fig. 5 is the time domain waveform of broken teeth failure original vibration signal in embodiment；
Fig. 6 is the time domain waveform of hypodontia failure original vibration signal in embodiment；
Fig. 7 is the atom waveform of crack fault sample time-frequency atom in embodiment；
Fig. 8 is the atom waveform of spot corrosion fault sample time-frequency atom in embodiment；
Fig. 9 is the atom waveform of broken teeth fault sample time-frequency atom in embodiment；
Figure 10 is the atom waveform of hypodontia fault sample time-frequency atom in embodiment；
Figure 11 is mechanical multiclass Fault Pattern Recognition result of the embodiment based on traditional RBF kernel function support vector machine；
Figure 12 is mechanical multiclass Fault Pattern Recognition result of the embodiment based on small echo Reproducing Kernel Function support vector machines.
Specific embodiment
Invention is further described in detail with reference to the accompanying drawings and examples.
As shown in Figure 1, the mechanical multi-fault Diagnosis of the invention based on signal atom driving small echo regeneration nuclear machine learning
Method, comprising the following steps:
(1) pass through the signal to be analyzed of acceleration transducer collection machinery equipment, then fault setting carried out to mechanical equipment,
And acquire the unsteady mechanical vibration signal of faulty equipment；
(2) residue signal R is initialized0F, the value range of specified atom dictionary parameter, and initialization with Ant colony optimization algorithm,
Wherein initialize residue signal R0F is equal to unsteady mechanical vibration signal, current residue signal RnF is equal to signal to be analyzed；
(3) ant colony optimization algorithm search and current residue signal R are utilizednThe Gabor atom of f best matchGabor is former
SonFor with current residue signal RnThe maximum atom of f inner product, Gabor atomExpression formula are as follows:
In formula: u is shift factor, and m is scale factor, and ν is frequency factor, and ω is phase factor, and t is the time, and n is orthogonal
The current decomposition number of match tracing method；
Wherein utilize the Gabor atom of ant colony optimization algorithm search best matchProcedural representation are as follows:
In formula: fitness indicates fitness function or objective function, and u is shift factor, u ∈ [0, N-1], m be scale because
Son, m ∈ [1, N] (N be signal sampling points), ν are frequency factor, value interval usually cover mechanical resonance frequency i.e. ν ∈ [0,
6000], ω is phase factor, and θ ∈ [0,2 π], t are the time, and n is the current decomposition number of orthogonal matching pursuit method；RnF is
Current residue signal；γ is a Gabor atomic parameter group；
(4) using Schmidt process to Gabor atomIt is orthogonalized processing, obtains atom unWith it is current residual
Remaining signal RnF is in atom unOn projection, wherein atom unWith residue signal Rn+1The formula of f are as follows:
In formula: n is the current decomposition number of orthogonal matching pursuit method, unFor atom, Rn+1F is residue signal；
(5) unsteady mechanical vibration is believed using the Gabor atom orthogonal matching pursuit method based on ant colony search strategy
It number is tracked and is decomposed, obtain the multiple atom components and residual component with signal best match to be analyzed；
(6) according to the tensor product theorem of reproducing kernel, wavelet structure Reproducing Kernel Function K (x, y)；Its expression formula are as follows:
In formula: cjFor small echo ψjCoefficient, j, k are that morther wavelet ψ () is flexible and translation parameters respectively；The present invention is using tool
There is the orthogonal wavelet Symmlet small echo of compact schemes for constructing Reproducing Kernel Function (Fig. 2)；
(6) machine learning is established --- small echo reproducing kernel support vector machine classifier decision model f (x), expression are as follows:
In formula: b*∈ R, α*For Lagrange multiplier；
(7) small echo reproducing kernel support vector machine classifier decision is inputted using obtained multiple atom components as feature samples
Study is trained and tested in model f (x), identifies mechanical multiple faults mode type.
The present invention is using Gabor atom orthogonal matching pursuit (OMP) method based on ant colony search strategy to non-stationary machine
Tool vibration signal is tracked and is decomposed, and multiple atom components and residual component with signal best match to be analyzed are obtained；It will
Obtain multiple atom components as feature samples input machine learning --- in small echo reproducing kernel support vector machine classifier into
Row training and test study, enable the classifier automatically and efficiently to realize the accurate identification of mechanical multiple faults mode.
Embodiment 1
The design parameter for testing epicyclic gearbox is as shown in table 1:
1 epicyclic gearbox design parameter of table
When test, motor drives test bearing rotation, and it is 30Hz, signal sampling frequencies 10kHz that wherein motor, which turns frequency,.
Failure gear is that the fault type of sun gear includes: tooth root crack fault, tooth surface abrasion failure, broken teeth failure, hypodontia failure.
Step 1: adding label 1,2,3,4 to 4 kinds of sun gear fault types respectively；Using 4096 points as data length, to every
Kind state has intercepted 50 groups of data (Fig. 3-Fig. 6 is the corresponding time domain waveform of four kinds of failures).Therefore, for being total under 4 kinds of states
200 groups of data carry out Fault Pattern Recognitions, extract wherein 60% sample for training RBF kernel function and multiple dimensioned reproducing kernel letter
Several SVM diagnostic models, remaining 40% sample is for testing.
Step 2: using Gabor atom orthogonal matching pursuit (OMP) method based on ant colony search strategy to state in 4
Unsteady mechanical vibration signal tracked and decomposed, obtain with multiple atom components of signal best match to be analyzed (such as
Fig. 7-10) and residual component.Then, 4 kinds of states are obtained and is constituted with multiple atom component marks of signal best match to be analyzed
Feature set, and it is divided into training sample and test sample.
Step 3: feature samples collection obtained in second step is inputted machine learning --- small echo regenerates kernel support vectors
Study is trained and tested in machine classifier, and the type of mechanical multiple faults mode is recognized.
Point of the support vector machines of traditional RBF kernel function and small echo Reproducing Kernel Function is set forth in wherein Figure 11, Figure 12
Class result.Compare two kinds of kernel functions, is used using orthogonal wavelet Symmlet regeneration Wavelet Kernel Function support vector machine classifier ratio
The accuracy of RBF kernel function support vector machine classifier improves 8.75 percentage points.
The above is not intended to limit the present invention in any form, although the present invention takes off through the foregoing embodiment
Show, however, it is not intended to limit the invention, any person skilled in the art, is not departing from technical solution of the present invention range
It is interior, made when the technology contents using the disclosure above and change or be modified to the equivalent embodiments of equivalent variations a bit, but it is all not
Be detached from technical solution of the present invention content, according to the technical essence of the invention it is to the above embodiments it is any it is simple modification,
Equivalent variations and modification, all of which are still within the scope of the technical scheme of the invention.
Claims (5)
1. the mechanical Multiple faults diagnosis approach based on signal atom driving small echo regeneration nuclear machine learning, which is characterized in that including
Following steps:
Step S10, the signal to be analyzed of collection machinery equipment, then fault setting is carried out to mechanical equipment, and acquire faulty equipment
Unsteady mechanical vibration signal；
Step S20, unsteady mechanical vibration is believed using the Gabor atom orthogonal matching pursuit method based on ant colony search strategy
It number is tracked and is decomposed, obtain the multiple atom components and residual component with signal best match to be analyzed；
Step S30, using multiple atom components of acquisition as feature samples input machine learning --- small echo reproducing kernel support to
It is trained and tests study in amount machine classifier, to identify mechanical multiple faults mode type.
2. a kind of mechanical multiple faults based on signal atom driving small echo regeneration nuclear machine learning according to claim 1 is examined
Disconnected method, which is characterized in that the detailed process of the step S20 are as follows:
Step S201, residue signal R is initialized0F, the value range of specified atom dictionary parameter, and initialization with Ant colony optimization is calculated
Method, wherein initialization residue signal R0F is equal to unsteady mechanical vibration signal, current residue signal RnF is equal to signal to be analyzed；
Step S202, ant colony optimization algorithm search and current residue signal R are utilizednThe Gabor atom of f best matchIts
Gabor atomExpression formula are as follows:
In formula: u is shift factor, and m is scale factor, and ν is frequency factor, and ω is phase factor, and t is the time, and n is orthogonal matching
The current decomposition number of method for tracing；
Step S203, to Gabor atomIt is orthogonalized processing, obtains atom unWith current residue signal RnF is in atom unOn
Projection, wherein atom unWith residue signal Rn+1The formula of f are as follows:
In formula: n is the current decomposition number of orthogonal matching pursuit method, unFor atom, Rn+1F is residue signal.
3. a kind of mechanical multiple faults based on signal atom driving small echo regeneration nuclear machine learning according to claim 2 is examined
Disconnected method, which is characterized in that the Gabor atom of ant colony optimization algorithm search best match is utilized in the step S202Mistake
Journey indicates are as follows:
In formula: fitness indicates fitness function or objective function, u are shift factor, and m is scale factor, and ν is frequency factor,
ω is phase factor, and t is the time, and n is the current decomposition number of orthogonal matching pursuit method；γ is a Gabor atomic parameter
Group；RnF is current residue signal.
4. a kind of mechanical mostly event based on signal atom driving small echo regeneration nuclear machine learning according to claim 1 or 3
Hinder diagnostic method, which is characterized in that the signal to be analyzed and unsteady mechanical vibration signal are adopted by acceleration transducer
Collection.
5. a kind of mechanical mostly event based on signal atom driving small echo regeneration nuclear machine learning according to claim 2 or 3
Hinder diagnostic method, which is characterized in that the detailed process of the step S30 are as follows:
Step S301, according to the tensor product theorem of reproducing kernel, wavelet structure Reproducing Kernel Function K (x, y)；Its expression formula are as follows:
In formula: cjFor small echo ψjCoefficient, j, k are that morther wavelet ψ () is flexible and translation parameters respectively；
Step S302, machine learning is established --- small echo reproducing kernel support vector machine classifier decision model f (x), expression
Are as follows:
In formula: b*∈ R, α*For Lagrange multiplier；
Step S303, it determines obtained multiple atom components as feature samples input small echo reproducing kernel support vector machine classifier
Study is trained and tested in plan model f (x), identifies mechanical multiple faults mode type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810617661.7A CN109029940A (en) | 2018-06-15 | 2018-06-15 | Mechanical Multiple faults diagnosis approach based on signal atom driving small echo regeneration nuclear machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810617661.7A CN109029940A (en) | 2018-06-15 | 2018-06-15 | Mechanical Multiple faults diagnosis approach based on signal atom driving small echo regeneration nuclear machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109029940A true CN109029940A (en) | 2018-12-18 |
Family
ID=64609674
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810617661.7A Pending CN109029940A (en) | 2018-06-15 | 2018-06-15 | Mechanical Multiple faults diagnosis approach based on signal atom driving small echo regeneration nuclear machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109029940A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110907162A (en) * | 2019-12-13 | 2020-03-24 | 北京天泽智云科技有限公司 | Rotating machinery fault feature extraction method without tachometer under variable rotating speed |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101620669A (en) * | 2008-07-01 | 2010-01-06 | 邹采荣 | Method for synchronously recognizing identities and expressions of human faces |
| CN105976021A (en) * | 2016-05-24 | 2016-09-28 | 北京工业大学 | Fault diagnosis method for roller assembly of belt conveyor |
| CN106017879A (en) * | 2016-05-18 | 2016-10-12 | 河北工业大学 | Universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals |
| WO2016183661A1 (en) * | 2015-05-15 | 2016-11-24 | Motion Metrics International Corp | Method and apparatus for locating a wear part in an image of an operating implement |
| CN107122802A (en) * | 2017-05-02 | 2017-09-01 | 哈尔滨理工大学 | A kind of fault detection method based on the rolling bearing for improving wavelet neural network |
-
2018
- 2018-06-15 CN CN201810617661.7A patent/CN109029940A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101620669A (en) * | 2008-07-01 | 2010-01-06 | 邹采荣 | Method for synchronously recognizing identities and expressions of human faces |
| WO2016183661A1 (en) * | 2015-05-15 | 2016-11-24 | Motion Metrics International Corp | Method and apparatus for locating a wear part in an image of an operating implement |
| CN106017879A (en) * | 2016-05-18 | 2016-10-12 | 河北工业大学 | Universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals |
| CN105976021A (en) * | 2016-05-24 | 2016-09-28 | 北京工业大学 | Fault diagnosis method for roller assembly of belt conveyor |
| CN107122802A (en) * | 2017-05-02 | 2017-09-01 | 哈尔滨理工大学 | A kind of fault detection method based on the rolling bearing for improving wavelet neural network |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110907162A (en) * | 2019-12-13 | 2020-03-24 | 北京天泽智云科技有限公司 | Rotating machinery fault feature extraction method without tachometer under variable rotating speed |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Grezmak et al. | Interpretable convolutional neural network through layer-wise relevance propagation for machine fault diagnosis | |
| Huo et al. | Incipient fault diagnosis of roller bearing using optimized wavelet transform based multi-speed vibration signatures | |
| Feng et al. | Atomic decomposition and sparse representation for complex signal analysis in machinery fault diagnosis: A review with examples | |
| Kuang et al. | Class-imbalance adversarial transfer learning network for cross-domain fault diagnosis with imbalanced data | |
| Lei et al. | EEMD method and WNN for fault diagnosis of locomotive roller bearings | |
| Lin et al. | Detection and classification of multiple power-quality disturbances with wavelet multiclass SVM | |
| CN111238807B (en) | Fault diagnosis method for planetary gear box | |
| Guo et al. | Rolling bearing fault classification based on envelope spectrum and support vector machine | |
| CN110543860B (en) | Mechanical fault diagnosis method and system based on TJM (machine learning model) transfer learning | |
| Li et al. | Hidden Markov model-based fault diagnostics method in speed-up and speed-down process for rotating machinery | |
| CN112036301B (en) | Driving motor fault diagnosis model construction method based on intra-class feature transfer learning and multi-source information fusion | |
| Kaya et al. | Classification of bearing vibration speeds under 1D-LBP based on eight local directional filters | |
| Wu et al. | Rub-impact fault diagnosis of rotating machinery based on 1-D convolutional neural networks | |
| CN106295684A (en) | A kind of the most continuous based on micro-Doppler feature/discontinuous gesture recognition methods | |
| CN103048133B (en) | Bayesian network-based rolling bearing fault diagnosis method | |
| Pinheiro et al. | Vibration analysis in turbomachines using machine learning techniques | |
| CN111089720A (en) | Regularization sparse filtering method suitable for gear fault diagnosis under variable rotating speed | |
| Wang et al. | Gear fault intelligent diagnosis based on frequency-domain feature extraction | |
| CN112329329B (en) | Simulation data driven rotary machine depth semi-supervised migration diagnosis method | |
| CN112414713A (en) | Rolling bearing fault detection method based on measured signals | |
| Islam et al. | Time–frequency envelope analysis-based sub-band selection and probabilistic support vector machines for multi-fault diagnosis of low-speed bearings | |
| Islam et al. | Motor bearing fault diagnosis using deep convolutional neural networks with 2d analysis of vibration signal | |
| CN114091504A (en) | Rotary machine small sample fault diagnosis method based on generation countermeasure network | |
| CN111796342A (en) | Tiny fault diagnosis method and device for cam type absolute gravimeter | |
| Liu et al. | Intelligent cross-condition fault recognition of rolling bearings based on normalized resampled characteristic power and self-organizing map |
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: 20181218 |