CN107036808A - Gearbox of wind turbine combined failure diagnostic method based on SVMs probability Estimation - Google Patents
Gearbox of wind turbine combined failure diagnostic method based on SVMs probability Estimation Download PDFInfo
- Publication number
- CN107036808A CN107036808A CN201710232999.6A CN201710232999A CN107036808A CN 107036808 A CN107036808 A CN 107036808A CN 201710232999 A CN201710232999 A CN 201710232999A CN 107036808 A CN107036808 A CN 107036808A
- Authority
- CN
- China
- Prior art keywords
- sample
- sensor
- svms
- probability
- gearbox
- 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.)
- Granted
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
Landscapes
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a kind of gearbox of wind turbine combined failure diagnostic method based on SVMs probability Estimation.The invention mainly comprises following two modules:First module is vibration signal processing and fault signature extraction algorithm based on overall experience mode decomposition (EEMD);Second module is the training pattern method and probability Estimation algorithm of SVMs.The present invention is respectively established for different sensors, is integrated the result of all graders during analysis, improves the degree of accuracy of diagnosis.The present invention gives detailed arthmetic statement to emulate signal as test object, and validity of the algorithm in terms of gear-box combined failure diagnosis by experimental verification.
Description
Technical field
Extracted the invention belongs to fault signature and method for diagnosing faults field, more particularly to it is a kind of general based on SVMs
The fault signature for wind-driven generator group wheel box of rate estimation is extracted and method for diagnosing faults.
Background technology
In the past few years, with the fast development of Wind Power Generation Industry, substantial amounts of Wind turbines are constantly built by deployment, wind
Electric single-machine capacity is also being continuously increased, and the accident because of caused by wind power generating set failure happens occasionally, and causes huge warp
Ji loss, hinders the development speed of Wind Power Generation Industry.Gear-box is located in engine rooms of wind power generators, is that wind-driven generator transmission is dynamic
The critical piece of power and the important pivot of connection main shaft and generator, its operation can be disturbed by many factors, such as wind speed ripple
Dynamic and load dynamic change.Simultaneously as the environment residing for Wind turbines is typically relatively more severe, fluctuations in wind speed and load change
Acutely and frequently, therefore gearbox parts easily Aging Damage during long-play, all kinds of failures are produced.Therefore, grind
Study carefully a set of feasible Fault Diagnosis of Gear Case method significant for the stable operation of Wind turbines.
Several vibrating sensors can be typically installed outside the running status of gear case of blower, gear-box in order to monitor, passed
The vibration signal measured is sent to host computer by sensor, and vibration signal is analyzed and speculated now using certain diagnostic method
The running status of gear-box.With the continuous accumulation of gear case of blower vibration data, several machine learning algorithms are also introduced into
To Fault Diagnosis of Gear Case field.SVMs is a kind of machine learning algorithm proposed on the basis of Statistical Learning Theory,
Suitable for the troubleshooting issue under Small Sample Size, it is widely used in mechanical breakdown field.The essence of algorithm of support vector machine
Disaggregated model exactly is set up on the basis of existing gear-box vibration data, when new sample data comes interim, is entered data into
Disaggregated model, so as to judge the running status of machinery.However, the SVMs that scholars propose at present still has some offices
It is sex-limited:First, the output result of existing algorithm of support vector machine is only category label, when SVMs is judged by accident
When, the output result is just entirely ineffective, it is impossible to provide more information.Secondly, what current algorithm of support vector machine was solved is most
It is single failure problem, i.e., the failure that sample occurs is preferably at most one, less to combined failure diagnosis research.Finally, really
Gearbox fault signal has non-stationary, it is non-linear the features such as, existing method shows poor when extracting the feature of such signal,
The accuracy rate of diagnosis for influenceing SVMs final.
The present invention for existing method weak point there is provided complete set based on SVMs probability Estimation
Method for diagnosing faults, algorithm mainly includes following two modules:First module is to be based on overall experience mode decomposition (EEMD)
Vibration signal processing and fault signature extraction algorithm;Second module is that the training pattern method and probability of SVMs are estimated
Calculating method.The novelty of the diagnostic method is mainly reflected in the following aspects:Processing is good at using overall experience mode decomposition
The characteristics of non-linear, non-stationary signal, the vibration signal that there is combined failure is decoupled, separate the feature of different faults.
Meanwhile, the energy ratio of faults energy and the approximate entropy of faults confusion degree, optimization event are chosen when extracting fault signature
Hinder Feature Selection.In addition, in Training Support Vector Machines model, this method is respectively established for different sensors, point
The result of these graders is integrated during analysis, the degree of accuracy of diagnosis is improved.
The content of the invention
In order to improve existing method, it can solve the problem that the diagnosis of gear-box combined failure is asked it is an object of the invention to provide one kind
The method of topic, and improve the degree of accuracy of diagnosis.
The purpose of the present invention is achieved through the following technical solutions, and is comprised the following steps:
Step (1), the vibration signal extracted from gear-box monitoring data storehouse under different working condition is used as sample.Assuming that
M sensor is installed on gear-box, according to the corresponding sample of each sensor of the sequential processes of sensor, m=1 is initialized,
To m-th of sensor, the fault signature of its respective sample is sought in the steps below.
Step (2) carries out EEMD decomposition to the sample point x (t) in sample, and EEMD decomposable processes are as follows:
(2-1) initializes iterations N=1, and predetermined maximum iteration is K;
(2-2) adds the white Gaussian noise w (t) of computer random generation as new x (t) in x (t), calculates x (t)
Upper lower envelope average value m1;
(2-3) calculates x (t) and m1Difference h10, judge h10Whether it is intrinsic mode function, if it is, making eigen mode
Measure c1=h10, and be transferred to (2-5), if it is not, then being transferred to (2-4);
(2-4) calculates k h repeatedly1(k-1)-m1k=h1k, until h1kIntrinsic mode function condition is met, c is made1=h1k;Its
In, m1kFor h1(k-1)Upper lower envelope average value;
(2-5) separation c from x (t)1, make r1=x (t)-c1, by r1Regard new x (t) as, constantly repeat (2-2)-(2-
4), often it is repeated once, it is possible to obtain new intrinsic modulus, i.e. r1-c2=r2,...,rn-1-cn=rn, standard is stopped until meeting
Then;
(2-6) record epicycle decomposes obtained intrinsic modulus and residual volume, it is assumed that isolates q intrinsic modulus, then makes c1N
=c1,c2N=c2,...,cqN=cq,rqN=rq, and iterations N is added 1, if N is less than predetermined maximum iteration K, turn
Enter (2-2), otherwise calculate final intrinsic modulus and residual volume, calculation formula isWith
Step (3) calculates each intrinsic modulus and the energy ratio and approximate entropy of residual volume, constitutes m-th of sensor correspondence sample
This characteristic vector.
Step (4) is entered using the characteristic vector construction optimal decision function of m-th of sensor respective sample to training sample
Row is classified, and categorised decision function is:
F (x)=sgn (wTx+b)
Wherein w and b are respectively the weight coefficient vector sum deviation of the categorised decision function, and x is sampling feature vectors, this
Two parameters can determine that object function is by solving object function:
s.t.yi(wTx+b)-1+ξi≥0
Wherein Q represents the number of samples in m-th of sensor, ξiFor slack variable, C is penalty coefficient, yiFor sample mark
Label, the grader now constructed is designated as SVMm.
Probability between the class of step (5) calculating sampleIts
Middle i, j represent i-th kind of running status, jth kind running status, and f is that sample point inputs the value that decision function is obtained, and A and B are to pass through
Following object function is minimized to obtain
Sample point x (t) belongs to the Probability p of i-th kind of running status under the model for step (6) estimationi=P (y=i | x),
Computational methods are the following object function of optimization
Step (7), to next sensor repeat step (2)-step (6), even m adds one, set up it is next support to
Amount machine grader;Until all the sensors analysis is finished, represent that fault diagnosis model is set up and complete;
Test sample input model is carried out fault diagnosis by step (8).Specifically, when test sample data acquisition is from multiple
Different sensors, then need to separate the vibration data for belonging to different sensors in sample, and sequentially input what is had built up
The output probability of all models, is finally averaging, obtains final result by the supporting vector machine model of respective sensor.
It is of the invention compared with existing method for diagnosing faults, beneficial effect includes:EEMD algorithms are introduced gear-box by the present invention
The fault signature of vibration signal is extracted, the characteristics of agreeing with gear-box vibration signal time-varying, non-linear, non-stationary, and can be effective
Solve the phenomenon that combined failure is mutually coupled;The sample fault signature that the present invention is extracted is energy ratio and approximate entropy, is demonstrate,proved through experiment
It is real that there is higher diagnosis;The present invention exports the probability that sample belongs to different conditions using SVMs, can solve the problem that
Traditional support vector machine model can not provide the shortcoming of fault message when judging by accident;The inventive method is for different sensors point
Corresponding diagnostic model is not set up, again the result of each diagnostic model is integrated during diagnosis, is effectively improved the spirit of algorithm
Activity and stability.
Brief description of the drawings
Fig. 1 algorithm frame schematic diagrames;
Fig. 2 simulation samples point time domain beamformer;
Intrinsic modulus and residual time domain waveform of the vibration signal of Fig. 3 simulation samples point sensor 1 after EEMD algorithm process
Figure.
Embodiment
The specific implementation method to the present invention is described further below in conjunction with the accompanying drawings:
Fig. 1 is total algorithm block schematic illustration of the invention, as shown in the figure, it is assumed that sample signal is gathered from gear-box
M sensor, then sample x x can be decomposed into according to sensor1,x2,...,xM, vibration performance f is extracted respectively1,f2,...,fM, and
Set up different supporting vector machine models.When new sample input, the diagnostic result of comprehensive all supporting vector machine models is obtained
Go out last conclusion.
In order to verify effectiveness of the invention, construct some groups of emulation signals by Matlab to simulate gear-box vibration
Signal.Emulation experiment setting is as follows:Number of sensors M=2, can use sensor 1 and sensor 2, while also illustrating that needs instruction
Practice two supporting vector machine models.Fig. 2 is a simulation sample point time domain beamformer randomly selected, it can be seen that it has two-way
Signal, represents the signal that two sensors are measured, x respectively1The signal surveyed for sensor 1, x2The signal surveyed for sensor 2.Failure
Type set is two kinds, is represented with failure 1 and failure 2.Simulation sample can be divided into four classes according to respective state:Normal sample,
The sample of failure 1, the sample of failure 2 and combined failure (failure 1 and failure 2 exist simultaneously) sample.Training sample be normal sample,
Three kinds of the sample of failure 1 and 2 sample of failure, combined failure sample is used as test sample.Understood through discussed above, it is proposed by the present invention
Supporting vector machine model is generated by normal sample and single failure sample training, but can be for diagnosis combined failure sample.
What is performed comprises the following steps that:
Step (1), generates some normal samples, the sample of failure 1 and the sample of failure 2 by the use of Matlab and is used as training sample.
As it is assumed that being provided with 2 sensors on gear-box, then all there is two-way sensor signal in all sample points, by the suitable of sensor
Each corresponding sample of sensor of sequence processing, initializes m=1.
Step (2) carries out EEMD decomposition, Tu3Wei Mou roads sensing to the sample point x (t) in m-th of sensor correspondence sample
Device signal decomposes obtained intrinsic modulus and surplus time domain beamformer through EEMD.The specific decomposable processes of EEMD are as follows:
(2-1) initializes iterations N=1, and predetermined maximum iteration is K;
(2-2) adds the white Gaussian noise w (t) of computer random generation as new x (t) in x (t), calculates x (t)
Upper lower envelope average value m1;
(2-3) calculates x (t) and m1Difference h10, judge h10Whether it is intrinsic mode function, if it is, making eigen mode
Measure c1=h10, and be transferred to (2-5), if it is not, then being transferred to (2-4);
(2-4) calculates k h repeatedly1(k-1)-m1k=h1k, until h1kIntrinsic mode function condition is met, c is made1=h1k.Its
In, m1kFor h1(k-1)Upper lower envelope average value;
(2-5) separation c from x (t)1, make r1=x (t)-c1, by r1Regard new x (t) as, constantly repeat (2-2)-(2-
4), often it is repeated once, it is possible to obtain new intrinsic modulus, i.e. r1-c2=r2,...,rn-1-cn=rn, standard is stopped until meeting
Then;
(2-6) record epicycle decomposes obtained intrinsic modulus and residual volume, it is assumed that isolates q intrinsic modulus, then makes c1N
=c1,c2N=c2,...,cqN=cq,rqN=rq, and iterations N is added 1, if N is less than predetermined maximum iteration K, turn
Enter (2-2), otherwise calculate final intrinsic modulus and residual volume, calculation formula isWith
Step (3) calculates each intrinsic modulus and the energy ratio and approximate entropy of residual volume, constitutes m-th of sensor correspondence sample
This characteristic vector.
Step (4) is classified using the characteristic vector construction optimal decision function of m sample classes to training sample, is classified
Decision function is:
F (x)=sgn (wTx+b)
Wherein w and b are respectively the weight coefficient vector sum deviation of the categorised decision function, and x is sampling feature vectors, this
Two parameters can determine that object function is by solving object function:
s.t.yi(wTx+b)-1+ξi≥0
Wherein Q represents the number of samples in m-th of sensor, ξiFor slack variable, C is penalty coefficient, yiFor sample mark
Label, the grader now constructed is designated as SVMm..
Probability between the class of step (5) calculating sampleIts
Middle i, j represent i-th kind of running status, jth kind running status, and f is that sample point inputs the value that decision function is obtained, and A and B are to pass through
Following object function is minimized to obtain
Sample point x (t) belongs to the Probability p of i-th kind of running status under the model for step (6) estimationi=P (y=i | x),
Computational methods are the following object function of optimization
Step (7), parameter m values Jia 1, to next sensor repeat step (2)-step (6), set up it is next support to
Amount machine grader;Until all the sensors analysis is finished, represent that fault diagnosis model is set up and finish.
Combined failure sample input model is carried out fault diagnosis by step (8).Specifically, will be gathered in combined failure sample
SVM1 is inputted from the vibration signal of sensor 1, the vibration signal gathered in combined failure sample from sensor 2 is inputted into SVM2,
Probability Estimation is respectively obtained, finally the probability of gained is averaging, final sumbission is obtained.
Table 1 is the diagnostic result of 5 combined failure samples.From table it can be seen that, either SVM1 or SVM2, all without
Method individually detects combined failure sample while having two kinds of failures, and after their probability Estimation is integrated, it finds that therefore
The probability that barrier 1 and failure 2 occur is closer to, thus can draw sample while having the conclusion of two kinds of failures, demonstrates this
The validity of invention algorithm.
Some combined failure sample diagnostic results of Table I
It should be appreciated that above-described embodiment is preferred embodiments of the present invention, but embodiments of the present invention are not
It is restricted to the described embodiments.For those of ordinary skills, it can according to the above description be improved or be converted, and
All these modifications and variations should all belong to the protection domain of appended claims of the present invention.
Claims (3)
1. a kind of gearbox of wind turbine combined failure diagnostic method based on SVMs probability Estimation, it is characterised in that:
Corresponding supporting vector machine model is set up for each sensor, the output result of model belongs to for certain sample point under the model
In corresponding probability under different running statuses, to each running status, the output result of all models is averaging, is somebody's turn to do
Sample point belongs to the true probability of the running status.
2. the gearbox of wind turbine combined failure diagnosis side according to claim 1 based on SVMs probability Estimation
Method, it is characterised in that:
Step (1), the vibration signal extracted from gear-box monitoring data storehouse under different working condition is used as sample;Assuming that in tooth
M sensor is installed, according to the corresponding sample of each sensor of the sequential processes of sensor on roller box.M=1 is initialized, to m
Individual sensor, seeks the fault signature of its respective sample in the steps below.
Step (2) carries out EEMD decomposition to the sample point x (t) in m-th of sensor correspondence sample, obtains final intrinsic modulus
And residual volume;
Step (3) calculates each intrinsic modulus and the energy ratio and approximate entropy of residual volume, constitutes the corresponding sample of m-th of sensor
Characteristic vector;
Step (4) is divided training sample using the characteristic vector construction optimal decision function of m-th of sensor respective sample
Class, categorised decision function is:
F (x)=sgn (wTx+b)
Wherein w and b are respectively the weight coefficient vector sum deviation of the categorised decision function, and x is sampling feature vectors, the two
Parameter can determine that object function is by solving object function:
s.t. yi(wTx+b)-1+ξi≥0
Wherein Q represents the number of samples in m-th of sensor, ξiFor slack variable, C is penalty coefficient, yiFor sample label, this
When the grader that constructs be designated as SVMm;
Probability between the class of step (5) calculating sampleWherein i, j
I-th kind of running status, jth kind running status are represented, f is that sample point inputs the value that decision function is obtained, and A and B are by minimum
Change following object function to obtain
Sample point x (t) belongs to the Probability p of i-th kind of running status under the model for step (6) estimationi=P (y=i | x), calculating side
Method is the following object function of optimization
Step (7), to next sensor repeat step (2)-step (6), even m adds one, sets up next SVMs
Grader;Until all the sensors analysis is finished, represent that fault diagnosis model is set up and complete;
Test sample input model is carried out fault diagnosis by step (8).Specifically, working as test sample data acquisition from multiple differences
Sensor, then need to separate the vibration data for belonging to different sensors in sample, and sequentially input the correspondence having had built up
The output probability of all models, is finally averaging, obtains final result by the supporting vector machine model of sensor.
3. the gearbox of wind turbine combined failure diagnosis side according to claim 2 based on SVMs probability Estimation
Method, it is characterised in that:Step 2) in EEMD decomposable processes it is as follows:
(2-1) initializes iterations N=1, and predetermined maximum iteration is K;
(2-2) adds the white Gaussian noise w (t) of computer random generation as new x (t) in x (t), calculates the upper of x (t)
Lower envelope average value m1;
(2-3) calculates x (t) and m1Difference h10, judge h10Whether it is intrinsic mode function, if it is, making intrinsic modulus c1
=h10, and be transferred to (2-5), if it is not, then being transferred to (2-4);
(2-4) calculates k h repeatedly1(k-1)-m1k=h1k, until h1kIntrinsic mode function condition is met, c is made1=h1k.Wherein, m1k
For h1(k-1)Upper lower envelope average value;
(2-5) separation c from x (t)1, make r1=x (t)-c1, by r1Regard new x (t) as, constantly repeat (2-2)-(2-4), often weigh
Again once, it is possible to obtain new intrinsic modulus, i.e. r1-c2=r2,...,rn-1-cn=rn, until meeting stopping criterion;
(2-6) record epicycle decomposes obtained intrinsic modulus and residual volume, it is assumed that isolates q intrinsic modulus, then makes c1N=c1,
c2N=c2,...,cqN=cq,rqN=rq, and iterations N is added 1, if N is less than predetermined maximum iteration K, it is transferred to (2-
2) final intrinsic modulus and residual volume, are otherwise calculated, calculation formula isWith
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710232999.6A CN107036808B (en) | 2017-04-11 | 2017-04-11 | Gearbox of wind turbine combined failure diagnostic method based on support vector machines probability Estimation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710232999.6A CN107036808B (en) | 2017-04-11 | 2017-04-11 | Gearbox of wind turbine combined failure diagnostic method based on support vector machines probability Estimation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107036808A true CN107036808A (en) | 2017-08-11 |
CN107036808B CN107036808B (en) | 2019-04-19 |
Family
ID=59535911
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710232999.6A Active CN107036808B (en) | 2017-04-11 | 2017-04-11 | Gearbox of wind turbine combined failure diagnostic method based on support vector machines probability Estimation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107036808B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107830996A (en) * | 2017-10-10 | 2018-03-23 | 南京航空航天大学 | A kind of vehicle rudder diagnosis method for system fault |
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 |
CN108872781A (en) * | 2018-05-08 | 2018-11-23 | 广东昊阳电力建设有限公司 | Analysis method and device based on electric power facility intelligent patrol detection |
CN112696481A (en) * | 2020-12-11 | 2021-04-23 | 龙源(北京)风电工程技术有限公司 | Intelligent diagnosis method and device for shaft temperature abnormity of wind turbine generator gearbox |
WO2022261805A1 (en) * | 2021-06-15 | 2022-12-22 | 大连理工大学 | Diesel engine gearbox fault diagnosis method |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20120027733A (en) * | 2010-09-13 | 2012-03-22 | 한국수력원자력 주식회사 | Rotating machinery fault diagnostic method and system using support vector machines |
CN102944416A (en) * | 2012-12-06 | 2013-02-27 | 南京匹瑞电气科技有限公司 | Multi-sensor signal fusion technology-based fault diagnosis method for wind turbine blades |
-
2017
- 2017-04-11 CN CN201710232999.6A patent/CN107036808B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20120027733A (en) * | 2010-09-13 | 2012-03-22 | 한국수력원자력 주식회사 | Rotating machinery fault diagnostic method and system using support vector machines |
CN102944416A (en) * | 2012-12-06 | 2013-02-27 | 南京匹瑞电气科技有限公司 | Multi-sensor signal fusion technology-based fault diagnosis method for wind turbine blades |
Non-Patent Citations (2)
Title |
---|
HU CHUN-ZHI等: "On the use of EEMD and SVM based approach for bearing fault diagnosis of wind turbine gearbox", 《 2016 CHINESE CONTROL AND DECISION CONFERENCE》 * |
TING-FAN WU等: "Probability Estimates for Multi-class Classification by Pairwise Coupling", 《JOURNAL OF MACHINE LEARNING RESEARCH》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107830996A (en) * | 2017-10-10 | 2018-03-23 | 南京航空航天大学 | A kind of vehicle rudder diagnosis method for system fault |
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 |
CN108872781A (en) * | 2018-05-08 | 2018-11-23 | 广东昊阳电力建设有限公司 | Analysis method and device based on electric power facility intelligent patrol detection |
CN108872781B (en) * | 2018-05-08 | 2020-11-06 | 广东昊阳电力建设有限公司 | Analysis method and device based on intelligent inspection of electric power facilities |
CN112696481A (en) * | 2020-12-11 | 2021-04-23 | 龙源(北京)风电工程技术有限公司 | Intelligent diagnosis method and device for shaft temperature abnormity of wind turbine generator gearbox |
WO2022261805A1 (en) * | 2021-06-15 | 2022-12-22 | 大连理工大学 | Diesel engine gearbox fault diagnosis method |
Also Published As
Publication number | Publication date |
---|---|
CN107036808B (en) | 2019-04-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105973594B (en) | A kind of rolling bearing fault Forecasting Methodology based on continuous depth confidence network | |
CN107036808B (en) | Gearbox of wind turbine combined failure diagnostic method based on support vector machines probability Estimation | |
Han et al. | Intelligent fault diagnosis method for rotating machinery via dictionary learning and sparse representation-based classification | |
CN109738776A (en) | Fan converter open-circuit fault recognition methods based on LSTM | |
CN104200396B (en) | A kind of wind turbine component fault early warning method | |
Cao et al. | Fault diagnosis of wind turbine gearbox based on deep bi-directional long short-term memory under time-varying non-stationary operating conditions | |
CN106769049A (en) | A kind of Fault Diagnosis of Roller Bearings based on Laplce's score value and SVMs | |
CN105846780A (en) | Decision tree model-based photovoltaic assembly fault diagnosis method | |
CN111878322B (en) | Wind power generator device | |
CN104407273B (en) | Electric energy quality disturbance source positioning method considering monitoring reliability | |
CN108510147B (en) | Electric energy quality comprehensive evaluation method based on residual error fluctuation model | |
CN106650122A (en) | Equipment variable working condition operation risk evaluation method | |
CN105065212A (en) | Checking method and system of wind generation sets of wind power plant | |
CN109655266A (en) | A kind of Wind turbines Method for Bearing Fault Diagnosis based on AVMD and spectral coherence analysis | |
CN113255458A (en) | Bearing fault diagnosis method based on multi-view associated feature learning | |
CN115453356B (en) | Power equipment operation state monitoring and analyzing method, system, terminal and medium | |
CN113408068A (en) | Random forest classification machine pump fault diagnosis method and device | |
CN105157821A (en) | Rotary machinery vibration fault diagnosis and quantitative analysis method | |
CN116625686A (en) | On-line diagnosis method for bearing faults of aero-engine | |
CN116306139A (en) | Intelligent monitoring method and system for service life of wind turbine blade | |
CN114897138A (en) | System fault diagnosis method based on attention mechanism and depth residual error network | |
CN112729825A (en) | Method for constructing bearing fault diagnosis model based on convolution cyclic neural network | |
CN114320773B (en) | Wind turbine generator system fault early warning method based on power curve analysis and neural network | |
CN106446503A (en) | Method for identifying time-varying working mode of auto-covariance matrix recursive principal component analysis with forgetting factor | |
CN114266013B (en) | Transmission system vibration decoupling method based on deep learning virtual perception network |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |